TRANSCRIPT 🔗
1
00:00:00,000 --> 00:00:08,320
Hello, and welcome to the AI Insiders, the flagship podcast of AI Insiders everywhere.
2
00:00:08,320 --> 00:00:09,840
And who are these AI Insiders?
3
00:00:09,840 --> 00:00:11,480
Well, we all are.
4
00:00:11,480 --> 00:00:16,360
We're all becoming AI Insiders as we find ourselves in this increasingly AI imbued world
5
00:00:16,360 --> 00:00:20,120
facing many possible AI imbued futures.
6
00:00:20,120 --> 00:00:23,560
Where appropriately, more and more of us are realizing that we all have a vested interest
7
00:00:23,560 --> 00:00:28,560
in making sure that we manage to get AI to help us realize the best of all these possible
8
00:00:28,560 --> 00:00:29,560
worlds.
9
00:00:29,560 --> 00:00:30,560
Hi, I'm Adam Russell.
10
00:00:30,560 --> 00:00:34,840
I'm the director of the AI division at University of Southern California's Information Sciences
11
00:00:34,840 --> 00:00:35,840
Institute.
12
00:00:35,840 --> 00:00:37,400
And I'm your podcast host.
13
00:00:37,400 --> 00:00:42,560
I'm serving as your virtual guide as we talk to some other AI Insiders here at ISI to get
14
00:00:42,560 --> 00:00:44,640
their sense of where things are headed.
15
00:00:44,640 --> 00:00:46,960
And as a result, what excites them the most?
16
00:00:46,960 --> 00:00:48,320
What worries them the most?
17
00:00:48,320 --> 00:00:50,280
What might confuse them the most?
18
00:00:50,280 --> 00:00:54,720
Because sometimes it feels like if you're not confused, you're not really thinking clearly.
19
00:00:54,720 --> 00:00:58,840
Let's call it informed confusion or infomution.
20
00:00:58,840 --> 00:01:03,120
So today I'm pleased to introduce another AI Insider who help ensure that I'm also sufficiently
21
00:01:03,120 --> 00:01:04,120
infomused.
22
00:01:04,120 --> 00:01:05,120
Amandeep Singh.
23
00:01:05,120 --> 00:01:09,120
Amandeep is a research programmer here at ISI.
24
00:01:09,120 --> 00:01:13,160
He's been here, I think for almost 10 years at this point.
25
00:01:13,160 --> 00:01:15,720
Amandeep will correct me, but let's get started.
26
00:01:15,720 --> 00:01:17,880
Amandeep, thank you for joining us on AI Insiders.
27
00:01:17,880 --> 00:01:19,680
Thanks for having me.
28
00:01:19,680 --> 00:01:20,680
All right.
29
00:01:20,680 --> 00:01:21,680
Have I got that right?
30
00:01:21,680 --> 00:01:23,680
How long have you been at ISI?
31
00:01:23,680 --> 00:01:25,560
Joined in 2015, it's almost 10 years.
32
00:01:25,560 --> 00:01:26,560
All right.
33
00:01:26,560 --> 00:01:29,280
So that is not that bad as a social scientist, excellent.
34
00:01:29,280 --> 00:01:30,280
Okay.
35
00:01:30,280 --> 00:01:32,520
So let's start with the vaguely insane questions.
36
00:01:32,520 --> 00:01:34,840
I am your 12 year old self.
37
00:01:34,840 --> 00:01:35,840
Work with me here.
38
00:01:35,840 --> 00:01:42,560
I would love for you to explain to me slash you, what is it you're doing now at ISI and
39
00:01:42,560 --> 00:01:46,800
in AI in particular, and why does it matter?
40
00:01:46,800 --> 00:01:55,840
So 12 year old myself would be mind blown if I told him that I work for and teach at
41
00:01:55,840 --> 00:02:04,120
the same university as one of the richest person in India goes to.
42
00:02:04,120 --> 00:02:07,280
His mind would be blown like, okay, you're just talking out of your mind.
43
00:02:07,280 --> 00:02:08,360
That's not true.
44
00:02:08,360 --> 00:02:15,240
But a 12 year old now, I would tell him that I work in building those knowledge graphs,
45
00:02:15,240 --> 00:02:23,400
which are powering things like Alexa, the AI assistant they're so used to talk to now.
46
00:02:23,400 --> 00:02:31,360
So we here at ISI help build those knowledge graphs, which power those AI assistants.
47
00:02:31,360 --> 00:02:32,360
Okay.
48
00:02:32,360 --> 00:02:35,520
So your 12 year old self probably would know this.
49
00:02:35,520 --> 00:02:38,000
Let's pretend it's my 12 year old self.
50
00:02:38,000 --> 00:02:39,520
What do you mean knowledge graph?
51
00:02:39,520 --> 00:02:41,600
Explain to me like why that's important.
52
00:02:41,600 --> 00:02:45,760
So let's say we have knowledge coming in from all different sources.
53
00:02:45,760 --> 00:02:48,400
So knowledge is just some facts.
54
00:02:48,400 --> 00:02:51,040
Now we build a graph.
55
00:02:51,040 --> 00:02:56,920
A graph is an interconnected system where entities are linked to each other.
56
00:02:56,920 --> 00:03:01,880
And by entities, I can mean like any person or a place or an event.
57
00:03:01,880 --> 00:03:08,240
So we build those graphs so that these entities connect to each other and we can find information
58
00:03:08,240 --> 00:03:12,160
by just following the links from one to another.
59
00:03:12,160 --> 00:03:14,240
So you're structuring the world in a way.
60
00:03:14,240 --> 00:03:15,400
Structuring the world, yes.
61
00:03:15,400 --> 00:03:19,720
So if you ask any AI assistant a question, so they break it down and then identify the
62
00:03:19,720 --> 00:03:26,680
entities and then they go over the knowledge graph and then it connects, gathers that information
63
00:03:26,680 --> 00:03:31,240
and then presents that to you in a way you understand like in a human language kind of
64
00:03:31,240 --> 00:03:39,920
speaking kind of way.
65
00:03:39,920 --> 00:03:46,160
So I do want to ask about your interest in knowledge graphs and then of course the explosion
66
00:03:46,160 --> 00:03:48,280
in areas like deep learning.
67
00:03:48,280 --> 00:03:53,560
What do you take of the sort of recent explosion in generative AI and deep learning for someone
68
00:03:53,560 --> 00:03:56,120
who works with knowledge graphs, who wants explainability?
69
00:03:56,120 --> 00:03:57,120
What do you think about that?
70
00:03:57,120 --> 00:03:58,440
Do you think it's a fool's errand?
71
00:03:58,440 --> 00:03:59,440
Do you think we're in trouble?
72
00:03:59,440 --> 00:04:04,680
Do you think we need to pick up our pace or something?
73
Load full transcript
00:04:04,680 --> 00:04:11,080
Like every other thing in the life there is pros and cons and anything should be consumed
74
00:04:11,080 --> 00:04:12,920
in moderation is what I believe in.
75
00:04:12,920 --> 00:04:16,720
Fair enough, even deep learning, right?
76
00:04:16,720 --> