David Primer

WEBVTT
1
00:00:00.600 --> 00:00:05.000
Greetings and welcome to the United States Transhumanist Party Virtual
2
00:00:05.120 --> 00:00:08.759
Enlightenment Salon. My name is Jannati stolier Off the second
3
00:00:09.000 --> 00:00:12.160
and I am the Chairman of the US Transhumanist Party.
4
00:00:12.480 --> 00:00:15.519
Here we hold conversations with some of the world's leading
5
00:00:15.560 --> 00:00:21.480
thinkers in longevity, science, technology, philosophy and politics. Like the
6
00:00:21.480 --> 00:00:24.480
philosophers of the Age of Enlightenment, we aim to connect
7
00:00:24.559 --> 00:00:27.839
every field of human endeavor and arrive at new insights
8
00:00:27.960 --> 00:00:32.320
to achieve longer lives, greater rationality, and the progress of
9
00:00:32.320 --> 00:00:37.280
our civilization. Greetings, ladies and gentlemen, and welcome to our
10
00:00:37.399 --> 00:00:42.119
US Transhumanist Party Virtual Enlightenment Salon. Today is Sunday, December
11
00:00:42.159 --> 00:00:46.000
twenty eighth, twenty twenty five, so this will be our
12
00:00:46.119 --> 00:00:50.000
final salon of this year, though we will be back
13
00:00:50.119 --> 00:00:54.039
next week as well. And we have been streaming our
14
00:00:54.119 --> 00:00:58.240
Virtual Enlightenment Salons for over five and a half years now,
15
00:00:58.280 --> 00:01:01.960
so thank you to our dedicated viewers for being with us,
16
00:01:02.359 --> 00:01:05.719
and today we bring you another fascinating conversation. We are
17
00:01:05.840 --> 00:01:11.200
joined by our esteemed US Transhumanist Party officers, Art Ramon Garcia,
18
00:01:11.239 --> 00:01:15.159
our director of Visual Art, and Daniel Tweet, our twenty
19
00:01:15.200 --> 00:01:19.239
twenty four US Vice presidential candidate and Director of Community
20
00:01:19.239 --> 00:01:23.200
and Citizen Science, and our special guest today is someone
21
00:01:23.280 --> 00:01:26.040
I have known for a long time, David Primer. He
22
00:01:26.159 --> 00:01:31.000
is the founder of Strataquant, a boutique data technology consulting
23
00:01:31.040 --> 00:01:36.239
company based in Chicago, Illinois. He is a certified professional
24
00:01:36.239 --> 00:01:39.640
in business intelligence and a Lean six Sigma black belt.
25
00:01:40.079 --> 00:01:44.200
He brings the discipline scientific approach he practiced performing cancer
26
00:01:44.280 --> 00:01:48.200
research to data analytics, and he received his Bachelor of
27
00:01:48.239 --> 00:01:52.239
Science degree in molecular biology from the University of Illinois
28
00:01:52.400 --> 00:01:57.079
at Urbana Champaign and his master's degree in cancer biology
29
00:01:57.439 --> 00:02:02.079
from the University of Wisconsin Madison. So he really brings
30
00:02:02.159 --> 00:02:07.159
quite a powerful combination of skill sets with his expertise
31
00:02:07.319 --> 00:02:11.960
in both biology and data science. I will also mention
32
00:02:12.400 --> 00:02:18.719
his website, strataquant dot com is a wealth of information
33
00:02:19.000 --> 00:02:24.039
and there he has episodes of his podcast called Profit
34
00:02:24.479 --> 00:02:27.319
is a Science. So if you want to learn more
35
00:02:27.360 --> 00:02:34.199
about the scientific approach for generating profits for your enterprise,
36
00:02:34.639 --> 00:02:37.840
then it is well worth a listen. I have listened
37
00:02:37.840 --> 00:02:40.479
to all of the episodes of Profit as a Science
38
00:02:40.520 --> 00:02:44.879
thus far, and I look forward to listening to more.
39
00:02:45.120 --> 00:02:51.120
But David welcome, and actually we know each other now
40
00:02:51.280 --> 00:02:54.680
for wow, twenty four years.
41
00:02:54.840 --> 00:02:55.560
Time flies.
42
00:02:56.080 --> 00:02:57.919
Oh wow, I was going to say sixteen, but you're right,
43
00:02:57.960 --> 00:02:59.159
I was off by a tens digit.
44
00:02:59.439 --> 00:03:00.319
Yes, each other.
45
00:03:00.360 --> 00:03:02.960
We met each other when we were fourteen. Yes, so
46
00:03:03.840 --> 00:03:05.680
it's been quite some time, so I'm sure I know that.
47
00:03:05.719 --> 00:03:05.840
Well.
48
00:03:05.879 --> 00:03:07.479
First of all, thank you so much for having me.
49
00:03:07.800 --> 00:03:11.759
Really honored to be a guest. This is a very
50
00:03:11.840 --> 00:03:16.000
long running interaction you have with your community, and I
51
00:03:16.039 --> 00:03:19.520
remember you speaking about these you know, I listened to
52
00:03:19.560 --> 00:03:21.879
your video at the beginning, and the passion that you
53
00:03:21.879 --> 00:03:24.879
show for this topic you showed when we met over
54
00:03:24.919 --> 00:03:27.159
twenty five years ago. So it's fantastic to see you
55
00:03:27.479 --> 00:03:31.520
continuing that. And I'd be happy to field, you know,
56
00:03:31.599 --> 00:03:34.199
after him through the talk, any questions that anybody has
57
00:03:34.400 --> 00:03:37.680
about what mister stoyerf was like as a fourteen fifteen
58
00:03:37.759 --> 00:03:41.400
year old, I'm more than happy to respond. Spoiler alert,
59
00:03:41.639 --> 00:03:46.120
he wasn't that different, not because he was basically a
60
00:03:46.159 --> 00:03:47.000
man at fourteen.
61
00:03:47.319 --> 00:03:48.919
I was also a spatial hair then.
62
00:03:49.240 --> 00:03:52.439
I'm jealous of the spacial hair now not much has changed,
63
00:03:52.639 --> 00:03:55.319
but I'd be happy to answer any questions that anybody has.
64
00:03:55.840 --> 00:03:59.319
All right, Well, if you have any questions on that,
65
00:03:59.479 --> 00:04:04.120
please feel free to post them in our chat. And
66
00:04:04.400 --> 00:04:09.759
in the meantime, David has a presentation prepared for us,
67
00:04:10.039 --> 00:04:15.159
so we will elevate it once it is ready to go.
68
00:04:15.759 --> 00:04:19.800
And I will say again for those of you who
69
00:04:19.879 --> 00:04:24.160
are new to data science, this presentation will teach you
70
00:04:24.240 --> 00:04:28.839
a lot. It will teach you some fundamental principles of
71
00:04:29.480 --> 00:04:36.240
looking at data and how medical analytics teaches us to
72
00:04:36.879 --> 00:04:41.399
be skeptical about certain narratives that may actually not be
73
00:04:41.519 --> 00:04:45.279
born out by the empirical facts. So David, thank you
74
00:04:45.439 --> 00:04:48.079
very much and we look forward to hearing what you
75
00:04:48.160 --> 00:04:48.680
have to say.
76
00:04:49.360 --> 00:04:49.839
Wonderful.
77
00:04:50.160 --> 00:04:52.519
So, before I really dive into this, I just I
78
00:04:52.519 --> 00:04:54.160
don't want you to be nervous, as want to mention
79
00:04:54.199 --> 00:04:56.079
a couple of things. First of all, I'm not going
80
00:04:56.079 --> 00:05:01.120
to assume that you know anything deeply about either medicine
81
00:05:01.600 --> 00:05:05.240
or about data science. I'm really trying to keep this
82
00:05:05.319 --> 00:05:10.560
at a level where we take some lessons that come
83
00:05:10.600 --> 00:05:13.680
from the medical community about how they approach analytics and
84
00:05:14.199 --> 00:05:17.000
allow those lessons to be applied to your daily life.
85
00:05:17.079 --> 00:05:20.720
Right from the Transhumanist Party is all about improving your
86
00:05:20.759 --> 00:05:23.639
quality of life, getting yourself to a place of longevity
87
00:05:23.879 --> 00:05:26.480
and just keep it at a very high level. Analytics
88
00:05:26.600 --> 00:05:29.240
is about the pursuit of truth and for us to
89
00:05:29.319 --> 00:05:32.120
extend our lives and to make those lives higher quality.
90
00:05:32.240 --> 00:05:34.240
We want to be operating a truth if we have
91
00:05:34.360 --> 00:05:37.439
better understanding as to what improves our physical health, about
92
00:05:37.439 --> 00:05:41.959
what improves our relationships, what improves our financial health, our businesses,
93
00:05:42.240 --> 00:05:45.519
our society. We want to have access to truth so
94
00:05:45.600 --> 00:05:47.879
we can implement those changes and make a positive impact.
95
00:05:47.920 --> 00:05:50.120
So don't be worried that I'm going to go deeply
96
00:05:50.160 --> 00:05:52.079
into biochemical pathways.
97
00:05:52.240 --> 00:05:52.600
I won't.
98
00:05:53.040 --> 00:05:54.639
Don't be worried that I'm going to be talking about
99
00:05:54.680 --> 00:05:57.000
the specifics of how the code and machine learning algorithm.
100
00:05:57.160 --> 00:05:57.759
I will not be.
101
00:05:58.040 --> 00:05:59.759
This is going to be focused on the lessons that
102
00:06:00.040 --> 00:06:03.040
and be applied to your daily life. So with that
103
00:06:03.120 --> 00:06:05.879
being said, I'm passionate about this topic. I spoke with it.
104
00:06:06.120 --> 00:06:08.240
I spoke with Tanadi about it when we saw each
105
00:06:08.279 --> 00:06:10.839
other a few months ago for our high school reunion,
106
00:06:11.120 --> 00:06:13.560
and I think that's what really spurred this conversation because
107
00:06:13.600 --> 00:06:15.199
I'm so passionate about it, and he said that he
108
00:06:15.240 --> 00:06:19.199
had a great medium for me to discuss it. With
109
00:06:19.240 --> 00:06:22.519
you and hopefully could provide you with some failure. So specifically,
110
00:06:22.839 --> 00:06:26.079
I'm passionate about this concept of narratives and how narratives
111
00:06:26.160 --> 00:06:30.000
can negatively impact how we go about making decisions in
112
00:06:30.040 --> 00:06:32.240
our daily lives, and that includes in business, not just
113
00:06:32.279 --> 00:06:36.000
in our personal lives. And medical analytics provides us with
114
00:06:36.759 --> 00:06:40.120
defenses we can use against those narratives. So that's what
115
00:06:40.360 --> 00:06:42.759
this conversation is going to be about. So the first
116
00:06:42.839 --> 00:06:45.399
kind of concept I want to introduce to you is
117
00:06:45.439 --> 00:06:49.639
that narratives are harmful. And before I can really justify that,
118
00:06:49.680 --> 00:06:52.000
we have to define what a narrative is. So if
119
00:06:52.040 --> 00:06:54.079
you plug in narrative into Google, you're going to see
120
00:06:54.079 --> 00:06:56.160
that it says that narrative is a spoken or written
121
00:06:56.160 --> 00:07:00.560
account of connected events or a story. Now I'm going
122
00:07:00.600 --> 00:07:02.480
to frame it in a little bit of a different
123
00:07:02.480 --> 00:07:05.519
way in the context of analytics. I'm saying a narrative
124
00:07:05.600 --> 00:07:09.680
is a statement that is framed as truth. So it's
125
00:07:09.680 --> 00:07:11.720
not really framed as a question. It's not even framed
126
00:07:11.759 --> 00:07:14.600
as a hypothesis. It's framed as this is the way
127
00:07:14.639 --> 00:07:17.959
that things are. It's got an error of authority around it.
128
00:07:17.480 --> 00:07:22.240
It is typically generalization and it's typically an untested hypothesis.
129
00:07:22.240 --> 00:07:24.519
So if I say something like it's forty five degrees outside,
130
00:07:24.560 --> 00:07:26.800
that's not a narrative. That's more of a fact. Now
131
00:07:26.800 --> 00:07:29.000
maybe you can dispute that, but it's a data point.
132
00:07:29.199 --> 00:07:32.319
Whereas if I say something like it's too cold to
133
00:07:32.360 --> 00:07:34.399
do any thing, or it's really cold, well, that's not
134
00:07:34.439 --> 00:07:38.040
really a data point. You can be more specific about it,
135
00:07:38.079 --> 00:07:41.279
because the word cold is somewhat subjective. It's kind of
136
00:07:42.000 --> 00:07:44.360
it's kind of general, right, it encompasses a large range
137
00:07:44.399 --> 00:07:47.279
of temperatures, so that's more of a narrative. So with
138
00:07:47.399 --> 00:07:50.240
that in mind, I want you to ask yourself, when's
139
00:07:50.279 --> 00:07:53.199
the last time you heard a statement that met these criteria,
140
00:07:53.439 --> 00:07:55.759
A statement that was said with an air of authority
141
00:07:57.040 --> 00:07:59.839
instead of providing specific details.
142
00:07:59.480 --> 00:08:01.480
It was kind of a generality.
143
00:08:02.360 --> 00:08:04.160
I'm pulling a fast room on you, because last time
144
00:08:04.160 --> 00:08:06.000
we hear a statement like that is when I told
145
00:08:06.079 --> 00:08:11.000
you that narratives were harmful. It's a generalization, it's not specific.
146
00:08:11.560 --> 00:08:14.600
I'm telling you a story. This is the bad guy, okay,
147
00:08:14.920 --> 00:08:18.720
because that's not really the best way to put it.
148
00:08:18.759 --> 00:08:21.800
So I'm not nullifying my entire talk here. I am
149
00:08:21.879 --> 00:08:24.560
saying there are better ways to phrase a statement like this,
150
00:08:24.639 --> 00:08:26.639
And I'm going to kind of acknowledge right now that
151
00:08:26.720 --> 00:08:30.160
narratives do have some benefits. So to sell you on this,
152
00:08:31.040 --> 00:08:33.519
I'm gonna make a couple of points. Hopefully agree with
153
00:08:33.559 --> 00:08:35.480
me on all of them. Number One, we want to
154
00:08:35.519 --> 00:08:39.399
live in a productive society. Number two, people are generally
155
00:08:39.559 --> 00:08:43.120
more productive when they are happy, which means that Number three,
156
00:08:43.679 --> 00:08:45.960
any statements or actions when in this state. You're talking
157
00:08:46.000 --> 00:08:50.080
about statements that are harmless in a vacuum, and they
158
00:08:50.200 --> 00:08:52.840
lead people to being happier are thus productive.
159
00:08:53.200 --> 00:08:54.159
So what's an example.
160
00:08:54.440 --> 00:08:57.000
If somebody's dog dies and they tell themselves that, well,
161
00:08:57.039 --> 00:08:58.279
the dog is in a better place.
162
00:08:58.759 --> 00:08:59.879
Is that a data point?
163
00:09:00.120 --> 00:09:02.440
Not really, that's a narrative, right, the dog is in
164
00:09:02.480 --> 00:09:07.039
a better place. It's not something that is really that disprovable.
165
00:09:07.320 --> 00:09:10.559
It's not a specific piece of data. But that is
166
00:09:10.600 --> 00:09:13.639
a story that serves to help that person feel better,
167
00:09:14.080 --> 00:09:17.879
because you know, dogs are important members of the family
168
00:09:18.039 --> 00:09:21.039
and when they die it can be very it.
169
00:09:20.919 --> 00:09:22.639
Can be difficult to deal with emotionally.
170
00:09:22.879 --> 00:09:25.279
So if the narrative that the dog is in a
171
00:09:25.279 --> 00:09:30.080
better place helps that person feel better, that is beneficial,
172
00:09:30.120 --> 00:09:32.720
and that narrative is helpful. The other thing I want
173
00:09:32.759 --> 00:09:35.639
to point out as to why narratives exist, and this
174
00:09:35.799 --> 00:09:39.879
is one of the most crucial aspects of our entire
175
00:09:39.919 --> 00:09:43.440
discussion today because it's rooted in human nature. It doesn't
176
00:09:43.480 --> 00:09:45.679
go away. So there's nothing going to be from this
177
00:09:45.759 --> 00:09:48.200
talk where you stop having narratives that pop up in
178
00:09:48.240 --> 00:09:51.279
your life. The issue isn't preventing them from popping up.
179
00:09:51.360 --> 00:09:54.039
It's identifying where they do pop up, because they're going
180
00:09:54.080 --> 00:09:56.919
to pop up at like when we feel hungry for food.
181
00:09:57.200 --> 00:09:59.039
It's about knowing what to do with them when they
182
00:09:59.039 --> 00:09:59.600
do pop up.
183
00:09:59.639 --> 00:10:05.720
So humans process stories better than raw data. When people
184
00:10:05.879 --> 00:10:08.679
that have the best memories in the world, So these
185
00:10:08.720 --> 00:10:12.679
are the people that can memorize pie out to three
186
00:10:12.759 --> 00:10:17.399
hundred digits, or people that can memorize sequences of cards,
187
00:10:17.440 --> 00:10:21.480
playing cards hundreds of a sequence of hundreds of playing
188
00:10:21.879 --> 00:10:24.879
cards long. What they'll all tell you is they'll all
189
00:10:24.960 --> 00:10:28.639
construct a story to help them remember the order of events.
190
00:10:28.679 --> 00:10:30.759
That's kind of how our brains are built. That's not
191
00:10:30.840 --> 00:10:32.559
a bad thing. Our brains are built in a way
192
00:10:32.559 --> 00:10:35.240
that allowed us to survive tens of thousands, if not
193
00:10:35.399 --> 00:10:38.320
hundreds of thousands of years ago, So that's the equipment
194
00:10:38.360 --> 00:10:40.440
that we still have though, so our brains are not
195
00:10:40.519 --> 00:10:43.159
built to deal with raw data. It's built to turn
196
00:10:43.240 --> 00:10:45.879
that raw data into a story, so it's something that
197
00:10:45.919 --> 00:10:49.720
we can remember. The issue is that that equipment that
198
00:10:49.759 --> 00:10:54.600
we have in our head was not built really for analytics.
199
00:10:54.799 --> 00:10:56.440
It was built to get us to an age where
200
00:10:56.480 --> 00:11:00.799
we can have offspring and raise those off so they
201
00:11:00.799 --> 00:11:02.039
can be self sufficient, and.
202
00:11:02.000 --> 00:11:03.279
Then we're not really needed anymore.
203
00:11:03.519 --> 00:11:06.759
Right, so we can do better than that, And that's
204
00:11:06.799 --> 00:11:09.159
why with the equipment that we have, we have to
205
00:11:09.159 --> 00:11:13.240
be aware of what narratives, how they can be hurting us,
206
00:11:13.519 --> 00:11:16.559
so we know that we operate better when we told
207
00:11:16.559 --> 00:11:17.000
the story.
208
00:11:17.320 --> 00:11:18.679
Narratives are also built for.
209
00:11:18.759 --> 00:11:21.000
Unique outcomes, so when you see a situation you've never
210
00:11:21.039 --> 00:11:23.240
seen before, it helps to have a story around it.
211
00:11:24.159 --> 00:11:28.399
And this last point kind of puts those last two together. Humans,
212
00:11:28.440 --> 00:11:31.480
we have a need and it's a primal instinct to
213
00:11:31.600 --> 00:11:35.200
have an explanation, even if it's not true. We do
214
00:11:35.320 --> 00:11:38.200
not do well with the idea that we don't know
215
00:11:38.360 --> 00:11:41.200
something and we just sit in that lack of knowledge.
216
00:11:41.320 --> 00:11:43.600
It's uncomfortable. Think about the.
217
00:11:43.600 --> 00:11:46.159
Last time you were sitting in your home and you
218
00:11:46.279 --> 00:11:48.759
heard some kind of a sound and you didn't know
219
00:11:48.799 --> 00:11:54.039
exactly what it was. Chances are your brain didn't think, oh,
220
00:11:54.240 --> 00:11:57.279
I don't know what that sound was. Chances are you
221
00:11:57.320 --> 00:12:00.480
thought to yourself that was probably the ice machine or
222
00:12:00.799 --> 00:12:03.399
a code fell down in the closet. Doesn't matter if
223
00:12:03.399 --> 00:12:06.360
it was true or not, you would feel better having
224
00:12:06.720 --> 00:12:09.440
an explanation even if you didn't have data around.
225
00:12:09.480 --> 00:12:11.360
It doesn't mean you've gone up and you checked. Maybe
226
00:12:11.360 --> 00:12:12.200
you didn't need to check.
227
00:12:12.320 --> 00:12:14.360
But the point is you came up with a story
228
00:12:14.759 --> 00:12:16.919
even though you weren't didn't have any data in front
229
00:12:16.919 --> 00:12:19.039
of you, because it was better to have a story
230
00:12:19.080 --> 00:12:22.559
that was unsubstantiated than not to have story at all.
231
00:12:22.799 --> 00:12:25.840
In terms of how you're feeling about it. That is primal.
232
00:12:25.879 --> 00:12:27.559
That's always going to be the case. That's the way
233
00:12:27.600 --> 00:12:29.799
it is with me. Also, I am not a robot.
234
00:12:30.159 --> 00:12:33.120
I just choose to be aware of when these narratives
235
00:12:33.159 --> 00:12:36.840
pop up because I want to obtain truth as much
236
00:12:36.840 --> 00:12:39.639
as I possibly can. So that's something to keep in
237
00:12:39.679 --> 00:12:41.559
minds to where these narratives come from. But they do
238
00:12:41.679 --> 00:12:44.159
have some benefits. So now let's talk about analytics, and
239
00:12:44.159 --> 00:12:45.519
I'm going to keep this at a very high level.
240
00:12:45.600 --> 00:12:48.120
I one thing, mister Stoller, I've been mentioned, I do
241
00:12:48.240 --> 00:12:51.440
teach a course in for MIT about analytics, and this
242
00:12:51.480 --> 00:12:54.600
is what I tell them. I tell them that if
243
00:12:54.639 --> 00:12:57.279
you were to boil down analytics to one thing, it's
244
00:12:57.279 --> 00:13:01.480
there to predict the future. And it sounds supernatural, but
245
00:13:01.639 --> 00:13:03.480
it's really not so. I mean, let's say that you
246
00:13:03.679 --> 00:13:07.799
are working in marketing and you are looking at your
247
00:13:07.960 --> 00:13:11.440
a lot of marketing data, different marketing campaigns. You're trying
248
00:13:11.440 --> 00:13:13.679
to identify what campaign is likely going to be the
249
00:13:13.679 --> 00:13:17.159
best for a particular product. That's an analytics question, right,
250
00:13:17.360 --> 00:13:19.120
But what it's boiling down to is I'm trying to
251
00:13:19.120 --> 00:13:21.519
predict the future. I'm trying to say which one of
252
00:13:21.559 --> 00:13:25.840
these marketing campaigns will provide me with the most profit
253
00:13:26.080 --> 00:13:26.879
or the most revenue.
254
00:13:27.759 --> 00:13:29.639
So most of the times, what it.
255
00:13:29.639 --> 00:13:32.919
Boils down to, and at the core of analytics is
256
00:13:33.120 --> 00:13:34.840
actionable intelligence.
257
00:13:35.519 --> 00:13:36.519
Both of these words.
258
00:13:36.919 --> 00:13:40.639
If you're not getting actionable intelligence, analytics is not a
259
00:13:40.639 --> 00:13:44.039
good application for what you're looking to do. So by intelligence,
260
00:13:44.080 --> 00:13:48.120
I mean information as to what is likely going to happen,
261
00:13:49.159 --> 00:13:51.360
and it has to be actionable, So it can't just
262
00:13:51.399 --> 00:13:55.399
be information for information's sake. Sometimes people will say, and
263
00:13:55.440 --> 00:13:58.840
I'm giving business examples, but this can happen in everyday life.
264
00:13:59.399 --> 00:14:02.440
Our sales were down three percent last quarter. That's not
265
00:14:02.480 --> 00:14:05.320
exactly actionable intelligence. I'm not saying it's not important.
266
00:14:06.240 --> 00:14:06.960
Of excuse me.
267
00:14:07.519 --> 00:14:12.320
I'm not saying it's not actually important. I am suggesting
268
00:14:12.679 --> 00:14:18.399
that it's a precursor to actionable intelligence. Actionable intelligence would
269
00:14:18.440 --> 00:14:22.120
be something like the return on investment is higher in
270
00:14:22.200 --> 00:14:26.840
marketing campaign A than it is a marketing campaign B.
271
00:14:27.000 --> 00:14:30.360
That's actionable because now I know, okay, that's very useful.
272
00:14:30.440 --> 00:14:33.360
I now have a direct link between where my money
273
00:14:33.399 --> 00:14:37.000
should be going and my profits. Right, So it needs
274
00:14:37.000 --> 00:14:39.399
to have both of these things. Now, the actions that
275
00:14:39.440 --> 00:14:42.639
are taken need should be leading outcomes that actually matter,
276
00:14:42.720 --> 00:14:47.360
shouldn't be meaningless. So in business, that's meaning those actions
277
00:14:47.360 --> 00:14:51.159
should lead directly to higher profits or revenue or credit costs.
278
00:14:51.440 --> 00:14:53.720
In medicine, it means improved health, and in our personal
279
00:14:53.759 --> 00:14:56.120
lives that means it should be leading to a higher
280
00:14:56.200 --> 00:14:58.480
quality of life. That's when we want to be applying
281
00:14:58.519 --> 00:15:03.200
these analytics. Another aspect of analytics that are absolutely required
282
00:15:03.639 --> 00:15:08.120
are singular optimization metrics. Now that's kind of a mouthful.
283
00:15:08.200 --> 00:15:11.159
But all I mean is that any kind of model
284
00:15:11.200 --> 00:15:13.480
that is built, right and so, like, a model is
285
00:15:13.559 --> 00:15:16.639
just an equation that relates variables to one another, and
286
00:15:16.679 --> 00:15:21.360
then you can use that equation to feed in future information,
287
00:15:21.799 --> 00:15:25.960
and that future information will tell you kind of what
288
00:15:26.000 --> 00:15:27.879
you should be doing next. So, as an example, if
289
00:15:27.879 --> 00:15:29.480
I'm trying to predict what my sales are going to
290
00:15:29.480 --> 00:15:31.960
be next quarter, right, you can have a model where
291
00:15:31.960 --> 00:15:34.000
I can feed in the sales from the previous quarter
292
00:15:34.000 --> 00:15:37.559
along with other information, and then my model will tell me, okay, well,
293
00:15:37.600 --> 00:15:41.279
next quarter, my projected sales will be X number of dollars.
294
00:15:41.360 --> 00:15:42.679
Right, I'll grow by y percent.
295
00:15:42.919 --> 00:15:43.159
Okay.
296
00:15:43.600 --> 00:15:45.240
Any kind of model that is built it could be
297
00:15:45.279 --> 00:15:48.159
a simple linear aggression model, which you've probably seen before.
298
00:15:48.399 --> 00:15:50.960
Imagine a scatterplot. You're trying to draw a line that
299
00:15:51.000 --> 00:15:55.840
fits those data points as closely as possible. That's a model.
300
00:15:56.320 --> 00:15:58.480
Or if you're talking about a model that's driven by
301
00:15:58.559 --> 00:16:00.759
artificial intelligence, it could be a model that's able to
302
00:16:00.759 --> 00:16:03.360
process an image. It can be a model that is
303
00:16:03.440 --> 00:16:06.360
able to take a large block of text and determine
304
00:16:06.360 --> 00:16:08.080
whether or not the text is positive or negative, or
305
00:16:08.080 --> 00:16:14.000
summarize it it's all training their parameters by optimizing a
306
00:16:14.080 --> 00:16:17.039
single metric. Now, linear aggression might have one or two parameters,
307
00:16:17.240 --> 00:16:19.720
an AI model might have a couple of billion, but
308
00:16:19.759 --> 00:16:23.679
they still in order to train their model. To train
309
00:16:23.799 --> 00:16:26.120
any kind of model, there has to be something that
310
00:16:26.159 --> 00:16:29.480
they determine is good and something that they're determined as bad.
311
00:16:29.759 --> 00:16:31.480
And in order to have that, you need need to
312
00:16:31.519 --> 00:16:34.240
have a singular metric, a singular dimension by what you're
313
00:16:34.279 --> 00:16:37.600
operating on. Now, that doesn't mean that that metric can't
314
00:16:37.600 --> 00:16:40.639
include a lot of different things in it, right, It
315
00:16:40.679 --> 00:16:44.159
can encompass a lot of different factors, but at the
316
00:16:44.200 --> 00:16:45.399
same time, it's.
317
00:16:45.240 --> 00:16:46.320
Still one singular metric.
318
00:16:46.399 --> 00:16:48.559
So what's an example of a singular metric that can
319
00:16:48.639 --> 00:16:51.080
be complex, but it's still singular. Think about a credit score.
320
00:16:51.240 --> 00:16:53.480
Think about everything that goes to a credit score. Lots
321
00:16:53.519 --> 00:16:55.399
of different things. How many things have you paid off,
322
00:16:55.600 --> 00:16:58.159
how long ago was it that you paid them off,
323
00:16:58.840 --> 00:17:01.559
what's the balance on your credit what are There's lots
324
00:17:01.559 --> 00:17:03.120
of things that go into it, but it does boil
325
00:17:03.159 --> 00:17:04.039
down to one number.
326
00:17:04.200 --> 00:17:05.519
Analytics works the same way.
327
00:17:05.759 --> 00:17:08.640
There's one number that any model is trying to optimize
328
00:17:08.839 --> 00:17:11.559
and then and without that clarity, as to that one
329
00:17:11.599 --> 00:17:13.920
number that it's trying to optimize, it won't be able
330
00:17:13.920 --> 00:17:17.799
to converge on a solution. So if we are operating
331
00:17:17.880 --> 00:17:21.079
in the realm of truth and not in the realm
332
00:17:21.200 --> 00:17:25.240
of emotion, we have to be clear about what we're
333
00:17:25.240 --> 00:17:28.000
trying to predict, otherwise we won't be able to converge
334
00:17:28.079 --> 00:17:30.200
on a solution. So that's a really important point.
335
00:17:30.240 --> 00:17:30.400
Now.
336
00:17:30.799 --> 00:17:32.279
The reason that I'm bringing that up is now I
337
00:17:32.319 --> 00:17:35.480
want to kind of put together narratives and analytics now
338
00:17:35.480 --> 00:17:37.759
that we've talked about both of them. So this is
339
00:17:37.759 --> 00:17:39.920
even diagram I mean that kind of tongue in cheek
340
00:17:40.319 --> 00:17:43.039
than diagrams. Typically the circles overlap. What I'm pointing out
341
00:17:43.079 --> 00:17:45.160
here is that the circles don't really overlap. Now I
342
00:17:45.200 --> 00:17:48.240
will get into in a couple of minutes how narratives
343
00:17:48.480 --> 00:17:51.440
can help analytics, but it's more like those circles are touching.
344
00:17:51.480 --> 00:17:55.799
They don't really overlap themselves. So, as we've discussed, narratives
345
00:17:55.839 --> 00:17:59.039
are very general. When I say something like narratives are harmful,
346
00:17:59.240 --> 00:18:02.359
that it's keeping at a very high level. Narratives they
347
00:18:02.400 --> 00:18:04.720
do function that you provide a benefit, But in the
348
00:18:04.839 --> 00:18:07.960
realm of emotion. Now, whenever I say that, I am
349
00:18:08.000 --> 00:18:10.359
not trying to say that human emotion doesn't matter.
350
00:18:10.400 --> 00:18:11.680
I'm not saying that by any stretch.
351
00:18:11.680 --> 00:18:14.680
I'm saying if anything, Early on I said human emotion
352
00:18:14.759 --> 00:18:17.119
is very important. If we are happy, we're more likely
353
00:18:17.160 --> 00:18:20.480
to be productive. We want people to be happy. Just
354
00:18:20.519 --> 00:18:23.359
stating as a statement of fact that if you're operating
355
00:18:23.359 --> 00:18:26.200
in the field of emotion, narratives can be helpful. But
356
00:18:26.240 --> 00:18:29.400
if you're operating in the field of actionable intelligence trying
357
00:18:29.400 --> 00:18:33.640
to obtain truth, then narratives should not be playing a
358
00:18:33.720 --> 00:18:37.039
role when it comes to having conclusive statements. And finally,
359
00:18:37.400 --> 00:18:42.200
narratives are generally inspired by unique events. If one of
360
00:18:42.200 --> 00:18:44.480
your friends goes to a restaurant and they get food
361
00:18:44.480 --> 00:18:46.400
poisoning and they come back to tell you about it,
362
00:18:46.559 --> 00:18:49.759
they're probably not coming back to tell you this restaurant
363
00:18:50.400 --> 00:18:51.039
is more.
364
00:18:50.960 --> 00:18:53.000
Likely going to give somebody food poisoning.
365
00:18:53.119 --> 00:18:55.519
They're going to tell you all about that one data
366
00:18:55.519 --> 00:18:58.200
point that they experienced. That doesn't mean their data point
367
00:18:58.359 --> 00:19:02.799
is useless, meaning lists. It does mean, however, that it
368
00:19:02.839 --> 00:19:04.799
is only one data point, so they can try to
369
00:19:04.839 --> 00:19:07.039
craft a story around it. But the fact is it's
370
00:19:07.079 --> 00:19:09.359
probably not going to have enough data to back it up. Now,
371
00:19:09.519 --> 00:19:11.160
I would understand that person saying, you know what, I
372
00:19:11.200 --> 00:19:14.839
don't feel comfortable going to that restaurant again. But maybe
373
00:19:14.839 --> 00:19:16.640
they were sick before they went, Maybe they got food
374
00:19:16.640 --> 00:19:18.720
poisoning or something they ate before or after going to
375
00:19:18.759 --> 00:19:25.400
the restaurant. Maybe maybe the chef that was cooking there
376
00:19:25.480 --> 00:19:27.960
got fired and now the new chef will be completely fine.
377
00:19:28.039 --> 00:19:30.400
The point is they're inspired by unique events, whereas in
378
00:19:30.480 --> 00:19:34.400
analytics it's very specific. We're looking at one very specific
379
00:19:34.480 --> 00:19:38.079
optimization metric when we're looking at developing a model. It
380
00:19:38.119 --> 00:19:41.119
also functions in the realm of actual intelligence. We're trying
381
00:19:41.160 --> 00:19:43.680
to get to truth. And finally, analytics are fed by
382
00:19:43.680 --> 00:19:46.640
sample size. No decent model is trained on a small
383
00:19:46.640 --> 00:19:48.400
sample size to try to feed it as large sample
384
00:19:48.440 --> 00:19:50.680
size as possible. If you think about any model that
385
00:19:50.720 --> 00:19:53.519
you really enjoy using, so like in today's day and age,
386
00:19:53.519 --> 00:19:55.720
as things like chat, CHPT or some of these large
387
00:19:55.759 --> 00:19:59.000
language models, they were fed by an enormous corpus of
388
00:19:59.039 --> 00:20:02.559
information literally billions and billions, if not trillions, of data
389
00:20:02.599 --> 00:20:05.920
points that helped feed what it was doing.
390
00:20:06.039 --> 00:20:08.119
So these things are very very different from each other.
391
00:20:08.160 --> 00:20:12.200
So it's important that we identify in which realm we're operating,
392
00:20:12.240 --> 00:20:14.319
so we know whether we should be leaning on narratives
393
00:20:14.400 --> 00:20:15.799
or whether we should be.
394
00:20:15.799 --> 00:20:16.960
Leaning on analytics.
395
00:20:18.039 --> 00:20:20.400
So I want to go over some sample narratives as
396
00:20:21.519 --> 00:20:22.559
I think they're interesting.
397
00:20:22.599 --> 00:20:25.039
It kind of points out where they're faulty.
398
00:20:25.359 --> 00:20:31.119
So the first example, as Jannati mentioned, I live I
399
00:20:31.119 --> 00:20:33.839
don't live downtown right now. I used to live downtown Chicago,
400
00:20:34.200 --> 00:20:37.960
and oftentimes I met with some of my wife's family
401
00:20:38.000 --> 00:20:40.920
that previously lived in Chicago and then moved to the suburbs,
402
00:20:41.079 --> 00:20:43.440
and they would tell me things have gotten really bad
403
00:20:43.480 --> 00:20:44.039
in Chicago.
404
00:20:44.359 --> 00:20:44.599
Now.
405
00:20:44.640 --> 00:20:47.200
I'm very passionate about narratives. I have been for a while.
406
00:20:47.519 --> 00:20:49.920
So when I heard this, I thought to myself, hmm,
407
00:20:50.519 --> 00:20:53.960
that's a narrative. It's a general statement. They're not asking
408
00:20:54.000 --> 00:20:56.759
a question, they're not really citing a specific piece of data.
409
00:20:57.319 --> 00:21:00.000
It's a reasonable statement, but it's not I didn't think
410
00:21:00.119 --> 00:21:02.200
was appropriate yet to have it give it an air
411
00:21:02.240 --> 00:21:04.640
of authority. So I said, well, I wonder if things
412
00:21:04.640 --> 00:21:07.880
have gotten really bad in Chicago. So I did some research,
413
00:21:08.440 --> 00:21:10.359
and what you're seeing here is a graph of the
414
00:21:10.440 --> 00:21:13.359
number of violent crimes committed between twenty twenty twenty. You
415
00:21:13.359 --> 00:21:15.160
could ignore the New York part of it. I'm only
416
00:21:15.200 --> 00:21:18.680
speaking to Chicago. I'm not even making the point here
417
00:21:18.799 --> 00:21:21.839
that things are that are they're better now thany were before.
418
00:21:22.640 --> 00:21:24.960
I'm also not making the point, to be clear, that
419
00:21:25.039 --> 00:21:27.599
things aren't bad in Chicago. This graph is showing you
420
00:21:27.640 --> 00:21:30.559
that there's over twenty I think twenty five thousand violent
421
00:21:30.640 --> 00:21:31.880
crimes per year in Chicago.
422
00:21:32.160 --> 00:21:34.759
That's far too many. There's no question that things are
423
00:21:34.759 --> 00:21:35.480
bad in Chicago.
424
00:21:35.960 --> 00:21:38.480
Where I would push back is the implication that things
425
00:21:38.480 --> 00:21:41.000
have gotten bad in Chicago. That implies that things were
426
00:21:41.039 --> 00:21:44.359
really good before and now it's a lot worse. The
427
00:21:44.440 --> 00:21:47.160
data doesn't really justify that. I couldn't really find any
428
00:21:47.240 --> 00:21:50.359
data that showed that things were really peaceful, calm, and
429
00:21:50.400 --> 00:21:54.440
wonderful in Chicago before and when this person lived there,
430
00:21:54.720 --> 00:21:57.440
and since they've moved back to the Chicago to the suburbs,
431
00:21:57.519 --> 00:21:58.839
things have gotten so much worse.
432
00:21:58.839 --> 00:22:00.599
So it doesn't justify going in there anymore.
433
00:22:01.480 --> 00:22:04.079
Now. That's not really what I found. It's been bad
434
00:22:04.119 --> 00:22:07.039
the whole time. So the narrative that which is really
435
00:22:07.039 --> 00:22:10.319
what this is implying is that you shouldn't this person
436
00:22:10.359 --> 00:22:12.359
shouldn't be going back into the city now because it's
437
00:22:12.359 --> 00:22:14.960
worse than it was before. This to make a big difference.
438
00:22:15.000 --> 00:22:17.160
I think Chicago's a fantastic, world class city. And now
439
00:22:17.119 --> 00:22:18.759
I would say, all right, listen, if you were comfortable
440
00:22:18.799 --> 00:22:20.839
being in the city before, and you found that your
441
00:22:20.880 --> 00:22:25.640
experience was, you enjoyed it, you knew where to go,
442
00:22:26.519 --> 00:22:28.799
you knew where not to go, more importantly, and when
443
00:22:28.880 --> 00:22:32.000
not to go there, and so you were safe, Well,
444
00:22:32.039 --> 00:22:33.599
then you should be able to do that now because
445
00:22:33.640 --> 00:22:36.359
things are not worse. So and notice how different this
446
00:22:36.400 --> 00:22:40.480
would be. Instead of saying things have gotten bad in Chicago,
447
00:22:40.599 --> 00:22:43.359
the question was have things gotten bad in Chicago?
448
00:22:43.799 --> 00:22:43.960
Now?
449
00:22:44.000 --> 00:22:47.039
It's opening up to curiosity and digging into data, the
450
00:22:47.119 --> 00:22:50.200
other narrative I wanted to get into. And I promise
451
00:22:50.240 --> 00:22:52.759
you this is not going to get political. Is we
452
00:22:52.839 --> 00:22:54.400
hear this every two years? For four years?
453
00:22:54.480 --> 00:22:56.240
Our country is falling apart? Right?
454
00:22:56.240 --> 00:22:58.480
This is something that we'll hear every single election cycle
455
00:22:58.799 --> 00:23:01.799
from generally party that's not in power. Sometimes it's the
456
00:23:01.799 --> 00:23:04.559
party that's in power, but they'll blame the other party. Well,
457
00:23:04.599 --> 00:23:06.640
first of all, does this statement make us feel good?
458
00:23:06.920 --> 00:23:09.759
Probably not, So it definitely doesn't operate. If we're just saying, hey,
459
00:23:09.799 --> 00:23:11.839
it's just about emotions, well, doesn't help us feel good.
460
00:23:12.200 --> 00:23:14.880
Is this really intelligence? Are we working with a fact here?
461
00:23:14.960 --> 00:23:17.839
Our country and by our country and some people might
462
00:23:17.880 --> 00:23:21.079
be I saw somebody who's listening from Bogata in Columbia.
463
00:23:21.519 --> 00:23:24.640
Regardless of what country you live in, any country has
464
00:23:24.680 --> 00:23:27.880
many many people. Is very very complex. So to say
465
00:23:27.920 --> 00:23:32.400
something's falling apart talk about a generality, there's lots of
466
00:23:32.400 --> 00:23:35.440
different buckets you could be speaking to. We don't really
467
00:23:35.480 --> 00:23:38.799
know what that means. It's kind of hard to disprove
468
00:23:38.839 --> 00:23:42.160
that statement. So it's not really intelligence. And is this
469
00:23:42.200 --> 00:23:45.039
really actionable to a single citizen? Even if I give
470
00:23:45.039 --> 00:23:47.599
you the benefit of the doubt, and I'd be idealistic
471
00:23:47.640 --> 00:23:50.079
to say yes, a single citizen can change the country.
472
00:23:50.359 --> 00:23:52.680
Is there enough evidence to show that the reason that
473
00:23:52.720 --> 00:23:55.640
it's falling apart is because of the other party?
474
00:23:56.000 --> 00:23:56.160
Right?
475
00:23:56.519 --> 00:23:58.160
Whatever party is saying this, they're saying, well, if you
476
00:23:58.240 --> 00:24:01.640
vote for the other party's going to continue to fall apart,
477
00:24:01.759 --> 00:24:03.559
and if you vote for our party then it won't.
478
00:24:03.920 --> 00:24:06.519
Is there really enough evidence to show that? Probably not,
479
00:24:06.839 --> 00:24:08.680
And I'm not again, that's not a political statement. We're
480
00:24:08.720 --> 00:24:10.720
just live in a very complex society. It'll be very
481
00:24:10.759 --> 00:24:15.559
extremely difficult to directly tie political decisions to the general
482
00:24:16.519 --> 00:24:19.079
state of the country. So this is not a helpful narrative.
483
00:24:19.200 --> 00:24:20.720
And the last one that I'm going to speak to
484
00:24:22.599 --> 00:24:25.240
that I think that these are kind of funny. When
485
00:24:25.240 --> 00:24:26.839
you ar a statement that we are in a bull market.
486
00:24:27.039 --> 00:24:30.279
I always think it's funny because this is operating specifically
487
00:24:30.319 --> 00:24:32.640
in the field of finance. Right, So in finance, we're
488
00:24:32.680 --> 00:24:34.400
trying to predict the future because if you're able to
489
00:24:34.400 --> 00:24:36.680
predict the future, you'll make lots and lots of money.
490
00:24:36.960 --> 00:24:39.519
So when someone says we are in a bull market,
491
00:24:41.039 --> 00:24:42.880
it is kind of a statement of facts. So if
492
00:24:42.880 --> 00:24:47.480
you're not familiar, there are technical definitions as to what
493
00:24:47.559 --> 00:24:49.920
a bear market is and what a bull market is.
494
00:24:50.000 --> 00:24:51.599
Now this kind of speaks to the third point I
495
00:24:51.640 --> 00:24:54.759
have here. So these are discreet category. So there is
496
00:24:54.839 --> 00:24:58.279
a day where like where you will see if you
497
00:24:58.279 --> 00:25:00.359
want to see NBC dot Com, it will say we
498
00:25:00.599 --> 00:25:03.000
just moved from a bullmarket into a bear market or
499
00:25:03.039 --> 00:25:04.799
a vice versa. It has to do with the crossing
500
00:25:04.880 --> 00:25:08.960
of certain moving averages. Now, what's kind of funny about
501
00:25:08.960 --> 00:25:11.839
that is that that means that there is a very
502
00:25:11.880 --> 00:25:14.279
specific line in the sand that we go from a
503
00:25:14.279 --> 00:25:16.920
bear market into a bull market. So and bear market
504
00:25:16.960 --> 00:25:19.519
in general is saying, oh my god, the market's falling down.
505
00:25:19.920 --> 00:25:23.200
The whole sentiment of everybody is terrible. People are afraid
506
00:25:23.240 --> 00:25:27.079
to invest. This is a really difficult time economically, and
507
00:25:27.119 --> 00:25:28.559
the bull market is the opposite.
508
00:25:28.640 --> 00:25:31.920
So I find it humorous that there.
509
00:25:31.839 --> 00:25:36.200
Is a single penny that exists that if the stock
510
00:25:36.279 --> 00:25:39.200
market goes above that penny, that means that everything's great,
511
00:25:39.400 --> 00:25:41.480
and if the market goes below that penny, that means
512
00:25:41.480 --> 00:25:45.559
that everything's terrible. That doesn't really make sense as a way.
513
00:25:45.599 --> 00:25:49.160
It doesn't make sense to discreetly categorize the state of
514
00:25:49.200 --> 00:25:52.400
the economy into two buckets based off of a single penny,
515
00:25:52.400 --> 00:25:55.680
which is what this model currently does number one and
516
00:25:55.759 --> 00:25:58.440
number two. Does this really help us predict the future?
517
00:25:58.440 --> 00:26:00.400
So the implication that we're in a bull market, it's
518
00:26:00.400 --> 00:26:03.920
saying that things have been going well. Like I can
519
00:26:03.960 --> 00:26:05.720
treat it as a statement of fact, like I said,
520
00:26:05.759 --> 00:26:09.400
because it is a technical definition that is retrospective. So
521
00:26:09.759 --> 00:26:12.599
if those moving averages are aligning with a bull market,
522
00:26:12.599 --> 00:26:13.920
then it's a statement of fact that we're in a
523
00:26:13.920 --> 00:26:14.480
bull market.
524
00:26:14.559 --> 00:26:16.960
But what's the point of talking about that if we're
525
00:26:17.000 --> 00:26:17.880
not trying to.
526
00:26:19.400 --> 00:26:22.880
Be fed information that's going to inform our decisions moving forward.
527
00:26:23.160 --> 00:26:25.920
So if the implication is that it's a full market,
528
00:26:26.079 --> 00:26:28.920
so that means we should invest, well, doesn't that also
529
00:26:29.000 --> 00:26:30.920
mean that you're saying that if it's a bear market,
530
00:26:30.960 --> 00:26:34.079
we shouldn't be investing. And I'm pretty sure that every
531
00:26:34.119 --> 00:26:38.799
single investment magnate by that, I mean the billionaires that
532
00:26:38.880 --> 00:26:43.160
have made I mean billions of dollars invested in.
533
00:26:43.160 --> 00:26:43.920
The stock market.
534
00:26:44.119 --> 00:26:46.599
They don't say I know when to invest in the
535
00:26:47.359 --> 00:26:49.559
bull market, I know not to invest in a bear market.
536
00:26:49.799 --> 00:26:53.480
They generally say the more time you spend in the market,
537
00:26:53.559 --> 00:26:56.880
the better. It's very difficult to predict market timing. That's
538
00:26:56.920 --> 00:26:59.039
not how they made their money. They made their money
539
00:26:59.039 --> 00:27:02.680
by staying invest for longer. That includes bear markets, which
540
00:27:02.720 --> 00:27:06.279
means this information is not actually so. First of all,
541
00:27:06.319 --> 00:27:08.000
I don't like the information because it's silly to me
542
00:27:08.000 --> 00:27:10.039
that a bullmarket and a bear market could change by
543
00:27:10.079 --> 00:27:12.759
one penny. But also it's not really not that actionable
544
00:27:12.880 --> 00:27:15.000
because I think you can agree that if you didn't
545
00:27:15.039 --> 00:27:17.920
invest during a bear market by definition. You're saying you're
546
00:27:17.920 --> 00:27:20.079
not buying into the stock market when the stock market
547
00:27:20.119 --> 00:27:23.559
is lower. I'm not slamming investing in a bull market either.
548
00:27:24.559 --> 00:27:27.359
I'm not even saying to trust my opinion. Just read
549
00:27:27.400 --> 00:27:34.519
what Warren Buffett, Ray Dahalio, Jack Bogo, what the highly
550
00:27:34.559 --> 00:27:39.079
successful finance professionals have advocated for, which is just staying
551
00:27:39.119 --> 00:27:41.400
invested in the market for longer. So these are narratives
552
00:27:41.400 --> 00:27:44.279
that we hear all the time that are not helpful
553
00:27:44.440 --> 00:27:46.759
because they don't provide actionable intelligence and they don't help
554
00:27:46.839 --> 00:27:49.240
us feel better. Now, I do want to touch on
555
00:27:49.720 --> 00:27:52.640
the place where narratives can exist in the analytics process,
556
00:27:53.039 --> 00:27:56.880
and that is narratives are a great way to provide questions.
557
00:27:57.079 --> 00:28:00.559
So analytics is the process of answering those questions. And
558
00:28:00.599 --> 00:28:03.319
I'm saying narratives are harmful because they're framed as answers.
559
00:28:03.480 --> 00:28:06.160
Things have gotten bad in Chicago, our countries falling apart.
560
00:28:06.200 --> 00:28:08.480
These are questions, These are answers. This is the way
561
00:28:08.519 --> 00:28:11.799
that things are. I think if you frame narratives as questions,
562
00:28:12.039 --> 00:28:14.079
you now the interesting thing is to look into and
563
00:28:14.119 --> 00:28:15.720
if you're coming up with a narrative, that means it's
564
00:28:15.720 --> 00:28:18.960
important to you, So it's probably worth looking into if
565
00:28:19.000 --> 00:28:22.319
you're operating in the field of truth. So if somebody
566
00:28:22.480 --> 00:28:24.759
is saying, I mean, let's go back to my family member,
567
00:28:25.279 --> 00:28:27.960
right in law, that it's saying things have gotten really
568
00:28:27.960 --> 00:28:31.759
bad in Chicago, well, that means they're considering spending more
569
00:28:31.799 --> 00:28:35.880
time in the city. And the assumption that the city
570
00:28:35.920 --> 00:28:39.039
is apt, that the city is not doing well, is
571
00:28:40.200 --> 00:28:45.039
is not is preventing them from going into the city.
572
00:28:45.119 --> 00:28:48.759
That's that's harmful. It matters to them, Hey buddy, please
573
00:28:48.759 --> 00:28:58.000
get that. Sorry about that, so that it's not doing well. Sorry,
574
00:28:58.039 --> 00:29:02.400
please excuse me those myself that it's not doing well.
575
00:29:02.440 --> 00:29:04.240
So it matters. But if you frame it as a question,
576
00:29:04.519 --> 00:29:06.400
now they can do research. They can discover the same
577
00:29:06.400 --> 00:29:08.319
thing that I discovered, which is that things actually haven't
578
00:29:08.319 --> 00:29:09.960
gotten worse than it was before. So if they were
579
00:29:09.960 --> 00:29:11.880
comfortable being in the city before, they could comformle being
580
00:29:11.880 --> 00:29:12.440
in the city now.
581
00:29:12.440 --> 00:29:14.319
So it can be helpful as a question.
582
00:29:14.599 --> 00:29:16.400
So now that we've gone on a narratives, hopefully I've
583
00:29:16.400 --> 00:29:18.359
sold you on this idea that narratives that while they
584
00:29:18.359 --> 00:29:21.519
do have some benefits in the field of pursuing truth,
585
00:29:22.279 --> 00:29:24.720
they are not helpful. Now, we're going to look at
586
00:29:24.759 --> 00:29:27.960
medical research and in medicine and how we can model
587
00:29:28.039 --> 00:29:31.279
how medical research treats narratives as a way that we
588
00:29:31.319 --> 00:29:34.599
can implement those same defenses in our daily lives. So
589
00:29:34.599 --> 00:29:37.880
I'm going to go through some narrative spins that the
590
00:29:37.920 --> 00:29:42.440
medical community does encounter. But you'll notice that the medical
591
00:29:42.480 --> 00:29:46.119
community has these defenses to fight against these spins.
592
00:29:46.519 --> 00:29:49.400
So one spin is that my patients are different enough.
593
00:29:49.400 --> 00:29:51.279
So if we think about any kind of medical study,
594
00:29:51.519 --> 00:29:54.039
they're going to collect a bunch of patients, and it'sportant.
595
00:29:53.519 --> 00:29:55.359
They're all patients of one type.
596
00:29:55.440 --> 00:29:57.839
They want to make sure that their patients are diverse
597
00:29:57.960 --> 00:30:00.920
enough right now. The problem here is that selection bias
598
00:30:01.200 --> 00:30:04.400
might be visible to the scientist that's choosing the patients
599
00:30:04.720 --> 00:30:06.559
to run that study on it, and they might have
600
00:30:06.640 --> 00:30:08.920
very good intentions, but it doesn't prevent the fact that
601
00:30:08.960 --> 00:30:11.799
there is bias assignment. So the way to prevent that
602
00:30:11.920 --> 00:30:16.039
is by having completely randomized sample collection. If it's randomized,
603
00:30:16.200 --> 00:30:19.400
then we can protect against our inherent human biases.
604
00:30:19.759 --> 00:30:21.799
Another spin is I can tell.
605
00:30:21.640 --> 00:30:25.880
When it's working right, I'm seeing all right, I'm treating
606
00:30:26.240 --> 00:30:29.839
patient A with something with the treatment I'm not treating
607
00:30:29.839 --> 00:30:34.039
patient B. Oh, looks like patient day is doing much better. Right, Well,
608
00:30:34.279 --> 00:30:36.400
if they know that Patient A is the one that's
609
00:30:36.480 --> 00:30:42.519
taking the treatment, that can impact how they are processing
610
00:30:42.519 --> 00:30:46.039
the results. That can impact how they're coming to their conclusion,
611
00:30:46.519 --> 00:30:50.000
and that can lead to faulty results. So as a result,
612
00:30:50.680 --> 00:30:54.000
there are the testing that takes place is double blinded,
613
00:30:54.440 --> 00:30:57.200
meaning that people that are administering the test don't know
614
00:30:57.279 --> 00:31:00.359
which group has the treatment, and the people seen in
615
00:31:00.400 --> 00:31:01.960
the test don't know whether or not they have the
616
00:31:02.000 --> 00:31:06.359
treatment or not. Right, so there's protection against that. Another
617
00:31:06.440 --> 00:31:11.599
mirror to spend. All these patients improved after the treatment. Well,
618
00:31:11.839 --> 00:31:16.400
the problem is time heals, right, so people will generally
619
00:31:16.440 --> 00:31:20.599
improve after treatment. And also, the placebo effect is not imaginary.
620
00:31:20.799 --> 00:31:23.119
We know the placebo effect is a very real thing.
621
00:31:23.519 --> 00:31:28.559
So to counteract this, control groups are created. And most
622
00:31:28.559 --> 00:31:31.039
of you probably understand this if you're not familiar with
623
00:31:31.079 --> 00:31:34.960
the idea of a control group, it is a set
624
00:31:35.000 --> 00:31:37.960
of people that don't receive anything, right, so anything that
625
00:31:37.960 --> 00:31:40.240
you're seeing from them in terms of improvements, we know
626
00:31:40.359 --> 00:31:44.839
our function of the placebo effect. So a final one here,
627
00:31:45.440 --> 00:31:47.039
this study proves our hypothesis.
628
00:31:47.160 --> 00:31:47.599
We did it.
629
00:31:47.920 --> 00:31:52.839
We performed the study, the treatments, they worked, So now
630
00:31:53.440 --> 00:31:56.240
we know we can move forward with this information. Well,
631
00:31:57.160 --> 00:31:59.440
this is about the pursuit of truth, and if it's
632
00:31:59.440 --> 00:32:01.880
really a true it should live out there in the universe.
633
00:32:02.119 --> 00:32:04.519
We should be able to access that truth again. So
634
00:32:05.079 --> 00:32:08.519
because we're always at the whim of a limited sample size,
635
00:32:08.680 --> 00:32:12.079
it's always good to have replication. So what you'll notice is,
636
00:32:12.119 --> 00:32:14.640
and I really love this about science, and I'm passionate
637
00:32:14.640 --> 00:32:19.359
about it. Science utilizes a red team blue team approach
638
00:32:19.640 --> 00:32:22.160
when it comes to validating results. By that, it means
639
00:32:22.480 --> 00:32:27.000
you have competing scientists that when one scientist runs a
640
00:32:27.039 --> 00:32:29.680
particular experiment, and this is not limited to the medical community,
641
00:32:29.799 --> 00:32:32.200
this is any kind of science. When one scientist comes
642
00:32:32.319 --> 00:32:37.759
up with a result, a meaningful result, a competing scientist
643
00:32:38.079 --> 00:32:41.960
will run an experiment, a similar experiment. Because the scientist
644
00:32:42.000 --> 00:32:46.079
that publishes the result, they also publish how exactly they
645
00:32:46.079 --> 00:32:47.880
came to that result. And the reason they do that
646
00:32:48.240 --> 00:32:50.400
is because it shows that if somebody else were to
647
00:32:50.480 --> 00:32:53.839
follow the same protocol that I did, you should see
648
00:32:53.880 --> 00:32:57.519
the same thing. So those other scientists, the competing scientists,
649
00:32:57.839 --> 00:33:01.440
will do the same thing, and if it shows that yes,
650
00:33:01.480 --> 00:33:03.440
I got the same results. That's what you see. That
651
00:33:03.599 --> 00:33:06.279
that becomes published. Another thing I want to point out here,
652
00:33:06.880 --> 00:33:09.400
and this is from the scientific community, So there's this
653
00:33:09.480 --> 00:33:11.079
concept of a pe value. This is going to be
654
00:33:11.079 --> 00:33:14.680
as technical as it gets, I'm saying in this talk,
655
00:33:15.079 --> 00:33:18.359
as technical as I'm going to get. So in any
656
00:33:18.400 --> 00:33:21.279
scientific experiment, it starts with a null hypothesis. So a
657
00:33:21.359 --> 00:33:25.960
null hypothesis is really a statement that encompasses our understanding
658
00:33:25.960 --> 00:33:28.519
as to how things are right now right. So a
659
00:33:28.559 --> 00:33:31.240
classic example I like to give. Let's say we're flipping
660
00:33:31.240 --> 00:33:33.640
a coin. The null hypothesis is that the coin is
661
00:33:33.640 --> 00:33:35.200
a fair coin, it's not weighted.
662
00:33:36.519 --> 00:33:38.000
I then can.
663
00:33:38.039 --> 00:33:42.440
Collect evidence to determine if an alternative hypothesis is supported
664
00:33:42.799 --> 00:33:47.160
enough to reject my null hypothesis. So, if I suspect
665
00:33:47.200 --> 00:33:50.880
the coin might be weighted, my alternative hypothesis is that
666
00:33:51.160 --> 00:33:54.000
the coin is weighted. I can then run an experiment
667
00:33:54.039 --> 00:33:56.160
where I flip the coin and I log whether or
668
00:33:56.160 --> 00:33:58.519
not it landed on heads or tails, and I'm looking
669
00:33:58.559 --> 00:34:01.880
to see if there's enough evidence that disputes my null hypothesis.
670
00:34:02.079 --> 00:34:04.920
Now this becomes a matter of probability. Let's say I
671
00:34:04.960 --> 00:34:07.200
flipped that coin ten times and it lands on heads
672
00:34:07.319 --> 00:34:11.360
six times. We need a weigh a quantitative way to say, okay,
673
00:34:11.360 --> 00:34:13.880
well is the coin weighted or not? I mean if
674
00:34:13.880 --> 00:34:15.559
we didn't have a quantitative way, because say, hey, look
675
00:34:15.559 --> 00:34:17.519
it's supposed to be heads five of the ten times,
676
00:34:17.519 --> 00:34:18.840
it was head six of the ten times.
677
00:34:18.960 --> 00:34:21.039
Sixty percent is very different than fifty percent.
678
00:34:21.599 --> 00:34:24.360
If I went to the casino and I won fifty
679
00:34:24.360 --> 00:34:26.960
to fifty bets sixty percent of the time, I would
680
00:34:26.960 --> 00:34:29.360
be a very rich man. Right, It's a meaningful difference.
681
00:34:29.639 --> 00:34:31.599
But we know the sample size is only ten. So
682
00:34:31.840 --> 00:34:33.800
how do we deal with this. Well, we have this
683
00:34:33.920 --> 00:34:36.559
value called a P value. So P value runs a
684
00:34:36.599 --> 00:34:38.960
specific statistical test. I'm not going to get into that
685
00:34:39.320 --> 00:34:41.440
what the test is, but I can run a statistical
686
00:34:41.480 --> 00:34:44.719
test to say, what are the odds that my coin
687
00:34:44.760 --> 00:34:48.599
would land on heads six or more times if it
688
00:34:48.760 --> 00:34:53.079
was a fair coin. The probability that it would do that,
689
00:34:53.199 --> 00:34:56.639
meaning the probability that I would see my results given
690
00:34:56.679 --> 00:35:00.679
that the null hypothesis is true is the P value.
691
00:35:01.280 --> 00:35:04.800
So it's the probability that my null hypothesis is true
692
00:35:05.639 --> 00:35:10.840
given the results. So generally, scientists, it's not as exciting.
693
00:35:11.119 --> 00:35:15.480
When you fail to disprove an all hypothesis, scientists kind
694
00:35:15.480 --> 00:35:18.280
of are hoping that then all hypothesis is disproven by
695
00:35:18.320 --> 00:35:21.400
the evidence, meaning smaller P values are better. It means, oh,
696
00:35:21.440 --> 00:35:24.199
it's less likely that my null hypothesis is true. Now
697
00:35:24.280 --> 00:35:26.920
in the scientific community, there's kind of a somewhat arbitrary
698
00:35:26.960 --> 00:35:29.280
line in the sand at a value of point oh five,
699
00:35:29.880 --> 00:35:32.480
meaning that if your P value is greater than point
700
00:35:32.480 --> 00:35:36.400
oh five, the community will treat that result as not
701
00:35:36.960 --> 00:35:40.480
You did not obtain sufficient enough evidence to dispute the
702
00:35:40.559 --> 00:35:43.480
null hypothesis. Doesn't mean you're definitely wrong. It might be
703
00:35:43.480 --> 00:35:46.000
worth doing again, or having a slightly different experiment, or
704
00:35:46.239 --> 00:35:48.800
gathering a larger sample. It just means that you haven't
705
00:35:48.840 --> 00:35:51.840
gotten there yet. Once you've gotten below point oh five,
706
00:35:52.159 --> 00:35:57.639
it's generally accepted. That's sometimes you'll see the word statistically significant.
707
00:35:58.119 --> 00:35:58.840
That's what they mean.
708
00:35:59.199 --> 00:35:59.400
Now.
709
00:35:59.599 --> 00:36:02.679
One that keep in mind is that five percent is
710
00:36:02.679 --> 00:36:06.039
actually a lot. So going back to this idea of
711
00:36:07.199 --> 00:36:10.480
red team blue team, and I know that I wanted
712
00:36:10.480 --> 00:36:14.159
this to be thirty minutes of just about done. Imagine
713
00:36:14.199 --> 00:36:16.360
that we wanted to see whether or not people could
714
00:36:16.639 --> 00:36:19.800
predict the outcome of coin flips, and we went around
715
00:36:19.840 --> 00:36:22.000
the United States. There's fifty states in our country, so
716
00:36:22.000 --> 00:36:23.960
you got fifty different people, and you said, Okay, we're
717
00:36:23.960 --> 00:36:25.679
going to flip a point one hundred times. Each flip,
718
00:36:25.679 --> 00:36:27.000
you're going to predict whether or not it's going to
719
00:36:27.000 --> 00:36:29.039
be heads details, then we're going to look at the results.
720
00:36:29.599 --> 00:36:31.960
If we ran that, and then we ran p values
721
00:36:32.239 --> 00:36:35.679
each on each individual person, what you would have is
722
00:36:35.800 --> 00:36:38.400
between two and three people whose PEE value were less
723
00:36:38.400 --> 00:36:41.199
than point oh five. Right, Because we've got fifty trials,
724
00:36:41.559 --> 00:36:44.400
point oh five means one out of twenty people, by
725
00:36:44.480 --> 00:36:51.039
random chance, will have something that is statistically significant. So
726
00:36:51.559 --> 00:36:53.760
that means we probably have two or three people that
727
00:36:53.800 --> 00:36:56.079
get below that number. Now am I saying that, Wow,
728
00:36:56.119 --> 00:36:58.639
that means those two or three people can definitely predict it. No,
729
00:36:58.960 --> 00:37:01.119
I'm making you are of this kind of You have
730
00:37:01.239 --> 00:37:04.599
to be skeptical sometimes just because point of five doesn't
731
00:37:04.599 --> 00:37:06.480
mean it's true. But that's just kind of the line
732
00:37:06.519 --> 00:37:08.119
in the sand. But the point is we're able to
733
00:37:08.119 --> 00:37:10.639
control for the sample size. That's why we have it.
734
00:37:10.960 --> 00:37:14.920
And science uses this and the medical community uses this
735
00:37:15.039 --> 00:37:17.039
to ensure that we get a large enough sample size
736
00:37:17.119 --> 00:37:20.920
to make sure we replicate things. So why does medicine
737
00:37:20.920 --> 00:37:24.119
have all of these defenses against narratives, And it's because
738
00:37:24.159 --> 00:37:28.159
of the cost false positives, meaning if we have a
739
00:37:28.239 --> 00:37:32.800
result that we incorrectly implement into treatment right, if they're
740
00:37:32.840 --> 00:37:35.400
not effective or they're harmful, can be catastrophic. We're talking
741
00:37:35.400 --> 00:37:38.960
about the loss of human life or drastically lowering the
742
00:37:39.039 --> 00:37:41.639
quality of people's lives. We don't want to do that, right,
743
00:37:41.679 --> 00:37:43.960
so the medical community will pay the price in poor
744
00:37:44.000 --> 00:37:47.280
health outcomes. And also it is worth mentioning there is
745
00:37:47.920 --> 00:37:52.519
extreme financial liability when incorrect conclusions are adopted. So as
746
00:37:52.519 --> 00:37:56.559
a result of both of those strong incentives, these high
747
00:37:56.559 --> 00:38:02.679
stakes force or incentivize the medical community to be extremely
748
00:38:02.800 --> 00:38:07.000
vigilant when it comes to treating a hypothesis as actionable intelligence.
749
00:38:07.239 --> 00:38:11.039
It wants to make sure that it is. When there's
750
00:38:11.079 --> 00:38:14.960
a possible narrative, it fights against those narratives to ensure
751
00:38:15.280 --> 00:38:17.559
that we're truly on the pursuit of truth. So I'm
752
00:38:17.559 --> 00:38:22.480
going to end with this. Analytics and narratives they can coexist.
753
00:38:22.800 --> 00:38:25.320
Narratives are wonderful tools for helping us feel better amid
754
00:38:25.400 --> 00:38:29.760
life ups and downs. They can inspire potentially helpful questions,
755
00:38:30.199 --> 00:38:33.679
right that have things gotten bad in Chicago? Is it
756
00:38:33.719 --> 00:38:35.920
better to invest in the market if it's a bear market?
757
00:38:36.000 --> 00:38:38.480
These are helpful questions, right, It gets us closer to truth,
758
00:38:38.599 --> 00:38:41.239
but they're not enough to provide the answer. Analytics is
759
00:38:41.280 --> 00:38:45.079
there to predict the future. It's so currency is truth.
760
00:38:45.119 --> 00:38:47.360
I'm not saying that if analytics pump something out it's
761
00:38:47.400 --> 00:38:50.159
definitely true, but it's trying to be true. It's not
762
00:38:50.159 --> 00:38:52.960
paying attention to emotion. And so if you find yourself
763
00:38:53.000 --> 00:38:56.719
providing a narrative, determine if you're operating in a spectrum
764
00:38:56.760 --> 00:38:59.360
of emotion or if you're operating a spectrum of truth.
765
00:38:59.440 --> 00:39:01.719
And if you're operating in a spectrum of truth, you
766
00:39:01.760 --> 00:39:04.079
want to make sure you reframe that narrative as being
767
00:39:04.079 --> 00:39:06.440
a question or provide more data to back it up.
768
00:39:06.840 --> 00:39:09.480
And now this is going to lead you with, leave
769
00:39:09.519 --> 00:39:14.480
you with. The medical community is extremely vigilant about defending
770
00:39:14.480 --> 00:39:18.039
against narratives to make sure that they don't contaminate their
771
00:39:18.199 --> 00:39:21.760
scientific process of obtaining truth. Why should you care any
772
00:39:21.840 --> 00:39:24.800
less about that than the medical community. Your life is important,
773
00:39:24.920 --> 00:39:28.320
whether it's your business, your physical health, your relationships, your
774
00:39:28.400 --> 00:39:33.480
quality of life, all aspects of your life are extremely important.
775
00:39:33.960 --> 00:39:37.599
You should be vigilant as well when you are looking
776
00:39:37.679 --> 00:39:41.280
at pursuing truth through ensuring that narratives aren't contaminating your
777
00:39:41.280 --> 00:39:44.559
decision making process. So don't fall victim to that primal
778
00:39:44.639 --> 00:39:46.920
human itch that will always be there, which is to
779
00:39:46.960 --> 00:39:49.840
have an explanation in the absence of knowledge. When you're
780
00:39:49.840 --> 00:39:53.480
in the spectrum of truth, it is vital that you
781
00:39:53.679 --> 00:39:57.320
set and be vigilant in that absence of knowledge until
782
00:39:57.360 --> 00:40:04.440
the evidence has given you sufficient information to justify implementing
783
00:40:04.440 --> 00:40:06.639
a conclusion, because if you don't do that, it will
784
00:40:06.719 --> 00:40:08.280
hurt you and it will get in the way of
785
00:40:08.360 --> 00:40:13.280
being more productive and living longer and living a higher
786
00:40:13.320 --> 00:40:14.920
quality of life, which is what we're trying to do
787
00:40:14.960 --> 00:40:16.239
in the Transhumanist Party.
788
00:40:16.440 --> 00:40:20.239
So with that I will open it up to any questions.
789
00:40:21.039 --> 00:40:26.119
Yes, thank you very much David for that highly informative
790
00:40:26.320 --> 00:40:31.400
and engaging presentation. And it's interesting we had a comment
791
00:40:31.840 --> 00:40:34.760
before you mentioned it that you should be a professor,
792
00:40:34.800 --> 00:40:38.239
and in fact you do teach a data science course
793
00:40:38.760 --> 00:40:45.400
for MIT, so you are applying your presentation skills to
794
00:40:46.000 --> 00:40:49.119
help the next generation of students understand these insights and
795
00:40:49.639 --> 00:40:53.639
our friend Jan Tabanaw rights that you are a very
796
00:40:54.079 --> 00:40:59.400
engaging presenter. Daniel Tweet rights narratives are kind of like
797
00:40:59.559 --> 00:41:03.079
Richard Nman's warning about you must not fool yourself because
798
00:41:03.079 --> 00:41:07.519
you are the easiest person to fool. On the other hand,
799
00:41:07.559 --> 00:41:11.280
he also thinks narratives like money should be considered a tool,
800
00:41:11.360 --> 00:41:15.599
and I'm curious to explore more to what extent narratives
801
00:41:15.639 --> 00:41:16.559
can be a tool.
802
00:41:16.679 --> 00:41:16.960
Now.
803
00:41:17.280 --> 00:41:21.000
One effective approach that you mentioned is to turn narratives
804
00:41:21.039 --> 00:41:24.280
into questions so that then they can be subjected to
805
00:41:24.400 --> 00:41:29.000
more rigorous testing, and I agree with that. I'm curious
806
00:41:29.079 --> 00:41:33.599
also if you think narratives can be effective memory aids.
807
00:41:33.639 --> 00:41:37.559
You mentioned that people's memories form more readily when they
808
00:41:37.599 --> 00:41:41.880
tell themselves stories. And indeed, the older I become, the
809
00:41:42.039 --> 00:41:45.760
more it seems I need to rely on that because
810
00:41:46.159 --> 00:41:50.599
as a kid, as a teenager, I had a very
811
00:41:50.639 --> 00:41:54.719
good memory. I think you can attest to that as well.
812
00:41:55.800 --> 00:42:00.639
Now it's interesting because the older I become, the more
813
00:42:00.800 --> 00:42:07.119
I need to essentially encapsulate what I know into stories
814
00:42:07.280 --> 00:42:11.480
or narratives in order to remember it more effectively. And
815
00:42:11.559 --> 00:42:15.159
I don't think it's entirely the result of aging. I
816
00:42:15.199 --> 00:42:19.760
think it's also the result of just the overwhelming number
817
00:42:19.800 --> 00:42:22.840
of facts that we are met with in our day
818
00:42:22.880 --> 00:42:26.320
to day experience. Whereas, say, when I was five years old,
819
00:42:26.519 --> 00:42:29.960
I could convincingly, at least for myself, say, well, I
820
00:42:30.039 --> 00:42:33.000
know all the planets of the solar system. I know
821
00:42:33.119 --> 00:42:36.280
the names of the numbers up to one Google. I
822
00:42:36.400 --> 00:42:39.800
know the countries of the world and their capitals. I
823
00:42:39.800 --> 00:42:44.079
can identify their flats. So I thought I know everything
824
00:42:44.400 --> 00:42:51.840
because that was the universe of knowledge that was disclosed
825
00:42:51.880 --> 00:42:54.440
to me and available to me, and I had memorized it.
826
00:42:54.480 --> 00:42:58.440
I had memorized laundering lists of facts. Now, over time
827
00:42:58.960 --> 00:43:05.280
that approach to essentially ingest laundry lists of facts and
828
00:43:05.840 --> 00:43:10.519
be able to repeat them gets replaced by more conceptual frameworks,
829
00:43:10.519 --> 00:43:13.559
and those conceptual frameworks are narratives until you get to
830
00:43:13.599 --> 00:43:15.760
a point in your life where there are just way
831
00:43:15.760 --> 00:43:19.039
too many facts to memorize, so you need stories to
832
00:43:19.599 --> 00:43:22.840
make sense of them. But would you say there are
833
00:43:22.960 --> 00:43:28.159
better or worse ways to use such stories when it
834
00:43:28.239 --> 00:43:32.000
comes to even aiding one's own memory.
835
00:43:32.119 --> 00:43:34.239
Yeah, so that's a great point. So I think you
836
00:43:34.360 --> 00:43:39.039
basically answered the question that I would. Narratives can absolutely
837
00:43:39.079 --> 00:43:43.599
be a tool that can be beneficial. You're accurate in
838
00:43:43.599 --> 00:43:47.079
that I read a book and I forgot what book
839
00:43:47.119 --> 00:43:48.679
it was. To be honest, I feel kind of bad
840
00:43:48.679 --> 00:43:50.480
about that. There was a book that talked to Oh
841
00:43:50.519 --> 00:43:51.079
I remember it was.
842
00:43:51.119 --> 00:43:53.440
It was called Peak. Forgot the author sorry.
843
00:43:53.760 --> 00:43:55.800
Peak was about the science of getting good at something,
844
00:43:56.320 --> 00:43:58.559
and one of the groups that it talked about were
845
00:43:58.599 --> 00:44:00.960
the people that extead of their memory.
846
00:44:01.079 --> 00:44:03.719
They literally compete in memory competitions.
847
00:44:03.599 --> 00:44:07.480
And the way that they remember these long sequences is
848
00:44:07.519 --> 00:44:09.199
by using narratives to connect them.
849
00:44:09.360 --> 00:44:10.800
So the way I want to connect that to what
850
00:44:10.840 --> 00:44:12.039
I mentioned during the.
851
00:44:13.840 --> 00:44:17.440
During the presentation is that you want to be clear
852
00:44:17.639 --> 00:44:20.960
as to what your objective is. If your objective is
853
00:44:20.960 --> 00:44:24.320
to help to remember something, then narratives are fine. I'm
854
00:44:24.360 --> 00:44:27.440
saying that narratives are harmful when you are trying to
855
00:44:27.480 --> 00:44:30.760
pursue truth or as you're trying to pursue truth, because
856
00:44:30.880 --> 00:44:34.599
narratives are not now, they are helpful as questions in
857
00:44:34.639 --> 00:44:36.760
that realm. So one thing, and this is what I
858
00:44:36.760 --> 00:44:41.400
tell like the MIT students, analytics is the process of
859
00:44:41.480 --> 00:44:44.159
moving from a question to an answer. But notice that
860
00:44:44.239 --> 00:44:47.000
the starting point is the question. Right, So if you're
861
00:44:47.039 --> 00:44:49.639
asking a question that isn't helpful. If you're asking a
862
00:44:49.719 --> 00:44:52.599
question where the even getting an answer to that would
863
00:44:52.639 --> 00:44:55.480
not be very helpful. It's it's not going to be
864
00:44:55.480 --> 00:44:57.800
as useful you can get you can have the correct answer,
865
00:44:57.800 --> 00:45:00.360
it doesn't help you too much. Having very many meaningful
866
00:45:00.480 --> 00:45:06.239
questions can lead to better outcomes. So I'll give you an
867
00:45:06.239 --> 00:45:08.360
example in business, just because I'm used to giving examples
868
00:45:08.400 --> 00:45:11.280
like that. Let's go back to that marketing campaign example.
869
00:45:12.159 --> 00:45:17.519
If I ask the question, given a certain marketing campaign,
870
00:45:18.079 --> 00:45:21.639
how many people can I expect to come to my website?
871
00:45:22.280 --> 00:45:24.159
A better question would be.
872
00:45:25.760 --> 00:45:29.599
Of a number of marketing campaigns which would provide the
873
00:45:29.760 --> 00:45:32.679
higher return on investment, right, because one of them is like, okay, great,
874
00:45:32.760 --> 00:45:34.119
if I told you that fifty people are going to
875
00:45:34.119 --> 00:45:35.760
go to your website if you go to marketing campaign A,
876
00:45:36.119 --> 00:45:37.679
that's not really enough to know whether or not you
877
00:45:37.679 --> 00:45:38.119
should do it.
878
00:45:38.159 --> 00:45:38.800
What do we care about?
879
00:45:38.840 --> 00:45:41.360
We care about making money in that context, right, So
880
00:45:41.400 --> 00:45:44.360
you want to make sure that you know the lane
881
00:45:44.360 --> 00:45:46.199
that errative is in. Now, one thing I would like
882
00:45:46.239 --> 00:45:47.800
to do, because I'm looking at some of the comments
883
00:45:47.840 --> 00:45:50.880
and I'm very passionate about this. If it's okay with you.
884
00:45:51.199 --> 00:45:54.000
There are some people saying that science cannot explain everything,
885
00:45:54.159 --> 00:45:56.320
and I want and science does not know everything. I'd
886
00:45:56.360 --> 00:46:01.840
love to address that if if you want that, okay, yes, please, okay,
887
00:46:01.920 --> 00:46:06.280
all right, wonderful. So some people say out there, they're
888
00:46:06.320 --> 00:46:09.760
saying that science cannot explain any science cannot explain everything,
889
00:46:10.280 --> 00:46:12.239
and that science can be wrong.
890
00:46:13.199 --> 00:46:14.360
So I want to.
891
00:46:14.280 --> 00:46:17.360
Dispute those but in a very specific way. I don't
892
00:46:17.360 --> 00:46:21.599
think anybody is suggesting that science can explain everything and
893
00:46:21.800 --> 00:46:24.360
science cannot be wrong. And here's what I mean by that.
894
00:46:25.440 --> 00:46:31.039
Science is a process. Science was not invented by people.
895
00:46:31.760 --> 00:46:38.800
Science is a process by which evidence is collected to
896
00:46:38.840 --> 00:46:42.400
pursue truth, to identify what is actually happening.
897
00:46:42.840 --> 00:46:45.599
So you can engage in the scientific process.
898
00:46:45.639 --> 00:46:48.519
And by the way, not only has this happened, we
899
00:46:48.719 --> 00:46:51.320
assume this is going to happen. This is how we
900
00:46:51.360 --> 00:46:54.840
evolve and how we accumulate knowledge is we assume that
901
00:46:54.920 --> 00:46:59.159
science is going to come up with conclusions that aren't
902
00:46:59.239 --> 00:47:02.840
sufficient or they might be incorrect. That's why there are
903
00:47:02.880 --> 00:47:06.840
competing scientists. So the idea that when there's a publication
904
00:47:07.719 --> 00:47:11.239
on any topic, it means that science is making a
905
00:47:11.280 --> 00:47:16.639
declaration that something is happening that is not That is
906
00:47:16.679 --> 00:47:20.079
not the correct framing. Science is the process that is
907
00:47:20.159 --> 00:47:24.199
outputting a specific conclusion. But one thing that all scientists,
908
00:47:24.239 --> 00:47:27.679
if they are I mean all might be an extreme,
909
00:47:27.840 --> 00:47:31.079
A vast majority of scientists, if they're being honest or
910
00:47:31.280 --> 00:47:34.960
if they're a good scientist, will acknowledge that any scientific
911
00:47:35.000 --> 00:47:38.000
conclusion is not the end of the path. If you
912
00:47:38.079 --> 00:47:42.000
look up any scientific paper, and I mean any scientific paper,
913
00:47:42.480 --> 00:47:46.159
you will find a large number of references, which means
914
00:47:46.159 --> 00:47:50.679
that that scientific result stemmed from a previous result that
915
00:47:50.719 --> 00:47:54.320
took place. You will also find that if you wait
916
00:47:54.880 --> 00:47:57.840
a month, you will have other scientific papers that are
917
00:47:57.880 --> 00:48:02.400
built off of what that scientific paper had come up with. Also,
918
00:48:02.440 --> 00:48:05.199
it's very common and encouraged that in the field of
919
00:48:05.239 --> 00:48:09.000
science you have multiple ideas, multiple so you can have
920
00:48:09.079 --> 00:48:11.760
multiple publications that can output the same thing. I'll give
921
00:48:11.760 --> 00:48:13.480
you this a great idea, and I'm not that well
922
00:48:13.559 --> 00:48:15.519
versed in this, I just know that this dispute exists.
923
00:48:16.239 --> 00:48:18.719
I've heard that eggs are good for you, and I've
924
00:48:18.719 --> 00:48:20.719
heard that eggs are terrible for you, and I hear
925
00:48:20.719 --> 00:48:23.840
it back and forth all the time. Those conclusions are
926
00:48:24.039 --> 00:48:27.719
both coming from science. There is some evidence that's collected,
927
00:48:28.159 --> 00:48:31.039
there's a conclusion that's being output. The reason that it
928
00:48:31.119 --> 00:48:34.760
continues to be debated is because science is a never
929
00:48:34.920 --> 00:48:40.800
ending process of collecting evidence, testing hypotheses, and coming up
930
00:48:40.840 --> 00:48:43.199
with conclusions that are more data driven. And as more
931
00:48:43.199 --> 00:48:46.800
time goes by, the more evidence that we have. So
932
00:48:46.840 --> 00:48:50.079
science can't be wrong because it's only the process of
933
00:48:50.199 --> 00:48:55.039
using logic and data to try to improve on our understanding.
934
00:48:56.199 --> 00:49:00.559
So hopefully, and I'm hoping, I'm not attacking anything. I'm
935
00:49:00.559 --> 00:49:05.920
not attacking you know, I'm not attacking religion. And what
936
00:49:06.000 --> 00:49:08.000
I think is kind of funny is that Jannati. Earlier
937
00:49:08.000 --> 00:49:09.679
in this talk, i'muted to the fact that one of
938
00:49:09.679 --> 00:49:14.960
the earliest recordings of us or Jani's digital presence was
939
00:49:15.000 --> 00:49:17.440
from a conversation we had we were fourteen or sixteen,
940
00:49:17.480 --> 00:49:20.119
I think, and coincidentally enough, I think the conversation was
941
00:49:20.159 --> 00:49:22.199
about this topic. So I thought it was kind of
942
00:49:22.480 --> 00:49:27.039
interesting to bring it up. So so there are other
943
00:49:28.239 --> 00:49:29.800
some comments bring in here. So I want to I
944
00:49:29.840 --> 00:49:30.880
want to let you. I don't want to just be
945
00:49:30.920 --> 00:49:33.079
going back and forth with the comment section. So Jannati,
946
00:49:33.119 --> 00:49:34.360
it's your show. I want to let you kind of
947
00:49:34.360 --> 00:49:35.880
guide where you'd like this conversation to.
948
00:49:35.800 --> 00:49:42.639
Go absolutely and thank you for that response. In regard
949
00:49:42.679 --> 00:49:47.239
to the question of what can science actually tell us
950
00:49:47.239 --> 00:49:50.519
on the fact that science is a process rather than
951
00:49:50.559 --> 00:49:54.880
a set of answers, though there are certain answers that
952
00:49:54.960 --> 00:49:59.880
the process can converge upon, Like we know that plan
953
00:50:00.079 --> 00:50:05.679
that's follow elliptical orbits around their respective stars, and that
954
00:50:05.719 --> 00:50:11.400
would be extraordinarily difficult to disprove unless you have some
955
00:50:11.599 --> 00:50:18.320
very unusual star system, Like if you could conceivably have
956
00:50:18.960 --> 00:50:23.119
multiple stars a binary star system, maybe the orbits will
957
00:50:23.159 --> 00:50:28.719
be different. But nonetheless, there has been a robust body
958
00:50:28.760 --> 00:50:32.800
of scientific research to indicate that planets follow elliptical orbits,
959
00:50:32.840 --> 00:50:37.840
and we can generally rely on that conclusion. Now, Mike
960
00:50:38.000 --> 00:50:41.920
Lusine writes that you are right on point. There are
961
00:50:41.960 --> 00:50:44.440
some scientists who do act like when they find something
962
00:50:44.480 --> 00:50:47.280
that is the end of the discussion. I think, though,
963
00:50:47.840 --> 00:50:53.239
that that may be less a feature of how scientists act.
964
00:50:53.480 --> 00:50:56.639
That may be more a feature of how certain people
965
00:50:57.039 --> 00:51:02.320
who create narratives that they allege our related to science
966
00:51:02.400 --> 00:51:05.960
will act. So sometimes politicians will say, well, this is
967
00:51:06.000 --> 00:51:10.960
the established science, whatever it may be, and maybe the
968
00:51:11.079 --> 00:51:16.199
science has more exploration to do, more hypothesis testing to do.
969
00:51:16.280 --> 00:51:19.519
And that doesn't mean that there are no conclusions from
970
00:51:19.559 --> 00:51:23.760
the science that can't be relied upon for practical purposes.
971
00:51:23.840 --> 00:51:28.039
But it does mean that there can always be more
972
00:51:28.960 --> 00:51:34.280
studies performed, more research done, more exploration that could potentially
973
00:51:34.360 --> 00:51:36.599
lead us to revisit prior conclusions.
974
00:51:37.239 --> 00:51:38.920
Yes, and I just want to kind of build on that.
975
00:51:39.199 --> 00:51:43.559
So there's this concept of incremental growth. We are on
976
00:51:43.599 --> 00:51:47.480
this incremental path to obtain all of the knowledge of
977
00:51:47.480 --> 00:51:50.639
the universe. I'm not suggesting we'll ever have that. When
978
00:51:51.000 --> 00:51:53.880
if they say that science is saying this, I don't
979
00:51:53.920 --> 00:51:57.840
think it's either extreme. They're not saying that it's meaningless
980
00:51:58.039 --> 00:52:00.719
because it could be disproven in five minutes. I also
981
00:52:00.719 --> 00:52:02.880
don't think they're saying this is it, We're done, we
982
00:52:02.920 --> 00:52:06.559
found it. I think they're basically saying, this is what
983
00:52:06.559 --> 00:52:10.840
we have right now. This is our best understanding, not
984
00:52:10.880 --> 00:52:13.159
to say it can't be improved upon, not to say
985
00:52:13.159 --> 00:52:17.440
that it might be incorrect, but it's our best understanding
986
00:52:17.599 --> 00:52:19.639
as of right now. And all we can do is
987
00:52:19.679 --> 00:52:22.440
operate off of the information that is available to us.
988
00:52:22.920 --> 00:52:25.519
And I think it should be treated with that level
989
00:52:25.559 --> 00:52:28.400
of nuance. Where now as you can tell, like, I'm
990
00:52:28.480 --> 00:52:31.000
very passionate about science. I think science is a great
991
00:52:31.119 --> 00:52:36.719
way to follow for us to get to a place
992
00:52:36.760 --> 00:52:40.320
of a better life. And I think too often we
993
00:52:40.360 --> 00:52:43.400
rely on narratives, we rely on emotions to choose what
994
00:52:43.440 --> 00:52:46.400
we should be listening to instead of just trying to
995
00:52:46.440 --> 00:52:47.840
obtain objective truth.
996
00:52:50.320 --> 00:52:51.920
Yes, yes, absolutely.
997
00:52:52.440 --> 00:52:59.960
And also in terms of the scientific process, I want
998
00:53:00.719 --> 00:53:05.159
if you would agree that there could be different degrees
999
00:53:05.199 --> 00:53:10.639
of acceptable statistical significance depending on what problem you're trying
1000
00:53:10.679 --> 00:53:16.039
to solve. So, if it's in regard to a medical treatment,
1001
00:53:16.199 --> 00:53:21.480
where whether a medical treatment actually is more effective than
1002
00:53:21.519 --> 00:53:25.519
a placebo, one would want to have a rather low
1003
00:53:25.639 --> 00:53:29.119
P value, maybe even a P value less than point
1004
00:53:29.199 --> 00:53:35.039
zero five in many cases, if there really is a
1005
00:53:35.199 --> 00:53:39.159
risk of, say some serious adverside effect stemming from that treatment.
1006
00:53:39.599 --> 00:53:42.400
On the other hand, let's say you're in a business
1007
00:53:42.440 --> 00:53:47.199
context and your objective is to determine do green widgets
1008
00:53:47.239 --> 00:53:52.360
sell better than other colors of widgets, and you run
1009
00:53:53.280 --> 00:53:55.920
a test of that hypothesis and you have a P
1010
00:53:56.119 --> 00:53:59.920
value of say point two. Now point two wouldn't meet
1011
00:54:00.079 --> 00:54:04.360
the test of statistical significance in most cases, but would
1012
00:54:04.360 --> 00:54:08.960
it still be actionable for the business person to decide, Oh,
1013
00:54:09.159 --> 00:54:13.679
I should probably invest more in producing green widgets because
1014
00:54:14.400 --> 00:54:18.719
I am not one hundred percent sure. Indeed, I am
1015
00:54:19.079 --> 00:54:22.480
twenty percent not sure that they will sell better than
1016
00:54:22.519 --> 00:54:26.960
other colors of widgets. But still I'll take my chances
1017
00:54:27.119 --> 00:54:31.159
and try to sell some green widgets because in four
1018
00:54:31.199 --> 00:54:34.800
out of five scenarios, I'm going to earn a greater
1019
00:54:34.880 --> 00:54:35.679
profit doing that.
1020
00:54:36.480 --> 00:54:37.519
Sure, that's a great question.
1021
00:54:38.000 --> 00:54:40.679
So the short answer to your question is that if
1022
00:54:40.679 --> 00:54:43.440
the P value is greater than zero point zero five,
1023
00:54:43.519 --> 00:54:47.199
especially like point two like you mentioned, I actually would
1024
00:54:47.239 --> 00:54:50.760
not go along with it. And here's why. Let's say
1025
00:54:50.840 --> 00:54:54.400
we're doing a simple linear aggression where you're saying, I'm
1026
00:54:54.440 --> 00:54:58.320
trying to see the relationship between investing in this marketing
1027
00:54:58.360 --> 00:55:01.519
campaign and my sales growth. Right, I want to make
1028
00:55:01.559 --> 00:55:03.960
sure that I'm selling more because I'm investing in this
1029
00:55:04.000 --> 00:55:04.599
marketing campaign.
1030
00:55:05.719 --> 00:55:08.039
If your P value, I actually no, I take it back.
1031
00:55:08.039 --> 00:55:09.559
That's not a good example because what you're pointing out
1032
00:55:09.559 --> 00:55:11.280
it's a good question is if it's in an area
1033
00:55:11.880 --> 00:55:15.159
that is under the stock market. You're saying, if it's
1034
00:55:15.159 --> 00:55:19.840
an area that's very unpredictable, shouldn't it be? Okay, if
1035
00:55:19.880 --> 00:55:23.519
there is some element of uncertainty there, right, And you're right,
1036
00:55:23.559 --> 00:55:26.199
but that would manifest itself in a different way. So
1037
00:55:26.760 --> 00:55:29.039
let's say, and this is going to sound kind of silly,
1038
00:55:29.920 --> 00:55:34.320
let me make it not sound silly. Let's say I'm
1039
00:55:34.360 --> 00:55:36.719
trying to make a model that's going to predict what
1040
00:55:36.760 --> 00:55:39.159
the stock market's going to do tomorrow, and my model
1041
00:55:39.199 --> 00:55:41.719
is going to take as input what the stock market
1042
00:55:41.719 --> 00:55:43.480
did yesterday. I'll just keep it really simple.
1043
00:55:43.719 --> 00:55:44.679
Okay.
1044
00:55:44.800 --> 00:55:47.159
If I were to run a statistical test.
1045
00:55:46.920 --> 00:55:50.119
On that, the P value would probably be pretty low.
1046
00:55:50.480 --> 00:55:51.760
I think we kind of meet these criteria.
1047
00:55:51.800 --> 00:55:54.679
Would be pretty low, but it wouldn't be We probably
1048
00:55:54.679 --> 00:55:55.760
wouldn't below point oh five.
1049
00:55:55.800 --> 00:55:57.079
It might be something like point two.
1050
00:55:57.440 --> 00:56:02.159
Right now, having some uncertainty is important, but it should
1051
00:56:02.199 --> 00:56:05.599
not be above point zero five if we're only looking
1052
00:56:05.639 --> 00:56:10.079
at the relationships between variables, right. So a different way
1053
00:56:10.119 --> 00:56:13.599
that we can look at uncertainty is there's something called
1054
00:56:13.639 --> 00:56:17.119
an R squared value. So our squared value is telling
1055
00:56:17.199 --> 00:56:21.960
us how well our model explains the variance that we
1056
00:56:22.039 --> 00:56:26.559
see in the variables that we're looking. So if our
1057
00:56:26.639 --> 00:56:31.679
model can explain thirty five percent of the variance that
1058
00:56:31.719 --> 00:56:36.000
we see in a realm. That's very hard to predict.
1059
00:56:36.400 --> 00:56:39.320
Then to your point, a thirty five percent our squared
1060
00:56:39.400 --> 00:56:42.119
value might be high enough, Like, look, this is an
1061
00:56:42.119 --> 00:56:45.800
incredibly difficult thing to predict. I'm able to remove thirty
1062
00:56:45.800 --> 00:56:48.440
five percent of the uncertainty, right, given the relationship between
1063
00:56:48.480 --> 00:56:51.239
these variables. The other way you can kind of map
1064
00:56:51.280 --> 00:56:54.079
that out is you should have an out of sample
1065
00:56:54.159 --> 00:56:57.199
data sets. By that, I mean when any model is
1066
00:56:57.320 --> 00:57:01.039
trained it that means that you're feeding in actual labeled
1067
00:57:01.119 --> 00:57:03.559
data to it. Not any model, but a supervised machine
1068
00:57:03.639 --> 00:57:06.440
learning model. It is what most models are. When you're feeding.
1069
00:57:06.480 --> 00:57:08.519
When you're teaching a model to be good, that means
1070
00:57:08.519 --> 00:57:10.840
it has to have an answer key first, to know
1071
00:57:10.920 --> 00:57:12.719
what is good and what is bad, right, so then
1072
00:57:12.760 --> 00:57:16.119
it will morph itself so it does as well as
1073
00:57:16.119 --> 00:57:19.199
it possibly can on that data set. But then when
1074
00:57:19.199 --> 00:57:22.800
you're using it, you're feeding it blind data or data
1075
00:57:22.840 --> 00:57:24.400
from the future, because we don't know what it is. Right,
1076
00:57:24.440 --> 00:57:25.960
So I get a data point as to what's happening
1077
00:57:25.960 --> 00:57:27.880
in the market today, and now my model is going
1078
00:57:27.920 --> 00:57:29.400
to try and predict what's going to happen. My model
1079
00:57:29.400 --> 00:57:31.039
doesn't know what's going to happen. Because nobody knows what's
1080
00:57:31.039 --> 00:57:33.280
going to happen, the model will pump something out. So
1081
00:57:33.280 --> 00:57:35.920
when you're training a model, you want to have out
1082
00:57:35.920 --> 00:57:38.719
of sample data because that out of sample data, meaning
1083
00:57:38.800 --> 00:57:41.760
data that the model has never seen before. The model
1084
00:57:41.800 --> 00:57:43.639
is the same relationship with that out of sample data
1085
00:57:43.760 --> 00:57:45.440
as it does with future data, and then it doesn't
1086
00:57:45.440 --> 00:57:47.880
know what it is. So my model might do fantastically
1087
00:57:48.039 --> 00:57:50.639
on training data. If it does horribly on out of
1088
00:57:50.639 --> 00:57:52.480
sample data, I could trust it will do horribly on
1089
00:57:52.519 --> 00:57:53.280
future data.
1090
00:57:53.480 --> 00:57:55.119
If it does decently.
1091
00:57:54.679 --> 00:57:57.280
On out of sample data, then I can say, okay,
1092
00:57:57.320 --> 00:57:59.719
well we probably we'll do decently well on future data.
1093
00:58:00.239 --> 00:58:03.519
That's a long way of saying if I test that
1094
00:58:03.639 --> 00:58:07.320
model on out of sample data, I can measure was
1095
00:58:07.360 --> 00:58:09.880
I able to make money with this model on the
1096
00:58:09.920 --> 00:58:13.320
out of sample data. It's possible the R squared value
1097
00:58:13.320 --> 00:58:15.960
can be zero point three to five and I made
1098
00:58:16.000 --> 00:58:18.119
money with it, because thirty five percent is a lot.
1099
00:58:18.559 --> 00:58:21.199
Going back to your P value example, if we were
1100
00:58:21.199 --> 00:58:25.519
to look at the P values of the X variable,
1101
00:58:25.519 --> 00:58:26.679
which is in this case I'm saying, is.
1102
00:58:26.679 --> 00:58:29.239
What the market of the day before that P value.
1103
00:58:29.239 --> 00:58:30.840
I don't think you're getting to an r square value
1104
00:58:30.880 --> 00:58:33.880
of point three five without the P value being below
1105
00:58:33.920 --> 00:58:36.960
point zero five. So the way that P value manifests
1106
00:58:37.000 --> 00:58:39.599
itself is that it wouldn't have that much of an impact.
1107
00:58:40.000 --> 00:58:43.719
And what is taught to analytics professionals is that if
1108
00:58:43.760 --> 00:58:47.760
you see a variable whose P value is inflated, you
1109
00:58:47.920 --> 00:58:49.800
probably want to leave it out because it's not a
1110
00:58:49.800 --> 00:58:54.159
strong enough relationship, because you should expect a very because
1111
00:58:54.199 --> 00:58:56.519
you understand these two separate questions, right, Whether or not
1112
00:58:56.760 --> 00:59:01.199
an ex variable impacts why variable is separate question of
1113
00:59:01.840 --> 00:59:05.119
can I model it efficiently? Right, especially with the right
1114
00:59:05.199 --> 00:59:07.800
kind of model. So when you're building a yagles MX
1115
00:59:07.800 --> 00:59:12.159
plus B model, right, it's linear, there are some relationships
1116
00:59:12.199 --> 00:59:18.039
that are nonlinear. So imagine that you see what's a
1117
00:59:18.039 --> 00:59:22.760
good example. If you see like four, If you see
1118
00:59:22.840 --> 00:59:25.880
kind of four circles like this, right, that seems like
1119
00:59:25.920 --> 00:59:26.880
something that's modelable.
1120
00:59:26.960 --> 00:59:28.239
Right, it's a very distinct shape.
1121
00:59:28.239 --> 00:59:30.079
If I gave you an X variable, you'd be able
1122
00:59:30.159 --> 00:59:32.239
to pinpoint like all right, I'm pretty sure that you're
1123
00:59:32.239 --> 00:59:34.519
going to see a point here and here right. If
1124
00:59:34.519 --> 00:59:36.400
you were to write a linear model around that it
1125
00:59:36.440 --> 00:59:41.119
would have no idea what's happening. So you can have
1126
00:59:41.239 --> 00:59:44.519
relationships that are meaningful. That doesn't necessarily mean the P
1127
00:59:44.719 --> 00:59:47.440
value is going to be there. And the P value
1128
00:59:47.480 --> 00:59:49.880
is not there, that means your model probably isn't effective.
1129
00:59:50.159 --> 00:59:53.920
So that's a long way of saying I do think
1130
00:59:53.960 --> 00:59:57.559
that your P values should have to be low, regardless
1131
00:59:57.599 --> 01:00:00.280
of how uncertain the relationships are.
1132
01:00:00.039 --> 01:00:02.079
Are uncertain what you're trying to predict.
1133
01:00:02.400 --> 01:00:05.960
I am not saying, however, that every single evaluation metric
1134
01:00:06.039 --> 01:00:08.559
needs to be extremely high, like I needed R square
1135
01:00:08.639 --> 01:00:12.440
value of point nine if it's something that's more difficult
1136
01:00:12.440 --> 01:00:15.039
to be predicted.
1137
01:00:15.360 --> 01:00:19.760
Yes, thank you for that answer, And indeed, it seems
1138
01:00:19.800 --> 01:00:24.880
like in many situations we could be more forgiving of
1139
01:00:24.920 --> 01:00:30.079
a lower our squared value than of a higher P value.
1140
01:00:30.559 --> 01:00:35.199
But I appreciate the rigor in insisting that the P
1141
01:00:35.400 --> 01:00:40.519
values should be below point zero five. Now, I'm also
1142
01:00:40.760 --> 01:00:46.960
curious as to your assessment of the relative merits and
1143
01:00:47.039 --> 01:00:50.840
perhaps some of the disadvantages or challenges of the various
1144
01:00:50.880 --> 01:00:54.679
types of predictive models that are available today. As you mentioned,
1145
01:00:54.719 --> 01:00:59.280
they range from simple linear regression models, which may fail
1146
01:00:59.360 --> 01:01:05.199
to capture nonlinear trends to AI models with potentially billions
1147
01:01:05.199 --> 01:01:10.480
of variables, which could be very precise, but sometimes they
1148
01:01:10.480 --> 01:01:13.159
may seem like a black box to a lot of
1149
01:01:13.199 --> 01:01:17.280
people because the inner workings are not as transparent. And
1150
01:01:17.480 --> 01:01:22.159
also there could be skepticism, at least in certain designs
1151
01:01:22.199 --> 01:01:26.320
of models, like large language models of hallucinations. So the
1152
01:01:26.400 --> 01:01:30.480
model may be outputting something that seems very precise, that
1153
01:01:30.519 --> 01:01:33.719
seems very cogent, but it could still be false because
1154
01:01:33.760 --> 01:01:36.960
of some artifact of how the model operates. And then
1155
01:01:37.000 --> 01:01:39.679
there are models that are in between. So there are
1156
01:01:40.039 --> 01:01:45.199
generalized linear models, for instance, there are decision tree based models,
1157
01:01:45.639 --> 01:01:50.559
which are more complex. They may have thousands of trees
1158
01:01:51.079 --> 01:01:57.400
of various shapes that may not be intuitively understandable anymore.
1159
01:01:57.519 --> 01:01:59.960
So it seems to me there's this trade off between
1160
01:02:01.239 --> 01:02:08.320
transparency understandability on the one hand, and then preciseness ability
1161
01:02:08.320 --> 01:02:12.199
to model very complex relationships on the other hand. But
1162
01:02:12.320 --> 01:02:15.440
you lose a certain degree of transparency along the way.
1163
01:02:15.960 --> 01:02:20.519
Is there, let's say, some sort of golden mean, or
1164
01:02:20.679 --> 01:02:25.559
is there a way to utilize each model appropriately that
1165
01:02:25.639 --> 01:02:27.840
you don't have as many of the drawbacks.
1166
01:02:29.119 --> 01:02:32.840
So that's an excellent question, given that we're in twenty
1167
01:02:32.840 --> 01:02:36.360
twenty six. This is the kind of question that one
1168
01:02:36.440 --> 01:02:38.599
hundred and fifty years, three hundred, five hundred years from now,
1169
01:02:38.639 --> 01:02:40.679
they're going to look back and they're going to say, yeah,
1170
01:02:40.719 --> 01:02:46.320
people did not know understandably how to how to deal
1171
01:02:46.400 --> 01:02:49.159
with this. Side is the fact that we've built these
1172
01:02:49.199 --> 01:02:53.760
models that are good. They can not every single instance,
1173
01:02:53.760 --> 01:02:56.599
but they can't all provide a lot of value. But
1174
01:02:56.639 --> 01:03:01.639
the problem is we can't interpret them. So all right,
1175
01:03:02.360 --> 01:03:07.360
let's say we're talking about an insurance rate for somebody. Okay, oftentimes,
1176
01:03:07.360 --> 01:03:09.480
and I'm not in the expert on this, I can
1177
01:03:09.519 --> 01:03:11.840
think of somebody who is. If we're looking at what
1178
01:03:11.960 --> 01:03:15.960
someone's insurance rates should be, you can understand if someone
1179
01:03:15.960 --> 01:03:18.159
applies a linear model. Right, Let's say I go out
1180
01:03:18.159 --> 01:03:20.719
to get car insurance and they say, okay, number one,
1181
01:03:21.199 --> 01:03:23.440
your mail, so your car insurance rate's going to go.
1182
01:03:23.480 --> 01:03:24.039
Up a little bit.
1183
01:03:24.199 --> 01:03:25.440
Okay, number two.
1184
01:03:25.559 --> 01:03:28.480
I'm thirty nine, So my car insurance rate is going
1185
01:03:28.519 --> 01:03:30.760
to be at some level because I'm thirty nine, I'm married.
1186
01:03:30.880 --> 01:03:32.599
Maybe that means it goes down a little bit. They
1187
01:03:32.639 --> 01:03:34.639
might look at my FICO score. That means, you know,
1188
01:03:34.880 --> 01:03:36.840
my MICO score is. You know, whatever it is, it's
1189
01:03:36.880 --> 01:03:39.199
going to impact what the insurance rate's going to be. Right,
1190
01:03:39.480 --> 01:03:42.400
So that makes sense to us. So if you look
1191
01:03:42.400 --> 01:03:44.320
at someone's looking at their insurance rate and they say, oh,
1192
01:03:44.400 --> 01:03:46.400
big charge, you know whatever, it's eighty five dollars a month,
1193
01:03:46.519 --> 01:03:48.679
they can look at those different variables, or the person
1194
01:03:48.679 --> 01:03:51.239
that's applying the charge can look at those variables and
1195
01:03:51.360 --> 01:03:54.719
understand where the numbers coming from. Why does a house
1196
01:03:54.760 --> 01:03:56.920
cost as much as it does? Oh, I get it.
1197
01:03:56.920 --> 01:03:58.960
It's got this many bedrooms, it's this many square feet,
1198
01:03:59.000 --> 01:04:01.239
it lives in this time. Right, You've got these these
1199
01:04:01.519 --> 01:04:05.480
different parameters that we can wrap our heads around. When
1200
01:04:05.480 --> 01:04:08.280
you feed a picture of a panda into a model
1201
01:04:08.840 --> 01:04:12.079
and the models able to correctly classify that it's a panda,
1202
01:04:12.159 --> 01:04:17.480
now let's put aside any any concern that it's wrong.
1203
01:04:17.599 --> 01:04:19.679
Let's just assume for the sake of this conversation that
1204
01:04:19.719 --> 01:04:21.119
the model is one hundred percent accurate.
1205
01:04:21.159 --> 01:04:22.800
It can recognize what a panda looks like.
1206
01:04:24.920 --> 01:04:27.599
We can't put a wrap our minds around how it
1207
01:04:27.719 --> 01:04:28.440
knows it's a panda.
1208
01:04:29.000 --> 01:04:30.079
You can't even print it out.
1209
01:04:30.119 --> 01:04:31.760
It's not going to say, all right, well, it's got
1210
01:04:31.760 --> 01:04:35.119
two ears and its face is white, so it's got
1211
01:04:35.159 --> 01:04:37.760
to be a panda, right. It's you can't even look
1212
01:04:37.800 --> 01:04:40.000
at it. I mean that like you could print out
1213
01:04:40.000 --> 01:04:42.119
the series of numbers. I could say, hey, here's all
1214
01:04:42.119 --> 01:04:43.639
the parameters, here's.
1215
01:04:43.639 --> 01:04:45.840
It's got these. I'm not going to even go into it.
1216
01:04:46.199 --> 01:04:48.440
I know how it works, but it's very complex and
1217
01:04:48.480 --> 01:04:52.519
I don't think your audience would appreciate it. The numbers
1218
01:04:52.559 --> 01:04:55.519
would be nothing to you. So we've got these models
1219
01:04:55.519 --> 01:04:57.960
that they work, but it's kind of like, I don't
1220
01:04:58.000 --> 01:04:59.920
know how it works, but we know that it works.
1221
01:05:00.159 --> 01:05:02.039
Because this is what a lot of people don't understand
1222
01:05:02.039 --> 01:05:08.320
about AI. AI looks intelligent and the abilities that these
1223
01:05:08.360 --> 01:05:12.519
models have are fantastic, and it absolutely mimkes intelligence. But
1224
01:05:12.599 --> 01:05:15.159
I think it's really funny about AI is the reason
1225
01:05:15.239 --> 01:05:17.599
the way that these models are come about is through
1226
01:05:17.639 --> 01:05:19.719
trial and error. So I want you to think about
1227
01:05:19.760 --> 01:05:21.760
that for a second. I want your audience really think
1228
01:05:21.800 --> 01:05:24.440
about when I say it's through trial and error. Imagine
1229
01:05:24.480 --> 01:05:26.119
that you want to go from your house to the
1230
01:05:26.199 --> 01:05:28.599
airport wherever you live, and you want to figure out
1231
01:05:28.639 --> 01:05:30.880
what's the best way to get there. The way that
1232
01:05:31.000 --> 01:05:32.920
you're probably going to try to get there is you
1233
01:05:32.960 --> 01:05:34.760
get in your car if you have a car, and
1234
01:05:34.800 --> 01:05:37.320
you kind of start driving in the direction right and
1235
01:05:37.360 --> 01:05:39.159
maybe a time how long it takes. Then if you
1236
01:05:39.280 --> 01:05:41.519
get stuck in traffic or there's some stoplights, think okay,
1237
01:05:41.760 --> 01:05:43.880
next time, instead of going left here, I'm going to
1238
01:05:43.920 --> 01:05:44.400
go right here.
1239
01:05:44.400 --> 01:05:45.679
But I should get a little bit closer.
1240
01:05:45.840 --> 01:05:48.559
The way that models get there is if the airport
1241
01:05:48.639 --> 01:05:50.960
is west of you, the model might start going east.
1242
01:05:51.679 --> 01:05:54.239
It might start going a completely different direction. So you
1243
01:05:54.239 --> 01:05:55.840
see it, walk out the door and like, hey, where
1244
01:05:55.840 --> 01:06:00.000
the heck are you going? Like Olympus is that way right?
1245
01:06:00.119 --> 01:06:02.599
So like the airport's a complete other direction.
1246
01:06:03.519 --> 01:06:04.440
So what it does is it.
1247
01:06:04.400 --> 01:06:07.800
Says, you know what, it took me forty five years
1248
01:06:07.800 --> 01:06:10.199
to drive to the airport. I simulated the distance. Instead
1249
01:06:10.199 --> 01:06:12.880
of driving west ten minutes, I drove east, and I
1250
01:06:12.960 --> 01:06:15.239
hit the ocean and then I drove around a while.
1251
01:06:15.320 --> 01:06:17.880
It took me forty five years. So I'm gonna try
1252
01:06:17.960 --> 01:06:20.039
doing that again. This time I'm gonna go south. And
1253
01:06:20.079 --> 01:06:21.960
if you're watching it and say, oh, why are you
1254
01:06:22.000 --> 01:06:24.360
going south? The airport is west right, what you're not
1255
01:06:24.400 --> 01:06:27.480
seeing is each one of those iterations takes a microsecond,
1256
01:06:27.679 --> 01:06:31.440
so it's able to go through these billions of different combinations,
1257
01:06:31.679 --> 01:06:33.320
and then when you're seeing the end result, you're saying,
1258
01:06:33.360 --> 01:06:35.599
oh my god, look at how brilliant this thing is.
1259
01:06:36.280 --> 01:06:38.440
Not only didn't know to go west, it took this
1260
01:06:38.480 --> 01:06:41.400
other side road that I didn't even know existed, and
1261
01:06:41.440 --> 01:06:43.599
it made it to the airport five minutes faster than
1262
01:06:43.639 --> 01:06:46.159
I ever have and I've lived here for fifteen years.
1263
01:06:46.320 --> 01:06:49.159
This model is amazing, Like, yes, that's amazing that it
1264
01:06:49.159 --> 01:06:51.880
came up with that route. Yes, the application is fantastic,
1265
01:06:52.039 --> 01:06:54.079
and it came up and it came up with that
1266
01:06:54.440 --> 01:06:57.159
conclusion in maybe like two seconds or something. But what
1267
01:06:57.199 --> 01:07:01.480
you're not seeing is the method of getting there is
1268
01:07:01.920 --> 01:07:05.599
like I apologize to not having a better word. It's
1269
01:07:05.639 --> 01:07:08.039
a stupid methodology. Like that's the best way we have
1270
01:07:08.039 --> 01:07:10.360
of going about it right now, which is say, all right,
1271
01:07:10.760 --> 01:07:12.159
we're just going to keep doing it as much as
1272
01:07:12.199 --> 01:07:12.480
we can.
1273
01:07:12.679 --> 01:07:12.840
Now.
1274
01:07:13.039 --> 01:07:15.960
There is something called gradient descent, which means that it
1275
01:07:15.960 --> 01:07:19.960
doesn't just randomly try different things. It will alter what
1276
01:07:20.119 --> 01:07:23.679
it's trying based off of the result of the previous iteration.
1277
01:07:24.039 --> 01:07:26.559
But it's still just iterating over and over again, and
1278
01:07:26.599 --> 01:07:29.440
it's still just doing that through brute force, as somebody
1279
01:07:29.440 --> 01:07:32.400
mentioned in there. So I think that that's a really
1280
01:07:32.440 --> 01:07:36.519
important concept because we're coming up with this just like.
1281
01:07:36.760 --> 01:07:37.960
Raw trial and error.
1282
01:07:38.440 --> 01:07:43.119
That does not mean that these computers have this fantastic
1283
01:07:43.280 --> 01:07:46.559
understanding of the different use cases that we have. So
1284
01:07:47.280 --> 01:07:49.400
given that, I think kind of the that was kind
1285
01:07:49.440 --> 01:07:52.360
of a long didn't mean to go on about that,
1286
01:07:52.440 --> 01:07:55.320
but I think the short answer to the question is
1287
01:07:55.400 --> 01:07:59.079
kind of we have to make sure that we're using models,
1288
01:08:00.239 --> 01:08:04.559
that we're applying them in the right way. It's important
1289
01:08:04.599 --> 01:08:08.400
that we monitor how accurate are the models that we're
1290
01:08:08.480 --> 01:08:11.079
using right and make sure that it's keeping up. Another
1291
01:08:11.119 --> 01:08:13.679
thing we have to be very very careful about is
1292
01:08:13.760 --> 01:08:16.199
we want to be aware of the data that our
1293
01:08:16.239 --> 01:08:19.720
models are constructed off of. As an example, a lot
1294
01:08:19.760 --> 01:08:23.000
of the models that we have on images are images
1295
01:08:23.039 --> 01:08:26.359
that are from the United States. So if you're not aware,
1296
01:08:26.800 --> 01:08:30.439
most people in the world live in Asia, so the
1297
01:08:30.520 --> 01:08:33.920
models actually are biased, not because there's any intention behind it,
1298
01:08:33.960 --> 01:08:36.159
but if the models are fed images of people from
1299
01:08:36.159 --> 01:08:39.000
the United States, that means that the models might not
1300
01:08:39.079 --> 01:08:42.359
be as accurate on people that are Asian because it
1301
01:08:42.399 --> 01:08:46.039
hasn't been trained on that. Now, that's not fair to them, right, So,
1302
01:08:46.439 --> 01:08:48.840
no matter what it is that we're looking at, if
1303
01:08:48.840 --> 01:08:52.520
the model is disproportionately accurate with different groups of people,
1304
01:08:53.039 --> 01:08:55.640
now that can start to be hard, that can become harmful.
1305
01:08:55.960 --> 01:08:58.399
We don't mean that I'm not suggesting that people training
1306
01:08:58.479 --> 01:09:00.960
the models we're racist or anything. That's just like a
1307
01:09:01.119 --> 01:09:04.600
natural consequence of feeding in a selective amount of data.
1308
01:09:04.920 --> 01:09:06.960
And the more we lean on these models, the more
1309
01:09:07.000 --> 01:09:10.680
important it is that we are not doing any unintentional
1310
01:09:10.800 --> 01:09:13.760
harm to groups because of that. That also takes place
1311
01:09:13.760 --> 01:09:18.840
in medicine. By the way, Let's say that we feed
1312
01:09:18.880 --> 01:09:21.560
in medical data into a model to predict what's going
1313
01:09:21.600 --> 01:09:24.119
to happen. Well, there's a lot of people that aren't
1314
01:09:24.119 --> 01:09:26.439
feeding in their medical data because maybe they don't live
1315
01:09:26.479 --> 01:09:28.720
near a hospital, maybe they can't afford to go to
1316
01:09:28.760 --> 01:09:33.760
a hospital, right, So those people's patient information isn't being
1317
01:09:33.760 --> 01:09:36.000
fed into the system. So if you've got a system
1318
01:09:36.000 --> 01:09:38.079
that's supposed to be able to take in let's say
1319
01:09:38.119 --> 01:09:42.039
a different medical attributes or biological attributes and pump out
1320
01:09:42.079 --> 01:09:45.039
a diagnosis, I think we can all agree that it's
1321
01:09:45.039 --> 01:09:48.239
not fair if your skin color is one thing that
1322
01:09:48.279 --> 01:09:50.560
the model is good at predicting what the medical diagnosis is,
1323
01:09:50.680 --> 01:09:52.439
but if your skin color is another then it's not
1324
01:09:53.079 --> 01:09:56.279
so Learning how to do this and applying these models
1325
01:09:56.319 --> 01:10:01.159
in situations it's definitely helpful, and I'm also giving the
1326
01:10:01.279 --> 01:10:03.479
model builders the benefit of the doubt when it comes
1327
01:10:03.520 --> 01:10:05.960
to intention but when it comes to like the challenges
1328
01:10:06.039 --> 01:10:09.319
of how we handle this moving forward. This is really like,
1329
01:10:09.720 --> 01:10:13.760
we have this amazing thing, and frankly, the value that
1330
01:10:13.840 --> 01:10:17.359
it provides is too much. It's too much to say
1331
01:10:18.039 --> 01:10:20.520
we're not going to use AI models right now until
1332
01:10:20.520 --> 01:10:22.399
we figured it out, so we're going to go back
1333
01:10:22.399 --> 01:10:25.039
to the way it was. It's too late to do that.
1334
01:10:25.119 --> 01:10:29.199
It's providing too much economic value. So we have to
1335
01:10:29.279 --> 01:10:31.359
kind of figure out as we go, how do we
1336
01:10:31.399 --> 01:10:34.439
deal with the fact that we don't understand when we cannot,
1337
01:10:34.479 --> 01:10:37.119
frankly understand where these models.
1338
01:10:37.359 --> 01:10:40.279
Parameters are coming from, right we just how does it know?
1339
01:10:40.319 --> 01:10:45.079
It's a PANDAM And we want to have better processes
1340
01:10:45.159 --> 01:10:50.920
and more standardization around ensuring that we're not discriminating against
1341
01:10:50.960 --> 01:10:54.840
any groups of people, whether intentional or not, when we're
1342
01:10:54.840 --> 01:10:57.199
constructing these models and utilizing these models.
1343
01:10:58.319 --> 01:11:00.079
Yes, thank you for that.
1344
01:11:00.159 --> 01:11:04.079
Response, and you actually have an episode of profit as
1345
01:11:04.119 --> 01:11:10.359
a science where you go through essentially a hypothetical example
1346
01:11:10.920 --> 01:11:14.760
of testing for whether a particular practice, this is a
1347
01:11:14.840 --> 01:11:20.880
hiring practice in a business, might be unintentionally discriminatory. So
1348
01:11:20.920 --> 01:11:25.800
there are data analysis tools that are geared toward this.
1349
01:11:26.319 --> 01:11:31.159
But to your point, it's interesting to consider essentially how
1350
01:11:31.199 --> 01:11:37.119
the data that one analyzes are compiled, what elements are included,
1351
01:11:37.159 --> 01:11:41.720
what elements might not be included, and to what extent
1352
01:11:42.399 --> 01:11:45.680
If certain data elements are not included, could there be
1353
01:11:46.720 --> 01:11:53.600
ways to supplement those, say through the lived experiences of patients.
1354
01:11:53.680 --> 01:11:57.119
Let's say a patient is not in a demographic group
1355
01:11:57.159 --> 01:12:01.920
whose experiences are commonly captured by data, and maybe there
1356
01:12:01.960 --> 01:12:06.840
should be ways to collect those experiences. But that patient
1357
01:12:06.880 --> 01:12:11.479
will relate something to a physician or to a researcher,
1358
01:12:11.600 --> 01:12:16.119
some sort of fact. Could there be a way to
1359
01:12:16.159 --> 01:12:20.239
make sure that that's taken into consideration because that might
1360
01:12:20.359 --> 01:12:25.399
make the difference for that patient in terms of attaining
1361
01:12:25.439 --> 01:12:27.079
a superior medical outcome.
1362
01:12:28.079 --> 01:12:30.680
Yeah, so there are different ways to kind of address
1363
01:12:30.720 --> 01:12:35.279
for this issue. So one is, I mean the simplest,
1364
01:12:35.279 --> 01:12:37.279
shouldn't say the easiest. The simplest way is just collect
1365
01:12:37.279 --> 01:12:39.960
more data, right, try to get more patients, try to
1366
01:12:39.960 --> 01:12:42.359
get their information in there. One way you can kind
1367
01:12:42.399 --> 01:12:45.479
of synthetically create more data is you can just replicate
1368
01:12:45.520 --> 01:12:48.399
some of the day that you have originally with those groups.
1369
01:12:48.600 --> 01:12:51.239
So if I have one person from group A and
1370
01:12:51.279 --> 01:12:53.560
I've got five people from group B, I just have
1371
01:12:53.880 --> 01:12:55.680
more replication of people from group BAY.
1372
01:12:55.720 --> 01:12:56.800
Now I'm not saying one person.
1373
01:12:57.199 --> 01:12:59.199
What I'm saying like, if I've got five times as
1374
01:12:59.239 --> 01:13:01.000
big with another group.
1375
01:13:00.960 --> 01:13:03.560
I can either leave out eighty percent.
1376
01:13:03.680 --> 01:13:06.119
I can randomly select twenty percent from that first group,
1377
01:13:06.119 --> 01:13:08.279
so that way you know they're about equal, or I
1378
01:13:08.319 --> 01:13:12.880
can replicate what's happening in the first group. Another thing
1379
01:13:12.920 --> 01:13:16.479
that you can do is there is a difference sometimes
1380
01:13:16.479 --> 01:13:18.760
with these models. There's a difference between predicting what's going
1381
01:13:18.760 --> 01:13:21.319
to happen and what the actual actionable intelligence would be.
1382
01:13:21.680 --> 01:13:24.520
So an example of this would be, let's say you
1383
01:13:24.560 --> 01:13:26.720
have a model that's predicting and we actually go over.
1384
01:13:26.600 --> 01:13:27.800
This example in the MIT course.
1385
01:13:28.000 --> 01:13:29.600
Let's say we have a model that's predicting whether or
1386
01:13:29.600 --> 01:13:32.479
not someone's going to default on a loan, and the
1387
01:13:32.520 --> 01:13:36.159
model thinks that one gender is more likely going to
1388
01:13:36.199 --> 01:13:38.239
default on the loan than the other one.
1389
01:13:38.560 --> 01:13:42.960
Right, let's say that's what the data says.
1390
01:13:43.720 --> 01:13:46.039
I think we can all agree that it's not fair
1391
01:13:46.279 --> 01:13:49.199
if the model is like pumping this out that we
1392
01:13:49.279 --> 01:13:52.840
have this bias presence that's in the model, and so therefore.
1393
01:13:52.439 --> 01:13:54.600
Like the outcomes are the same. But we can alter though,
1394
01:13:54.680 --> 01:13:55.479
or the thresholds.
1395
01:13:55.680 --> 01:13:57.760
So a model like that is going to predict a
1396
01:13:57.800 --> 01:14:00.439
probability that someone's going to default on a loan. So
1397
01:14:01.079 --> 01:14:06.520
if the gender that is that generally is thought to
1398
01:14:06.560 --> 01:14:13.600
default on a loan more has a say a higher
1399
01:14:13.640 --> 01:14:18.079
threshold with which you'd reject them, you can then normalize
1400
01:14:18.319 --> 01:14:20.479
how often the loads are being granted out right, So
1401
01:14:20.479 --> 01:14:22.079
you can say, okay, so for the I'm not going
1402
01:14:22.119 --> 01:14:23.600
to say which genders, which I don't want to get
1403
01:14:23.600 --> 01:14:26.920
anybody upset. If one gender you say, if there's a
1404
01:14:26.920 --> 01:14:28.920
twenty percent chance they're going to fault on the loan
1405
01:14:29.239 --> 01:14:31.359
or higher, we're not going to give them the loan,
1406
01:14:31.640 --> 01:14:33.680
and the other gender we say, if there's a sixteen
1407
01:14:33.680 --> 01:14:36.199
percent chance they default on the loan or higher, we're
1408
01:14:36.199 --> 01:14:38.680
not going to give them. You can alter those thresholds
1409
01:14:38.800 --> 01:14:42.800
so the actual outcome is equalized. Even if the model
1410
01:14:42.840 --> 01:14:46.119
itself is still still reflects that bias.
1411
01:14:48.640 --> 01:14:53.079
Yes, thank you very much for that answer, and yes,
1412
01:14:53.119 --> 01:14:58.880
it's encouraging that there are data analysis techniques that would
1413
01:14:59.000 --> 01:15:05.119
allow for consideration of such biases and their mitigation. Our
1414
01:15:05.279 --> 01:15:09.479
viewer GS four twenty two recommends the book Invisible Women,
1415
01:15:09.800 --> 01:15:15.279
which highlights statistical bias, and here's the link to the
1416
01:15:15.279 --> 01:15:19.439
book on Amazon. It's a book by Caroline Creato Perez.
1417
01:15:19.520 --> 01:15:24.640
So thank you for that recommendation. Now, another aspect of
1418
01:15:24.760 --> 01:15:30.520
your comments, how aisystems today essentially work by brute forcing
1419
01:15:30.520 --> 01:15:35.640
a solution through trial and error of numerous iterations. It's
1420
01:15:35.720 --> 01:15:40.439
just each iteration takes a fraction of a second to perform.
1421
01:15:40.880 --> 01:15:43.960
It reminds me of this paper. This was a landmark
1422
01:15:43.960 --> 01:15:47.399
paper in twenty seventeen called Attention Is All You Need,
1423
01:15:47.680 --> 01:15:54.000
which essentially launched this revolution of large language models generative
1424
01:15:54.319 --> 01:15:59.600
AI models, and within the ensuing five years those models.
1425
01:16:00.319 --> 01:16:05.560
By having access to more computing resources, essentially we're able
1426
01:16:05.680 --> 01:16:09.399
to generate outcomes and solutions that a lot of people
1427
01:16:09.479 --> 01:16:14.439
found plausible, either visually or in terms of the text
1428
01:16:14.560 --> 01:16:18.800
they received as responses to their inquiries. But that gives
1429
01:16:18.880 --> 01:16:25.159
rise to a question that's commonly posed in the transhumanist realm,
1430
01:16:25.359 --> 01:16:30.760
and that is when do you think artificial superintelligence or
1431
01:16:30.800 --> 01:16:34.800
ASI will arrive? This is from our viewer, my Clausine. So,
1432
01:16:35.800 --> 01:16:40.479
if the way current AI systems work is essentially by
1433
01:16:40.520 --> 01:16:44.840
brute forcing a solution, is that enough to get to
1434
01:16:46.039 --> 01:16:51.720
superhuman capabilities not just in one field, but more generally
1435
01:16:51.800 --> 01:16:57.640
across a variety of fields, across essentially the full array
1436
01:16:57.680 --> 01:17:01.159
of fields that humans would be involved in, except to
1437
01:17:01.159 --> 01:17:05.239
be even better than the best humans are. Or do
1438
01:17:05.279 --> 01:17:10.359
you think there's an inherent limitation to that architecture of
1439
01:17:10.560 --> 01:17:14.319
AI to how far it can advance to the quality
1440
01:17:14.359 --> 01:17:16.920
of the answers it could give us, as there's still
1441
01:17:16.960 --> 01:17:21.279
some human edge that will be needed to supplement these
1442
01:17:21.359 --> 01:17:25.359
models or interpret them. Or are they just bounded by
1443
01:17:25.479 --> 01:17:30.359
the sum total of accumulated human data and experience And
1444
01:17:30.399 --> 01:17:33.600
if you don't have data or experience in certain realms,
1445
01:17:33.680 --> 01:17:38.560
then they can't surpass the collective attainments of humankind.
1446
01:17:40.079 --> 01:17:44.640
So that's a very big question. So first I'm going to.
1447
01:17:47.479 --> 01:17:48.199
I'm going to operate.
1448
01:17:48.239 --> 01:17:52.279
When you say artificial superintelligence, I'm assuming you mean kind
1449
01:17:52.319 --> 01:17:55.840
of imagine one model that is kind of all knowing
1450
01:17:55.840 --> 01:18:01.680
in every single realm, I would say one of the
1451
01:18:01.720 --> 01:18:03.479
things you're going to run into and I'm gonna I'm
1452
01:18:03.479 --> 01:18:05.399
going to go back to a point that I brought
1453
01:18:05.439 --> 01:18:09.279
up during my talk. Every single model needs to have
1454
01:18:09.279 --> 01:18:16.359
a single optimization metric. And I think you're always going
1455
01:18:16.439 --> 01:18:18.760
to need a person. You're always going to need a
1456
01:18:18.800 --> 01:18:23.520
person to choose what the proper optimization metric is for
1457
01:18:23.560 --> 01:18:25.640
a given point in time. And let me just go
1458
01:18:25.680 --> 01:18:28.000
back to like a really basic example, you're trying to
1459
01:18:28.039 --> 01:18:31.239
launch a business. When you're trying to launch your business,
1460
01:18:31.920 --> 01:18:34.600
your first step is probably not to maximize your profit,
1461
01:18:35.600 --> 01:18:39.319
right first person, probably not even trying to minimize your costs,
1462
01:18:39.520 --> 01:18:41.039
probably just trying to maximize your revenue.
1463
01:18:41.279 --> 01:18:41.479
Right.
1464
01:18:41.640 --> 01:18:44.279
You want to get people invested in whatever your product, services,
1465
01:18:44.319 --> 01:18:46.520
You want to get moyal customers, right, you want to
1466
01:18:46.520 --> 01:18:48.520
build from there. It's okay if for not making money
1467
01:18:48.560 --> 01:18:50.880
to start, right, you also are not operating off of
1468
01:18:50.880 --> 01:18:52.760
economies at scale, So your raw cost per unit it's
1469
01:18:52.760 --> 01:18:54.640
probably gonna be the highest ever going to be, right,
1470
01:18:54.840 --> 01:18:59.119
So operating off of this, at some point you shift
1471
01:18:59.439 --> 01:19:02.039
from being from being focused on revenue to say, okay,
1472
01:19:02.039 --> 01:19:04.039
now my revenue is fine, I want to make more profit,
1473
01:19:04.119 --> 01:19:08.600
so you shift to going into profit when you choose
1474
01:19:08.640 --> 01:19:12.000
to do that is something that within a given model
1475
01:19:12.640 --> 01:19:15.039
is going to be outside of its scope.
1476
01:19:15.640 --> 01:19:16.840
Another thing to really consider.
1477
01:19:16.880 --> 01:19:18.840
And we're just on the edge of this and not
1478
01:19:18.920 --> 01:19:25.079
in a good way. We've got these models that provide
1479
01:19:25.640 --> 01:19:28.159
Now I'll just speak from personal experience. So I use
1480
01:19:28.279 --> 01:19:31.760
chatbots in my work. So I build machine learning algorithms.
1481
01:19:32.159 --> 01:19:35.239
I built a program this past year in a programming
1482
01:19:35.279 --> 01:19:37.840
language that I had never spoken before. Now this is
1483
01:19:37.840 --> 01:19:39.640
a fully functioning program. It built it in less than
1484
01:19:39.680 --> 01:19:42.760
a year. It launched two months ago. There is no
1485
01:19:42.880 --> 01:19:46.439
way I would have even accepted the project, let alone
1486
01:19:46.479 --> 01:19:49.520
built it if it weren't for AI chatbots. Now perfectly
1487
01:19:50.600 --> 01:19:54.279
personally or professionally, I use it for education. I didn't say, hey,
1488
01:19:54.319 --> 01:19:57.159
build me this. I said, okay, I don't know how
1489
01:19:57.199 --> 01:20:00.920
to import a library in this new program language. I
1490
01:20:00.960 --> 01:20:03.119
was able to learn it through infinite feedback loop. If
1491
01:20:03.119 --> 01:20:05.039
you were to put a dollar value as to how
1492
01:20:05.119 --> 01:20:07.560
much it would cost me to have somebody that knew
1493
01:20:07.640 --> 01:20:10.039
the programming language to the level that the chappot knew
1494
01:20:10.079 --> 01:20:14.039
the programming language and have them in my office ready
1495
01:20:14.119 --> 01:20:15.920
twenty four to seven. If I couldn't sleep and I
1496
01:20:16.000 --> 01:20:17.479
was up at one in the morning and I say, hey,
1497
01:20:18.560 --> 01:20:19.920
you know, how do I make it variable?
1498
01:20:20.039 --> 01:20:20.279
Do this?
1499
01:20:20.720 --> 01:20:22.319
And they say, oh, here's how you do it.
1500
01:20:22.520 --> 01:20:24.920
That would cost me tens if not like one hundred
1501
01:20:24.960 --> 01:20:27.520
thousand dollars a year. I got access to that tool
1502
01:20:27.760 --> 01:20:31.319
pretty much for free, so you could put a dollar value.
1503
01:20:31.319 --> 01:20:33.960
And now this is anecdotal, but this is pretty sure.
1504
01:20:34.039 --> 01:20:37.760
This is valid worldwide. The value that this is providing me,
1505
01:20:37.880 --> 01:20:41.319
that the economic value is providing me. So on one hand,
1506
01:20:41.359 --> 01:20:44.199
you've got this amazing thing that's providing all this economic value.
1507
01:20:44.960 --> 01:20:46.560
The downside is this is.
1508
01:20:46.520 --> 01:20:50.119
Going to eat a lot of jobs, right, and so
1509
01:20:50.560 --> 01:20:53.720
if we probably don't want to go into a society that,
1510
01:20:53.840 --> 01:20:58.840
even though it is economically beneficial and I'm not going
1511
01:20:58.880 --> 01:21:00.680
to get political, I'm really not trying to do that,
1512
01:21:02.840 --> 01:21:05.359
the option of the game might not be to be
1513
01:21:05.479 --> 01:21:10.199
as economically as productive as possible. We also might want
1514
01:21:10.199 --> 01:21:14.880
to have our civilization right, our private citizens be a
1515
01:21:14.960 --> 01:21:18.279
place where they have a purpose and they have a function.
1516
01:21:18.520 --> 01:21:22.479
So if I were to say, Jannati, guess what I've
1517
01:21:22.520 --> 01:21:25.720
done it, I did it. I built it. The super intelligence.
1518
01:21:25.800 --> 01:21:28.520
It can do everything. We have no need for accountants,
1519
01:21:28.600 --> 01:21:30.399
we have no need for lawyers, we have no need
1520
01:21:30.439 --> 01:21:32.279
for doctors, we have no need for any of this thing.
1521
01:21:32.319 --> 01:21:34.439
You feed whatever you want into my model, I'll make
1522
01:21:34.439 --> 01:21:36.640
it available to you for free. You're going to put
1523
01:21:36.680 --> 01:21:40.039
a lot of people out of work, right, and I
1524
01:21:40.119 --> 01:21:42.319
don't think we like the idea of just handing out
1525
01:21:42.359 --> 01:21:42.960
money for free.
1526
01:21:43.399 --> 01:21:44.000
Another thing I.
1527
01:21:43.960 --> 01:21:46.880
Will point out is that people need purpose, and I
1528
01:21:46.920 --> 01:21:51.279
think purpose it matters. So I do think these AI
1529
01:21:51.319 --> 01:21:53.279
models are going to grow. I do think it's going
1530
01:21:53.319 --> 01:21:55.600
to eat jobs. I'm even kind of taken away that
1531
01:21:55.720 --> 01:21:57.279
I'm not even going to make a statement as to
1532
01:21:57.279 --> 01:21:59.479
whether or not that's like, obviously it's bad that job's
1533
01:21:59.520 --> 01:21:59.840
being taken.
1534
01:22:00.119 --> 01:22:02.159
Saying that means that the AA model shouldn't be implemented.
1535
01:22:02.359 --> 01:22:06.520
I'm just being realistic as to what the downsides are
1536
01:22:06.640 --> 01:22:09.479
of implementing this AI. So you've kind of defined this
1537
01:22:09.640 --> 01:22:12.680
end of the spectrum where we have this massive, super
1538
01:22:12.720 --> 01:22:16.560
intelligent algorithm that's able to handle everything. It's able to
1539
01:22:16.600 --> 01:22:19.359
dispatch its robots to do construction on your house.
1540
01:22:19.600 --> 01:22:23.079
It's going to be perfect. Okay, that may be the case.
1541
01:22:23.319 --> 01:22:26.720
So if you're living on the world by yourself, right,
1542
01:22:26.920 --> 01:22:30.000
and you've got all the money you could ever want, then, yeah,
1543
01:22:30.119 --> 01:22:31.880
you don't have to pay anybody anything because you just
1544
01:22:31.960 --> 01:22:34.399
use this model and do everything. But we have to
1545
01:22:34.439 --> 01:22:36.039
And I mean, this is a question I don't mean
1546
01:22:36.079 --> 01:22:41.199
to retortically answer. It is that what we want we
1547
01:22:41.279 --> 01:22:44.159
have to define what good looks like. And I'll give
1548
01:22:44.159 --> 01:22:45.800
you my opinion because I think that it sounds like
1549
01:22:45.800 --> 01:22:48.880
with your audience, like you'll vibe with this. I feel
1550
01:22:49.000 --> 01:22:53.279
very passionately about the pursuit of science right now. Science
1551
01:22:53.399 --> 01:22:56.960
is all about testing what we don't know, right, It's
1552
01:22:57.000 --> 01:23:00.119
coming up with a question we don't have the answer to. Now,
1553
01:23:00.159 --> 01:23:03.000
you said yourself that models a lot of times they hallucinate.
1554
01:23:03.279 --> 01:23:06.560
Models are trained, they're trained off of previous information. I
1555
01:23:06.560 --> 01:23:11.520
don't think models make for great scientists because how can
1556
01:23:11.560 --> 01:23:15.199
a model really learn new information? Now, maybe at some
1557
01:23:15.279 --> 01:23:18.079
point it's like, maybe at some point it can get there.
1558
01:23:18.159 --> 01:23:20.800
But again, we have to be the ones set aside.
1559
01:23:20.840 --> 01:23:23.920
What are those questions that we want to be asking? Right, So,
1560
01:23:25.119 --> 01:23:28.720
given that I and I think your audience and you
1561
01:23:28.920 --> 01:23:33.960
are interested in how do we progress our civilization so
1562
01:23:34.600 --> 01:23:39.439
we have a better chance of evading our own destruction. Right,
1563
01:23:39.479 --> 01:23:43.600
so this includes colonizing other planets, right, so we can like,
1564
01:23:43.720 --> 01:23:44.880
what do we do at some point?
1565
01:23:44.920 --> 01:23:46.279
And wean't go extreme here?
1566
01:23:46.359 --> 01:23:48.600
Right, It's fun to talk about the Sun's going to
1567
01:23:49.039 --> 01:23:51.279
get to explode at some point. It's going to swallow
1568
01:23:51.279 --> 01:23:54.479
the entire Solar System and we've got time, but it's
1569
01:23:54.479 --> 01:23:57.920
going to happen. Can we get ourselves off our planet
1570
01:23:58.359 --> 01:24:01.239
somewhere else to ensure that, like we could keep living
1571
01:24:01.720 --> 01:24:05.000
the universe might collapse some point, Right, there's different The
1572
01:24:05.000 --> 01:24:07.119
thoughts about that some is that it's always expanding. If
1573
01:24:07.159 --> 01:24:10.880
I expand and then and then at some point come back, right,
1574
01:24:11.439 --> 01:24:14.000
can we counteract that? I mean, it sounds insane, but hey,
1575
01:24:14.039 --> 01:24:16.119
we can divert an ask, Right, that sounds pretty insane
1576
01:24:16.279 --> 01:24:18.880
in saying a couple hundred years ago. To answer these questions,
1577
01:24:18.880 --> 01:24:21.880
we have to progress scientifically, and to progress scientifically, we
1578
01:24:21.920 --> 01:24:24.119
have to have all of our needs met. That means
1579
01:24:24.119 --> 01:24:26.119
that we have to have our children be educated. That
1580
01:24:26.159 --> 01:24:27.720
means we have to have food to eat, That means
1581
01:24:27.760 --> 01:24:29.199
we need to be taken care of. That means we
1582
01:24:29.239 --> 01:24:30.920
need to be able to operate in the society. And
1583
01:24:30.960 --> 01:24:32.520
for all those things that happen to take it back
1584
01:24:32.520 --> 01:24:34.760
towarded by very first points. I think we need people
1585
01:24:34.760 --> 01:24:37.239
to be happy, and so I think we have to
1586
01:24:37.319 --> 01:24:38.439
balance those things.
1587
01:24:38.840 --> 01:24:39.720
And one of the.
1588
01:24:39.720 --> 01:24:42.000
Things I mentioned is that these models operate off of
1589
01:24:42.039 --> 01:24:45.600
singular optimization metrics. I don't think that super intelligence is
1590
01:24:45.600 --> 01:24:48.199
going to be good at balancing. It's going to say, Okay,
1591
01:24:48.359 --> 01:24:51.760
here's the best way to do it, and what is
1592
01:24:51.880 --> 01:24:55.880
best will be shifting and so like. For that reason,
1593
01:24:56.319 --> 01:24:59.880
I guess the short answer is that I don't think
1594
01:25:00.079 --> 01:25:03.560
we're going to get there, not because we can't, but
1595
01:25:03.600 --> 01:25:06.279
I think we shouldn't get it to that level because
1596
01:25:06.279 --> 01:25:09.039
I think we shouldn't because I think it would be
1597
01:25:09.520 --> 01:25:14.560
it would take away from what we're trying to work towards.
1598
01:25:15.560 --> 01:25:20.760
Yes, thank you for that response, And indeed this area
1599
01:25:21.399 --> 01:25:25.800
is a conundrum for a lot of transhumanists because we
1600
01:25:25.880 --> 01:25:29.720
are trying to reach the best possible outcomes for humanity
1601
01:25:29.760 --> 01:25:35.520
and the advancement of human civilization. And Daniel Tweed rights,
1602
01:25:36.039 --> 01:25:39.119
every human now alive actually has a new job keeping
1603
01:25:39.159 --> 01:25:44.520
civilization human or humane. And the vision you articulated that
1604
01:25:45.119 --> 01:25:50.640
perhaps all humans should become scientists in their outlook, that
1605
01:25:50.680 --> 01:25:53.800
they should be the ones to ask the questions, the
1606
01:25:53.800 --> 01:25:57.920
ones to pose the hypotheses, the ones to direct the
1607
01:25:58.159 --> 01:26:00.960
AI systems to the extent that they can help in
1608
01:26:01.039 --> 01:26:04.680
testing those hypotheses. But humans do need to be the
1609
01:26:04.720 --> 01:26:08.840
ones to set the objectives, to determine what a good
1610
01:26:08.880 --> 01:26:13.000
answer is, what a meaningful answer is. And you mentioned
1611
01:26:13.039 --> 01:26:18.079
meaning and purpose. It's interesting to our friend more life,
1612
01:26:18.119 --> 01:26:21.600
please John H Write's purpose. Yes, but perhaps we need
1613
01:26:21.640 --> 01:26:25.840
to disambiguate our sense of purpose from what we do
1614
01:26:26.359 --> 01:26:29.520
to make a living. And I think right now, for
1615
01:26:29.640 --> 01:26:32.279
a lot of people, their purpose is tied up with
1616
01:26:32.800 --> 01:26:35.359
their jobs, and on the one hand, that motivates them
1617
01:26:35.399 --> 01:26:39.920
to acquire resources that help them need better lives. But
1618
01:26:40.039 --> 01:26:43.760
what if we do reach a society where the material
1619
01:26:43.880 --> 01:26:48.920
prerequisites of living are fulfilled for the vast majority of people,
1620
01:26:49.520 --> 01:26:52.439
then I think people will still need a sense of purpose.
1621
01:26:52.600 --> 01:26:59.199
But answering these greater civilizational questions or even cosmic scale
1622
01:26:59.359 --> 01:27:05.319
questions that will determine whether our species survives into the long,
1623
01:27:05.520 --> 01:27:08.720
long term future, that could be a great motivator. That
1624
01:27:08.720 --> 01:27:12.439
could be a great sense of purpose for humans. Aided
1625
01:27:12.560 --> 01:27:16.479
by AI and robots and other kinds of machines that
1626
01:27:16.920 --> 01:27:17.520
we come.
1627
01:27:17.399 --> 01:27:19.960
Up with yes, I agree. So I just want to
1628
01:27:20.000 --> 01:27:21.800
I just want to speak to a couple of points.
1629
01:27:22.000 --> 01:27:24.800
So the first thing I want to respond to is
1630
01:27:24.840 --> 01:27:29.600
that the reason I'm equating purpose with jobs is I'm
1631
01:27:29.640 --> 01:27:33.600
assuming that people won't be given money. So if you're
1632
01:27:33.640 --> 01:27:37.239
saying that, look, we've built this model that can handle
1633
01:27:37.359 --> 01:27:41.279
every aspect of society, including farming, right, which means that
1634
01:27:41.359 --> 01:27:44.479
nobody has to worry about money anymore. Okay, you don't
1635
01:27:44.479 --> 01:27:46.079
wan to worry about food, you want to worry about
1636
01:27:46.079 --> 01:27:49.439
your house. We're so efficient that you don't have to
1637
01:27:49.479 --> 01:27:51.960
make money. You can live fine. Then I totally agree
1638
01:27:52.000 --> 01:27:55.239
with you. Yes, yes, And I'm not saying that like
1639
01:27:55.439 --> 01:27:56.119
I support that or not.
1640
01:27:56.119 --> 01:27:57.399
I'm trying not to be super political.
1641
01:27:58.079 --> 01:28:00.920
I totally agree with you that that, yeah, your purpose
1642
01:28:01.000 --> 01:28:02.960
can be around like being a good father, being a
1643
01:28:03.000 --> 01:28:06.039
good husband, being a good friend, right, help in your community,
1644
01:28:06.359 --> 01:28:09.039
more power to you. I'm operating off of the assumption
1645
01:28:09.199 --> 01:28:13.279
that generally we're not close to the idea that people
1646
01:28:13.319 --> 01:28:15.119
are just gonna be like handed out enough money so
1647
01:28:15.119 --> 01:28:17.760
they can do whatever they want to, which means that
1648
01:28:17.920 --> 01:28:20.039
people will need a way to make money. They will
1649
01:28:20.079 --> 01:28:21.960
need something to do in society, so they can they
1650
01:28:22.000 --> 01:28:24.439
can pay for themselves. So I'm gonna use that as
1651
01:28:24.439 --> 01:28:26.359
a bridge. Would be okay if I responded to something
1652
01:28:26.359 --> 01:28:27.640
that I saw in the comments.
1653
01:28:28.439 --> 01:28:29.920
So yes, absolutely, okay.
1654
01:28:29.960 --> 01:28:32.359
So I saw somebody write that it doesn't make sense
1655
01:28:32.439 --> 01:28:35.520
to operate the economy less efficiently for a sense of purpose.
1656
01:28:35.840 --> 01:28:36.880
I want to push back on that.
1657
01:28:37.159 --> 01:28:38.720
Now, on one hand, I want to say, I'm gonna
1658
01:28:38.720 --> 01:28:40.159
give an extreme, But on the other hand, this whole
1659
01:28:40.199 --> 01:28:42.199
conversation is talking about an extreme. We're talking about this
1660
01:28:42.279 --> 01:28:44.960
like massive superintelligence. There's no need for jobs anymore, so
1661
01:28:45.199 --> 01:28:49.399
we're already there. Let's say there was one person, one
1662
01:28:49.479 --> 01:28:52.479
person that was able to write this thing, and they
1663
01:28:52.560 --> 01:28:55.760
were and they were able to generate trillions and trillions
1664
01:28:55.800 --> 01:28:59.680
and trillions of dollars of wealth, right because they can
1665
01:28:59.720 --> 01:29:03.840
provide everything to everyone, right. But the problem was they
1666
01:29:03.880 --> 01:29:06.800
took all the jobs, and I mean all the jobs.
1667
01:29:07.000 --> 01:29:12.479
No doctors, no teachers, no lawyers, no accountants, no, no contractors, everything. So,
1668
01:29:13.600 --> 01:29:17.359
assuming that we're not just handing out money, do you
1669
01:29:17.520 --> 01:29:19.800
think that that person would want to live in a
1670
01:29:19.880 --> 01:29:23.960
world where every single other person is starving and is
1671
01:29:24.079 --> 01:29:27.119
desperate to survive and is probably going to become pounding
1672
01:29:27.159 --> 01:29:32.279
at the gate because like, because they're desperate. Right, In
1673
01:29:32.399 --> 01:29:35.479
other words, we're talking about extremes. I don't think it
1674
01:29:35.600 --> 01:29:38.279
makes sense if we're there, that we're generating that much
1675
01:29:38.319 --> 01:29:41.760
wealth and we're that efficient. Is it worth going from
1676
01:29:41.800 --> 01:29:45.600
here to here if it means putting billions of people
1677
01:29:46.279 --> 01:29:49.239
out of work? I mean, because that's what we're talking about.
1678
01:29:52.680 --> 01:29:54.760
If jobs are done to provide what we need, no
1679
01:29:55.159 --> 01:29:57.159
jobs are done to provide with a person with money
1680
01:29:57.239 --> 01:29:59.880
needs because if you don't have money, how are you
1681
01:30:00.159 --> 01:30:03.439
to do it? Because no matter what that algorithm is doing. Again,
1682
01:30:03.800 --> 01:30:06.359
assuming that it's not for free, you need a doctor,
1683
01:30:06.560 --> 01:30:09.039
you need a lawyer, right, you need all these things.
1684
01:30:09.399 --> 01:30:12.840
So if everything is provided, it's provided if you have
1685
01:30:12.960 --> 01:30:15.680
the money, so we're not so again, if it's provided.
1686
01:30:15.359 --> 01:30:17.760
For free, then then I agree with you.
1687
01:30:18.119 --> 01:30:21.960
But until we're there, I think people need a purpose,
1688
01:30:22.000 --> 01:30:25.199
and more specifically, people need a way to both provide
1689
01:30:25.279 --> 01:30:28.199
value and get money for themselves. And I'm saying there
1690
01:30:28.279 --> 01:30:32.439
is economic value by the way to a person making
1691
01:30:32.560 --> 01:30:35.520
money and being a happy, productive member of society. A
1692
01:30:35.720 --> 01:30:39.600
desperate person will be a weight on the system. They're
1693
01:30:39.600 --> 01:30:42.199
more likely going to commit crimes, they're more likely going
1694
01:30:42.279 --> 01:30:44.079
to have negative health outcomes, which is going to be
1695
01:30:44.239 --> 01:30:48.359
a tax on our health system. Right, And that's aside
1696
01:30:48.640 --> 01:30:51.159
anything talking about religious as to like, do we want
1697
01:30:51.199 --> 01:30:52.279
to be humane to other people?
1698
01:30:52.560 --> 01:30:54.920
Do we want other people to suffer? I'm putting that aside.
1699
01:30:54.960 --> 01:30:56.319
If you made it a math problem.
1700
01:30:57.079 --> 01:31:00.920
If someone is like if you gave someone some money,
1701
01:31:01.079 --> 01:31:03.319
just like flat out gave them money, but it meant
1702
01:31:03.319 --> 01:31:05.560
that they were more likely going to be happy and productive,
1703
01:31:06.079 --> 01:31:08.319
it's worth giving them that money because it's better than
1704
01:31:08.319 --> 01:31:11.239
the alternative. Right, there's not this false choice that a
1705
01:31:11.319 --> 01:31:14.600
person can just have nothing and have no purpose. And
1706
01:31:15.840 --> 01:31:17.319
by the way, it's not going to cost us anything.
1707
01:31:17.359 --> 01:31:19.239
There's still going to be an economic cost. So the
1708
01:31:19.359 --> 01:31:22.199
question is is it better to be economically less efficient
1709
01:31:23.000 --> 01:31:25.520
and provide them with something or is it better to
1710
01:31:25.560 --> 01:31:28.319
be economically more efficient but now have this person be
1711
01:31:28.399 --> 01:31:30.520
a weight on the system. Now do I know the
1712
01:31:30.560 --> 01:31:34.680
specific answer to that question, I'm not sure, Like, obviously
1713
01:31:35.079 --> 01:31:37.640
it depends on the person, but I am saying that
1714
01:31:38.000 --> 01:31:40.479
it's very dependent on whether or not that person is
1715
01:31:40.600 --> 01:31:43.479
provided for, and we have to consider what is the
1716
01:31:43.680 --> 01:31:48.000
economic cost. I'm not even making humane argument, what's the
1717
01:31:48.199 --> 01:31:51.159
economic cost of having people that are desperate? And that
1718
01:31:51.279 --> 01:31:54.239
really ties into our conversation. I'm not even talking about
1719
01:31:54.239 --> 01:31:55.079
a super intelligence.
1720
01:31:55.199 --> 01:31:57.520
We're talking about AI that's going to take a ton
1721
01:31:57.800 --> 01:31:58.439
of jobs.
1722
01:31:58.960 --> 01:32:03.239
Right, It's fantastic because of the economic value, But we
1723
01:32:03.359 --> 01:32:06.479
do need to be careful about how we're balancing what
1724
01:32:06.760 --> 01:32:09.039
negative impact is this going to have in society? Is
1725
01:32:09.760 --> 01:32:11.199
it good that we have all those people out of work.
1726
01:32:11.239 --> 01:32:13.319
I'm not advocating for one thing or another. I am
1727
01:32:13.479 --> 01:32:16.319
advocating that we ask the question right so we can
1728
01:32:16.399 --> 01:32:18.479
get at what is the best way we can go
1729
01:32:18.600 --> 01:32:20.319
about this, because right now we don't know the answer.
1730
01:32:21.199 --> 01:32:23.880
Yes, And I think it is important to ask the question.
1731
01:32:24.279 --> 01:32:28.880
If I were the person who developed the artificial superintelligence
1732
01:32:28.920 --> 01:32:31.960
and this hypothetical scenario that would have this effect, I
1733
01:32:32.000 --> 01:32:34.760
would consider it to be in my rational self interest
1734
01:32:35.239 --> 01:32:41.159
to essentially give other people enough so that they would
1735
01:32:41.239 --> 01:32:45.720
not even object to the arrival of that artificial superintelligence,
1736
01:32:45.760 --> 01:32:48.319
so they would have a standard of living higher than
1737
01:32:48.359 --> 01:32:53.199
they would have had previously, in what economists would call
1738
01:32:53.520 --> 01:32:57.119
a Parareto efficient move. So a Pareto efficient move is
1739
01:32:57.159 --> 01:33:01.079
a move that benefits at least one person and doesn't
1740
01:33:01.199 --> 01:33:05.960
disadvantage any others. But hypothetically, if you have some engine
1741
01:33:06.000 --> 01:33:11.079
of great wealth generation, it should be possible, under some system,
1742
01:33:11.199 --> 01:33:15.760
under some framework, to ensure that everybody benefits from it,
1743
01:33:15.920 --> 01:33:18.680
that those who would be in the default state of
1744
01:33:18.760 --> 01:33:22.680
things net losers from it because they're displaced from their
1745
01:33:22.800 --> 01:33:26.720
jobs or they're otherwise unable to earn an income, they
1746
01:33:26.760 --> 01:33:30.119
could be compensated to some extent that at least makes
1747
01:33:30.199 --> 01:33:33.439
them similarly better off to what they were before. The
1748
01:33:33.560 --> 01:33:38.359
question is what's the kind of political arrangement that could
1749
01:33:38.439 --> 01:33:41.640
make it happen. Is there a will within the society
1750
01:33:41.760 --> 01:33:45.560
to make it happen? What are the logistical difficulties of
1751
01:33:45.640 --> 01:33:48.720
making it happen? AI might help with that. So those
1752
01:33:48.760 --> 01:33:52.000
are all fascinating questions to address. There is, of course,
1753
01:33:52.039 --> 01:33:56.399
a popular science fiction series that does address it, and
1754
01:33:56.520 --> 01:33:59.800
that is Star Trek. Mike Lausine mentions that, so the
1755
01:34:00.000 --> 01:34:03.439
Star Trek economy is one where you have replicators to
1756
01:34:03.800 --> 01:34:09.920
create essentially small scale objects, food items or personal property
1757
01:34:10.520 --> 01:34:14.840
types of items for people to utilize. But people still
1758
01:34:14.880 --> 01:34:17.720
have meaning and purpose because they go off and join
1759
01:34:17.920 --> 01:34:23.239
Starfleet and explore other worlds and meet new civilizations and
1760
01:34:24.079 --> 01:34:29.199
there is this sense of meaning that people find through
1761
01:34:30.520 --> 01:34:35.880
essentially the kind of mentality that you mentioned, scientific discovery
1762
01:34:36.319 --> 01:34:41.039
oriented mentality. So hopefully we can have that kind of future.
1763
01:34:41.600 --> 01:34:47.399
Now we have our friend Zach Varkaris here from Austin Biotech,
1764
01:34:47.479 --> 01:34:50.640
and he has been listening to the conversation on YouTube
1765
01:34:50.680 --> 01:34:53.560
as well, so he has a question for David about
1766
01:34:53.600 --> 01:34:57.039
the human condition and economics. So Zach, please go ahead.
1767
01:34:58.520 --> 01:35:01.960
Thanks, Thanks Reddie, and then highgate this I've bolt to
1768
01:35:02.000 --> 01:35:03.479
my camera's not working very.
1769
01:35:04.920 --> 01:35:06.880
Real, so the question.
1770
01:35:06.920 --> 01:35:08.479
I've been listening to what you're saying, and I think
1771
01:35:08.479 --> 01:35:11.119
it's ironic with the name like primer and you work
1772
01:35:11.199 --> 01:35:14.039
with the numbers. I think it's quite amusing. Crime and numbers.
1773
01:35:14.399 --> 01:35:16.520
I think it's quite amusing. But that's not a hint
1774
01:35:16.520 --> 01:35:17.920
of that. But there's a lot to be said in
1775
01:35:17.960 --> 01:35:20.439
the name. But the thing I was going to ask
1776
01:35:20.479 --> 01:35:22.359
you is this, I listened to a lot of what
1777
01:35:22.479 --> 01:35:26.199
you say. I'm a little bit of an amateurst statistic
1778
01:35:26.319 --> 01:35:29.039
in myself, a little bit. But but the thing is
1779
01:35:29.199 --> 01:35:32.119
is that we talked about economic value, and you rightly
1780
01:35:32.199 --> 01:35:35.399
point out, you know, if someone made prillions of dollars,
1781
01:35:35.479 --> 01:35:38.479
do you think that some of our science of the
1782
01:35:38.720 --> 01:35:41.000
tech industry today would actually.
1783
01:35:40.760 --> 01:35:43.319
Give up their power. And while if they had it,
1784
01:35:43.399 --> 01:35:43.760
and the.
1785
01:35:43.800 --> 01:35:45.720
Answer, we know what the answer is in it it's
1786
01:35:45.840 --> 01:35:49.199
two letters, it's not three, right, So the reality is
1787
01:35:49.279 --> 01:35:52.800
is that is that the Kiven condition is that how
1788
01:35:52.880 --> 01:35:56.560
many times in history has someone created a lot of.
1789
01:35:56.640 --> 01:35:59.159
Power and then given it away? I mean, I'm just
1790
01:35:59.279 --> 01:36:01.079
giving you two very.
1791
01:36:01.279 --> 01:36:05.199
Top line one. One is a guy called Cincinnatus. He
1792
01:36:05.319 --> 01:36:08.920
was he was a Roman sam so he took power
1793
01:36:08.960 --> 01:36:11.359
and then David Away went back to his farm. Another
1794
01:36:11.439 --> 01:36:15.119
one more closely to our time, as the double President Polk.
1795
01:36:15.399 --> 01:36:17.039
He said I'm going to get the West and I'm
1796
01:36:17.079 --> 01:36:19.520
going to retire, and he did that. He rated as
1797
01:36:19.560 --> 01:36:22.319
a as I said, you know, sort of racist or whatever.
1798
01:36:22.399 --> 01:36:23.920
But the reality is he did do what he said
1799
01:36:23.960 --> 01:36:27.359
he was going to do. Would our leaders today, would
1800
01:36:27.399 --> 01:36:32.279
our both industrial and technological and political, would they have
1801
01:36:32.520 --> 01:36:36.560
the courage there, I say, the balls to actually give
1802
01:36:36.680 --> 01:36:39.640
up that power and say we give it back to
1803
01:36:39.760 --> 01:36:40.199
the people.
1804
01:36:40.359 --> 01:36:41.119
And also.
1805
01:36:43.319 --> 01:36:45.960
That's not the anconomic value is more about mental and
1806
01:36:46.039 --> 01:36:47.359
spiritual and physical health.
1807
01:36:47.399 --> 01:36:49.520
And the reason I say that is we tend to
1808
01:36:49.680 --> 01:36:56.840
conflate economics with with what people, and I don't quite.
1809
01:36:56.800 --> 01:37:01.279
Understand why why we do that, and I would work
1810
01:37:01.319 --> 01:37:03.680
to know my opinions on that as Christian.
1811
01:37:05.479 --> 01:37:08.199
Sure, So before I get into that, Jann I had
1812
01:37:08.199 --> 01:37:09.800
a little trouble understanding him.
1813
01:37:09.800 --> 01:37:12.640
So I think I think your first question was around.
1814
01:37:12.560 --> 01:37:15.319
If like this hypothetical person that would have mass all
1815
01:37:15.319 --> 01:37:17.680
his power, would would be able to give it back?
1816
01:37:19.359 --> 01:37:22.760
Jenny were able to what the second question was with
1817
01:37:22.920 --> 01:37:25.680
regards to health and economics.
1818
01:37:25.840 --> 01:37:27.079
I didn't hear that too.
1819
01:37:27.119 --> 01:37:33.279
Well, yeah, so Zach, feel free to repeat if I
1820
01:37:33.399 --> 01:37:37.159
didn't get the gist of it. But essentially, it's about
1821
01:37:37.640 --> 01:37:43.199
the connection between economics and the well being of people.
1822
01:37:43.399 --> 01:37:46.520
So you could see the well being of people from
1823
01:37:46.560 --> 01:37:50.159
the standpoint of quality of life or health outcomes or
1824
01:37:51.159 --> 01:37:54.720
Zach's own term which is freuden shpahna or joy span,
1825
01:37:55.520 --> 01:38:01.560
which he tries to pursue with his patience. So essentially,
1826
01:38:02.840 --> 01:38:07.159
is there a way to connect an economic system to
1827
01:38:07.880 --> 01:38:13.640
those kinds of outcomes so that they are necessarily fulfilled
1828
01:38:13.840 --> 01:38:15.279
as an economy grows?
1829
01:38:16.720 --> 01:38:21.359
Okay, all right, well, okay, that's a lot. So I'll
1830
01:38:21.359 --> 01:38:23.840
start with the first one with regards to giving back.
1831
01:38:24.000 --> 01:38:27.159
So in the context that we were discussing with having
1832
01:38:27.199 --> 01:38:30.319
the super intelligence, I would think given giving back means
1833
01:38:30.439 --> 01:38:34.840
turning off large segments of debt. Super intelligent model, right,
1834
01:38:34.960 --> 01:38:39.239
So like giving back could mean that people can And
1835
01:38:39.600 --> 01:38:40.960
I want to go back to the sense of purpose,
1836
01:38:41.000 --> 01:38:42.439
which actually ties to your second point.
1837
01:38:43.880 --> 01:38:49.119
I think it goes beyond just having money. And I
1838
01:38:49.159 --> 01:38:50.880
don't mean this in a materialistic way.
1839
01:38:51.960 --> 01:38:53.359
I think that we have.
1840
01:38:55.119 --> 01:38:58.159
In an there's an innate part of human nature where
1841
01:38:58.159 --> 01:38:59.720
how I'm turned my light on it looks like I'm
1842
01:39:00.000 --> 01:39:08.399
I think so, I think that in human nature there
1843
01:39:08.520 --> 01:39:14.239
is an innate part of us that we want to contribute,
1844
01:39:14.920 --> 01:39:17.159
and of course there's different ways to that we can contribute.
1845
01:39:17.159 --> 01:39:19.680
But I mean like in a daily in a daily sense.
1846
01:39:19.960 --> 01:39:21.560
I don't know if you know anybody like this, but
1847
01:39:21.800 --> 01:39:26.439
I know people that were tired and died shortly thereafter,
1848
01:39:27.000 --> 01:39:29.800
and not because of horrible health, like they didn't have
1849
01:39:29.920 --> 01:39:33.880
like a crazy disease, They weren't very depressed. It was
1850
01:39:34.000 --> 01:39:37.159
because I think it's a fundamental component of the human
1851
01:39:37.239 --> 01:39:40.600
condition that we want to be challenged. This is why
1852
01:39:40.840 --> 01:39:42.840
if there's five things going on in your life that
1853
01:39:42.880 --> 01:39:46.760
are good and one that is not, your brain will gravitate.
1854
01:39:46.399 --> 01:39:47.039
To one that is not.
1855
01:39:47.239 --> 01:39:50.319
And it's not because we're negative. It's actually a survival instinct.
1856
01:39:50.600 --> 01:39:53.039
If we go back fifty thousand years and you're in
1857
01:39:53.079 --> 01:39:56.920
a cave and there's a line that's about to eat you, right,
1858
01:39:57.279 --> 01:39:59.079
and you and you haven't eaten.
1859
01:39:59.199 --> 01:40:02.479
In three days, but you're really warm, You're not like,
1860
01:40:02.840 --> 01:40:03.119
you know what.
1861
01:40:03.439 --> 01:40:05.399
I'm actually like, I don't even to worry about this
1862
01:40:05.439 --> 01:40:08.199
line right now because I'm really warm, it's really comfortable.
1863
01:40:08.199 --> 01:40:09.640
You'd be like, oh my god, I need to run
1864
01:40:09.680 --> 01:40:11.359
as fast as I can. I'm gonna get eaten, and
1865
01:40:11.399 --> 01:40:12.479
then I need to find some food.
1866
01:40:12.720 --> 01:40:12.840
Right.
1867
01:40:12.920 --> 01:40:15.319
So it's a part of our human condition to be
1868
01:40:15.600 --> 01:40:19.199
challenged to be dissatisfied and to push to be comfortable.
1869
01:40:19.520 --> 01:40:25.199
So if we're talking about this fantastical utopian scenario where
1870
01:40:26.000 --> 01:40:31.279
everyone has everything provided for them, I can't guarantee anything.
1871
01:40:31.640 --> 01:40:33.640
I'll just make it a statement of fact. I think
1872
01:40:34.039 --> 01:40:34.680
could be wrong.
1873
01:40:35.279 --> 01:40:39.359
I think that people would still be unhappy because I
1874
01:40:39.439 --> 01:40:43.359
think people are at their happiest when we are making
1875
01:40:43.600 --> 01:40:47.359
progress and when we feel that we're contributing to something
1876
01:40:48.000 --> 01:40:50.239
like that that is larger than ourselves. So so I'm
1877
01:40:50.239 --> 01:40:52.880
not being paid for it. It's about having my identity
1878
01:40:52.960 --> 01:40:55.239
being like, okay, I am a doctor, I.
1879
01:40:56.039 --> 01:40:56.560
Am a one.
1880
01:40:57.600 --> 01:41:01.960
Sorry, go ahead, no, keep there was just okay, okay.
1881
01:41:02.079 --> 01:41:05.960
So I think that there and and I'm going to
1882
01:41:06.000 --> 01:41:08.079
go as far to say, there's economic value in that
1883
01:41:09.119 --> 01:41:11.800
people are going to be happier, and there's there's value
1884
01:41:11.840 --> 01:41:13.439
to their being happier. And I'll tie that to the
1885
01:41:13.479 --> 01:41:16.199
second point that you made, Zach, that I.
1886
01:41:16.279 --> 01:41:17.319
Think tying.
1887
01:41:18.680 --> 01:41:24.720
Economic output to happiness is something that can I mean,
1888
01:41:24.880 --> 01:41:27.239
those are those are kind of different values. Like, I
1889
01:41:27.319 --> 01:41:28.920
thought it would be really cool if at some point
1890
01:41:29.000 --> 01:41:32.039
we can actually quantify something like the amount of dopamine
1891
01:41:32.039 --> 01:41:35.760
that someone receives. So when you have a I remember
1892
01:41:35.800 --> 01:41:38.479
reading a statistic that when you are laughing really hard
1893
01:41:38.560 --> 01:41:40.720
with a friend for two minutes, you get the same
1894
01:41:40.840 --> 01:41:43.479
level of dopamine that you get from smoking roughly one.
1895
01:41:43.439 --> 01:41:44.359
Hour of marijuana.
1896
01:41:45.000 --> 01:41:48.760
I think that that's that's I forget the exact study,
1897
01:41:48.800 --> 01:41:50.159
but you can see like the vibe of that. So
1898
01:41:50.880 --> 01:41:53.199
it'd be interesting if we could quantify the amount of dopamine.
1899
01:41:53.199 --> 01:41:57.119
I don't know if dopamine captures we're talking about what
1900
01:41:57.239 --> 01:41:58.039
it is for someone.
1901
01:41:57.920 --> 01:41:58.439
To be happy.
1902
01:41:58.479 --> 01:42:00.840
I think there's different kind of levels that it's a
1903
01:42:00.920 --> 01:42:02.960
very complex term. So I don't know if go pu
1904
01:42:03.039 --> 01:42:06.159
mean exactly encompasses all of what it is to be happy.
1905
01:42:06.600 --> 01:42:09.000
But if we could quantify that, I think it would
1906
01:42:09.119 --> 01:42:12.760
absolutely be correlated with economic output, but it's not going
1907
01:42:12.800 --> 01:42:15.239
to be exactly one to one. And I think what
1908
01:42:15.399 --> 01:42:17.720
kind of makes sense and hopefully we've hit an inflection
1909
01:42:17.840 --> 01:42:20.479
point with AI is we're at this point that our
1910
01:42:20.800 --> 01:42:27.359
total collective economic output can still massively increase. We still
1911
01:42:27.479 --> 01:42:30.840
haven't yet figured out how do we turn that economic
1912
01:42:30.960 --> 01:42:35.840
output into having people be as happy as possible, right,
1913
01:42:35.880 --> 01:42:37.720
And that's a balance and we feel that, you know,
1914
01:42:37.800 --> 01:42:41.039
we feel that kind of tension politically where it's like, Okay,
1915
01:42:41.159 --> 01:42:43.760
well do we give money to people that are doing
1916
01:42:43.840 --> 01:42:46.600
as well or do we prioritize economic output. That's a
1917
01:42:46.640 --> 01:42:49.439
really hard question for us to ask to answer because
1918
01:42:49.439 --> 01:42:51.680
there's so much growth in either direction. And I think
1919
01:42:51.760 --> 01:42:53.560
just kind of what's a little bit different about where
1920
01:42:53.640 --> 01:42:56.119
this conversation is gone is that if we're talking about
1921
01:42:56.159 --> 01:43:00.159
the superintelligence and we're talking about a place where where
1922
01:43:00.560 --> 01:43:03.039
we've really hit this place of diminishing returns, where it's
1923
01:43:03.079 --> 01:43:07.720
harder to find a significant increase in economic value because
1924
01:43:07.760 --> 01:43:10.520
we're so efficient, then I would think.
1925
01:43:10.479 --> 01:43:13.319
The only thing that's really left is to make people happier.
1926
01:43:13.439 --> 01:43:17.039
So when I say, like, is less economically efficient, Yeah,
1927
01:43:17.359 --> 01:43:19.840
but it's not like to an extent than it matters.
1928
01:43:19.840 --> 01:43:21.600
So it's like if you have a trillion dollars and
1929
01:43:21.680 --> 01:43:24.000
I'm saying, okay, but instead of a trillion dollars, you
1930
01:43:24.159 --> 01:43:26.600
have nine hundred and ninety nine billion dollars. And by
1931
01:43:26.600 --> 01:43:28.760
the way, there's a one billion dollar difference between those
1932
01:43:28.800 --> 01:43:34.199
two things. Right, Like that could be a better answer overall,
1933
01:43:34.520 --> 01:43:37.319
depending on how you want to define a best answer
1934
01:43:37.760 --> 01:43:39.239
and to Noati. The other thing I want to mention
1935
01:43:39.399 --> 01:43:41.039
to kind of get on what you mentioned about science.
1936
01:43:42.239 --> 01:43:45.479
I think we have to acknowledge that a not everybody
1937
01:43:45.560 --> 01:43:48.199
is going to want to be invested in science. I
1938
01:43:48.319 --> 01:43:50.479
will one hundred percent believe that it should be glorified
1939
01:43:50.520 --> 01:43:54.279
more and there should be more attention paid to it.
1940
01:43:54.520 --> 01:43:56.439
So one thing I do with my family is that
1941
01:43:56.720 --> 01:43:58.760
we now my kids are too young to really know
1942
01:43:58.920 --> 01:44:01.680
that they'll start to learn it is we kind of
1943
01:44:01.760 --> 01:44:03.359
like they have the oscars. We do something like that
1944
01:44:03.439 --> 01:44:06.039
with the Nobel prize winners where we talk about who
1945
01:44:06.119 --> 01:44:07.800
won the Nobel Prize. And what's so great about that
1946
01:44:08.239 --> 01:44:10.640
is that with science, this is what I love about it.
1947
01:44:11.760 --> 01:44:16.119
We're moving forward and it's all good news. Like there's
1948
01:44:16.119 --> 01:44:17.920
so many things that happened in the world that are terrible.
1949
01:44:18.159 --> 01:44:21.039
When things come out in science, I mean generally like
1950
01:44:21.439 --> 01:44:23.479
scientific breakthroughs, it's good news.
1951
01:44:23.560 --> 01:44:25.479
Oh we found the god particle.
1952
01:44:25.840 --> 01:44:26.840
Oh we had a.
1953
01:44:26.880 --> 01:44:29.680
Reaction where it generated more energy than what we put
1954
01:44:29.720 --> 01:44:31.760
into it. Right, these are exciting things to talk about.
1955
01:44:31.960 --> 01:44:35.119
So I think we need to glorify science, but we
1956
01:44:35.239 --> 01:44:37.960
don't want everybody doesn't want to be participating in science,
1957
01:44:38.239 --> 01:44:41.159
and we don't want everybody participating in science, right, I
1958
01:44:41.199 --> 01:44:44.319
mean just some people that just maybe maybe it's just
1959
01:44:44.359 --> 01:44:46.800
a character thing, maybe some people we don't trust. Right,
1960
01:44:47.079 --> 01:44:49.520
by the way, in science, and I'm not taking away
1961
01:44:49.560 --> 01:44:53.640
from scientists, it can be easy to fake data.
1962
01:44:53.960 --> 01:44:56.279
I worked in a lab where the raw data was
1963
01:44:56.319 --> 01:44:57.039
a Western blot.
1964
01:44:57.119 --> 01:44:59.560
If we've ever seen a Western blot, it's basically invented
1965
01:44:59.600 --> 01:45:02.720
in the night teen sixties, like imagine and I'm not
1966
01:45:02.800 --> 01:45:06.640
even exaggerating, Like a smudge of ink is the data, right,
1967
01:45:06.720 --> 01:45:09.119
So you get these like five little columns, and it's
1968
01:45:09.119 --> 01:45:11.319
showing kind of like how big the proteins are, if
1969
01:45:11.319 --> 01:45:13.680
the protein is, if a protein is present. I won't
1970
01:45:13.720 --> 01:45:15.560
get more detailed than that. I just want you to
1971
01:45:15.640 --> 01:45:17.640
have in your head of picture where the data that
1972
01:45:17.800 --> 01:45:21.159
is being published is these five like little strips, and
1973
01:45:21.239 --> 01:45:23.279
each strip has a smudge of ink on it. You
1974
01:45:23.319 --> 01:45:26.000
know how easy it would be for someone to say, hey,
1975
01:45:26.319 --> 01:45:28.840
I saw a smudge of ink, when in reality they didn't.
1976
01:45:28.880 --> 01:45:31.680
They just photoshopped it in. Especially with AI. Now you
1977
01:45:31.760 --> 01:45:34.039
can make that thing wherever you want. So we want
1978
01:45:34.079 --> 01:45:37.359
to make sure that people are of the correct character
1979
01:45:37.800 --> 01:45:41.359
to be participating in science because it's important, right. So
1980
01:45:41.800 --> 01:45:44.800
I still think that people are going to want a
1981
01:45:47.000 --> 01:45:49.840
people are going to want a sense of purpose, want
1982
01:45:49.920 --> 01:45:52.520
a way to contribute independent of like being paid a
1983
01:45:52.520 --> 01:45:54.079
lot of money for it, because I agree that it
1984
01:45:54.119 --> 01:45:56.079
shouldn't just be a round of I feel good because
1985
01:45:56.079 --> 01:45:58.720
I'm being paid money. I'm saying being contributing in a
1986
01:45:58.720 --> 01:45:59.319
particular way.
1987
01:45:59.359 --> 01:46:03.159
So I'll speak to myself. I love building models. I
1988
01:46:03.279 --> 01:46:04.680
love it because I love math.
1989
01:46:05.039 --> 01:46:07.399
I love the models provide value, and I love the
1990
01:46:07.479 --> 01:46:09.800
fact that I'm actually building something that like after it's done,
1991
01:46:09.840 --> 01:46:11.880
it can operate on its own. Sometimes I'll name the
1992
01:46:11.960 --> 01:46:14.000
models that I'm building so that like I like short
1993
01:46:14.079 --> 01:46:20.960
short ways, I cannot stand creating like visual art. I'm colorblind.
1994
01:46:21.239 --> 01:46:23.399
I've ever been good at it. I don't value it.
1995
01:46:24.000 --> 01:46:25.760
So like if someone said, hey, let's make a painting,
1996
01:46:25.840 --> 01:46:29.359
I say, I would really really rather I would literally
1997
01:46:29.520 --> 01:46:32.239
pay to not paint something because I just I really
1998
01:46:32.319 --> 01:46:35.079
dislike it. Right, I'm not saying painting's worthless somebody. It
1999
01:46:35.159 --> 01:46:37.359
gives value to some people, there's more power to I
2000
01:46:37.359 --> 01:46:39.199
don't want to be participating in that. And I think
2001
01:46:39.239 --> 01:46:42.119
that everyone feels some way about different things. So in
2002
01:46:42.159 --> 01:46:44.800
a perfect society, I would love it if they said, David,
2003
01:46:44.880 --> 01:46:47.159
look we've took taken care of everything. You don't have
2004
01:46:47.199 --> 01:46:49.479
to worry about food security, you got everything. I would
2005
01:46:49.560 --> 01:46:53.079
still say, please give me some way that I can
2006
01:46:53.279 --> 01:46:55.319
build something for you. I want to build some kind
2007
01:46:55.359 --> 01:46:57.319
of math equation that's going to improve your life. Don't
2008
01:46:57.319 --> 01:46:59.319
worry about paying me anything. I want to have that
2009
01:46:59.439 --> 01:47:02.800
sense of purpose recognizing that yes, maybe the super intelligence
2010
01:47:02.840 --> 01:47:04.560
thing can do that, but I'm gonna be a happier person,
2011
01:47:04.840 --> 01:47:05.560
and which means.
2012
01:47:05.359 --> 01:47:06.199
I'm going to be a better father.
2013
01:47:06.239 --> 01:47:07.560
I'm gonna be a better husband, I'm gonna be a
2014
01:47:07.560 --> 01:47:09.880
better friend, I'm gonna be a better brother, community member.
2015
01:47:10.239 --> 01:47:14.680
If I have some place where I feel I'm giving
2016
01:47:14.720 --> 01:47:17.640
in a meaningful way, and so who cares. It's a
2017
01:47:17.680 --> 01:47:20.800
super you know, the hyper super if the artificial superintelligence
2018
01:47:20.880 --> 01:47:23.560
was able to do it. If I'm able to contribute
2019
01:47:23.600 --> 01:47:26.760
something and I'm not taking away significantly from the economy,
2020
01:47:27.119 --> 01:47:29.399
I think it's worthwhile. If like, I'm not the only
2021
01:47:29.439 --> 01:47:32.039
person that's doing that. If we do that for like people,
2022
01:47:33.000 --> 01:47:35.680
then we can be contributing, we can be productive, and
2023
01:47:35.760 --> 01:47:38.520
will be moving towards this like utopian society.
2024
01:47:39.600 --> 01:47:43.199
Yes, thank you very much for that reply. David and
2025
01:47:44.039 --> 01:47:48.319
Zach expressed agreement with you in our chat. He also
2026
01:47:48.520 --> 01:47:54.319
says stagnation is death and figuring out ways to maximize
2027
01:47:54.399 --> 01:47:57.600
the happiness of patients is what he and his team
2028
01:47:57.720 --> 01:48:02.520
does an Austin Biotech so Zach was our guest on
2029
01:48:03.079 --> 01:48:07.199
multiple prior Virtual Enlightenment salons as well, and he notes, yes,
2030
01:48:07.319 --> 01:48:10.479
Western blots can totally be faked.
2031
01:48:11.079 --> 01:48:13.840
Actually, the example of.
2032
01:48:15.319 --> 01:48:20.680
People who retire and then die shortly thereafter is epitomized
2033
01:48:20.800 --> 01:48:26.000
by James Knox Polk, the President of the United States,
2034
01:48:26.199 --> 01:48:31.600
whom Zach referred to if you look at his Wikipedia page.
2035
01:48:31.880 --> 01:48:38.079
He ceased to be president in March of eighteen forty nine.
2036
01:48:38.159 --> 01:48:42.399
He died on June fifteenth, eighteen forty nine, and he
2037
01:48:42.520 --> 01:48:48.399
most likely had underlying health conditions, but he decided not
2038
01:48:48.520 --> 01:48:51.319
to run for a second term, and he was dead
2039
01:48:51.720 --> 01:48:56.079
essentially three months after the expiration of his first term.
2040
01:48:56.159 --> 01:48:58.680
So it is essential for humans to have a sense
2041
01:48:58.720 --> 01:49:02.800
of meaning and purpose and to continue to be productive
2042
01:49:02.880 --> 01:49:06.520
in some way that is personally fulfilling, whether or not
2043
01:49:06.680 --> 01:49:09.720
it makes money. So if one does retire, one should
2044
01:49:09.760 --> 01:49:15.119
continue working on personal projects, watching salons like these, participating
2045
01:49:15.359 --> 01:49:20.760
in stimulating, thought provoking conversations. And I think that's an
2046
01:49:20.760 --> 01:49:25.760
important point to keep in mind as our technological capabilities advance.
2047
01:49:26.279 --> 01:49:29.800
Now we only have about four and a half minutes
2048
01:49:29.880 --> 01:49:33.720
left in our salon. I'm hoping that our tramone might
2049
01:49:33.840 --> 01:49:37.199
have some time to ask a brief question or make
2050
01:49:37.439 --> 01:49:41.159
a brief comment in that remaining time. So ur t ramone,
2051
01:49:41.199 --> 01:49:41.880
please go ahead.
2052
01:49:43.319 --> 01:49:48.039
Yeah, when you started off inventional narrative, and that just
2053
01:49:48.119 --> 01:49:50.279
got me thinking of an episode of The Wi Files
2054
01:49:50.399 --> 01:49:54.119
when they he was talking about syops, So how to
2055
01:49:54.279 --> 01:49:59.520
identify syops like kind of you know, triggered that recall.
2056
01:50:00.199 --> 01:50:03.640
And you know, it's very convenient that as the population
2057
01:50:03.800 --> 01:50:07.520
of people global globally is going down, the robots are
2058
01:50:07.640 --> 01:50:11.079
getting better. It's going to make it very easy to
2059
01:50:11.199 --> 01:50:13.680
replace a lot of the manual labor with the robots
2060
01:50:13.680 --> 01:50:17.800
that are advancing. So and also some people consider that
2061
01:50:18.720 --> 01:50:25.000
possibly conspiracy, that something's behind that helping the population world
2062
01:50:25.279 --> 01:50:29.359
get reduced, you know, lovers, firm count and whatnot.
2063
01:50:30.319 --> 01:50:31.560
So what are your thoughts about.
2064
01:50:34.399 --> 01:50:39.039
You mean about like if robots, so Hans, I heard
2065
01:50:39.039 --> 01:50:41.119
two very different things there. I heard about like robots
2066
01:50:41.159 --> 01:50:43.439
are taking jobs, right, and the cause of it that's
2067
01:50:43.920 --> 01:50:46.760
heard psyops and the siops that like psyche is that.
2068
01:50:48.800 --> 01:50:53.159
When it's a government, uh uh, what you call propaganda,
2069
01:50:53.319 --> 01:50:55.279
it's a sort of propaganda about the government in order
2070
01:50:55.359 --> 01:50:58.880
to write that narrative and get people to believe. So
2071
01:50:58.960 --> 01:51:02.920
it's say the now why you would invade another country,
2072
01:51:03.279 --> 01:51:06.000
So you would have to convince the population through the
2073
01:51:06.079 --> 01:51:08.640
propaganda that would be put out through the news news media,
2074
01:51:09.279 --> 01:51:13.520
and basically that would be what convinces people the narrative
2075
01:51:14.000 --> 01:51:17.760
that they're crafting to you know, shape it in people's minds,
2076
01:51:17.800 --> 01:51:18.920
so they're accepting of it.
2077
01:51:19.560 --> 01:51:21.520
So, uh, that's socie.
2078
01:51:23.199 --> 01:51:25.560
So so you're saying the propaganda is saying that there
2079
01:51:25.600 --> 01:51:27.680
are robots that are taking people's jobs, and that would
2080
01:51:27.760 --> 01:51:29.680
make people want to that that.
2081
01:51:29.720 --> 01:51:33.319
Would be a the more the conspiracy theory, U narrative.
2082
01:51:34.439 --> 01:51:38.680
But it's just that your your narratives sure had to
2083
01:51:38.760 --> 01:51:42.399
recall the particular episode of the wi fis sure.
2084
01:51:42.720 --> 01:51:47.880
Yeah, so I mean, yeah, so I think that, Uh,
2085
01:51:48.279 --> 01:51:54.159
I think robots are taking some people's jobs. I don't
2086
01:51:54.159 --> 01:51:58.359
think that's a conspiracy. I think that it is beneficial
2087
01:51:58.479 --> 01:52:02.079
to most people when it happens. So like your car
2088
01:52:02.640 --> 01:52:06.760
is less expensive because a robot is contributing, if not
2089
01:52:06.880 --> 01:52:08.880
doing the whole thing and putting it together because it's
2090
01:52:09.159 --> 01:52:12.680
it costs the company less to produce that robot. That's
2091
01:52:12.720 --> 01:52:15.359
good for the consumer, right, the product's cheaper. It's not
2092
01:52:15.520 --> 01:52:17.840
good for the person whose job it was to like
2093
01:52:18.000 --> 01:52:20.319
hammer in the parts to the car. I think both
2094
01:52:20.359 --> 01:52:24.079
are true. I think capitalism was the force behind that.
2095
01:52:24.119 --> 01:52:26.039
I don't think it was a conspiracy. Like I don't
2096
01:52:26.079 --> 01:52:29.199
think that was a conspiracy theory. And I think I
2097
01:52:29.279 --> 01:52:31.039
mean I think where you're going with it, which is
2098
01:52:31.119 --> 01:52:36.000
that like in general, and this is what I think
2099
01:52:36.119 --> 01:52:39.880
is interesting about today's day and age. We used to
2100
01:52:40.000 --> 01:52:44.079
be more prone to propaganda because we did not have
2101
01:52:44.359 --> 01:52:46.840
access to the information. Right. So if you think about
2102
01:52:46.920 --> 01:52:50.520
things like the news, the only way that you could
2103
01:52:50.520 --> 01:52:52.319
see the news, like let's say there was some court
2104
01:52:52.399 --> 01:52:56.439
case that was going on, is every night for thirty minutes.
2105
01:52:56.520 --> 01:52:58.039
Now this is I mean, I don't know this was
2106
01:52:58.119 --> 01:52:59.640
this was CNN was around when I was a kid.
2107
01:52:59.640 --> 01:53:01.560
This is up to like when my parents were kids
2108
01:53:01.600 --> 01:53:04.199
in that generation, Like you had the news for thirty
2109
01:53:04.199 --> 01:53:06.199
minutes a night, maybe twice a night, eight thirty and
2110
01:53:06.239 --> 01:53:07.920
ten thirty or something, where they would give a recap
2111
01:53:08.000 --> 01:53:09.800
as to what happened, and that was really your access.
2112
01:53:09.840 --> 01:53:11.800
There was no Internet. I mean maybe there was radio
2113
01:53:11.960 --> 01:53:14.199
that had that as well, Right, you were limited to that.
2114
01:53:14.399 --> 01:53:15.960
So if you want to know what happened in the courthouse,
2115
01:53:15.960 --> 01:53:18.319
you had two choices. You could physically go down to
2116
01:53:18.359 --> 01:53:19.840
the courthouse and try to get your hands on the
2117
01:53:19.880 --> 01:53:22.880
physical papers that had like oh, here here's the indictment
2118
01:53:23.000 --> 01:53:25.399
and here's the testimony, and here's everything that happened, or
2119
01:53:25.600 --> 01:53:28.119
what's more realistic is you sit in your couch and
2120
01:53:28.199 --> 01:53:30.239
you hear what they told you happened, right, So you're
2121
01:53:30.319 --> 01:53:33.159
very prone to propaganda. What's interesting about today's day and
2122
01:53:33.159 --> 01:53:36.039
age is that we have access to the primary sources.
2123
01:53:36.399 --> 01:53:39.479
You can read indictments, you could read testimony, you could
2124
01:53:39.560 --> 01:53:40.479
read studies.
2125
01:53:40.720 --> 01:53:42.039
You don't need to rely on.
2126
01:53:44.039 --> 01:53:46.359
These, you know, third parties to tell you what's going on.
2127
01:53:46.760 --> 01:53:49.520
So if there's some entity be it. It could be
2128
01:53:49.560 --> 01:53:51.680
a government, it could be a business, it could be media,
2129
01:53:51.920 --> 01:53:53.800
it could be your friend, doesn't matter. If there's some
2130
01:53:54.119 --> 01:53:57.399
entity that is trying to push a narrative on you,
2131
01:53:57.840 --> 01:54:03.079
it is so easy to just for ten minutes do
2132
01:54:03.279 --> 01:54:08.800
some research to see, well, what does the primary literature say, right,
2133
01:54:09.239 --> 01:54:11.479
and if you have trouble understanding it, you can have
2134
01:54:11.600 --> 01:54:12.640
chatbots help you to learn.
2135
01:54:12.720 --> 01:54:14.880
I actually think that's the best use of chatbots.
2136
01:54:14.600 --> 01:54:18.319
Is for education, not for production, and then and then
2137
01:54:18.319 --> 01:54:20.279
you can go about it. So I think that, like,
2138
01:54:20.399 --> 01:54:21.760
I think that's a really good point that when you're
2139
01:54:21.760 --> 01:54:24.039
fed a narrative, you should just be alert.
2140
01:54:24.159 --> 01:54:26.479
That's a narrative. This is helping you feel better. That's fine.
2141
01:54:26.520 --> 01:54:28.359
Going back to my talk, like bringing it full circle,
2142
01:54:28.800 --> 01:54:31.159
if it's meant for the purposes of information, she'd be like,
2143
01:54:31.319 --> 01:54:34.399
wait a minute, that's a statement. Let's make let's turn
2144
01:54:34.439 --> 01:54:37.159
that statement into a question, into a hypothesis, And I'm
2145
01:54:37.199 --> 01:54:40.199
going to go about like proving or disproving that hypothesis
2146
01:54:40.239 --> 01:54:42.800
as objectively as I can so you can obtain truth
2147
01:54:42.880 --> 01:54:43.439
and move forward.
2148
01:54:44.880 --> 01:54:48.439
Yes, thank you very much for that response, David, and
2149
01:54:48.760 --> 01:54:52.520
for bringing the discussion full circle. So this has been
2150
01:54:52.600 --> 01:54:56.039
such a fascinating conversation. The time has just flown by,
2151
01:54:56.680 --> 01:55:00.520
and just because of our time constraint, we will need
2152
01:55:00.640 --> 01:55:03.920
to conclude our virtual Enlightenment salon. But I do want
2153
01:55:04.159 --> 01:55:07.800
to once again recommend that our viewers visit the website
2154
01:55:07.840 --> 01:55:13.279
of Strata quant and particularly to listen to the episodes
2155
01:55:13.319 --> 01:55:17.479
of profit as a Science that are available there. Now,
2156
01:55:18.039 --> 01:55:21.760
we have two questions about my past. Very briefly, did
2157
01:55:21.800 --> 01:55:24.199
I wear bow ties at fourteen?
2158
01:55:25.840 --> 01:55:26.039
Yes?
2159
01:55:26.159 --> 01:55:27.560
Next question yes?
2160
01:55:27.760 --> 01:55:28.039
Okay.
2161
01:55:29.119 --> 01:55:31.560
Did I seem like the type who would be involved
2162
01:55:31.600 --> 01:55:34.520
in politics back then? Yes?
2163
01:55:34.720 --> 01:55:37.760
Next question okay, those were I believe? Correct me if
2164
01:55:37.760 --> 01:55:39.800
I'm wrong, Correct me if I'm wrong. He was the
2165
01:55:39.880 --> 01:55:44.800
president of her student council the freshman year. Freshman year,
2166
01:55:44.880 --> 01:55:45.840
I reelected he.
2167
01:55:45.920 --> 01:55:46.800
Was involved in college.
2168
01:55:47.000 --> 01:55:49.479
I also distinctly remember that he made the R and
2169
01:55:49.520 --> 01:55:51.600
Brascal argumentator that is.
2170
01:55:51.640 --> 01:55:53.560
An artifact for when he was in high school.
2171
01:55:53.600 --> 01:55:57.000
That was a long, long time ago, So one hundred
2172
01:55:57.239 --> 01:56:00.520
everything we're talking about here, Jannati has been on that
2173
01:56:00.720 --> 01:56:03.880
train from the day I met him when we were fourteen.
2174
01:56:05.399 --> 01:56:08.640
So yes, all those questions were easy.
2175
01:56:09.319 --> 01:56:13.640
Yes, indeed, Well, thank you David for providing that confirmation.
2176
01:56:14.159 --> 01:56:16.119
We were thrilled to have you as a guest on
2177
01:56:16.239 --> 01:56:19.960
our Virtual Enlightenment Salon today and I think you and
2178
01:56:20.079 --> 01:56:22.439
your insights are going to help us all to live
2179
01:56:22.520 --> 01:56:23.880
long and prosper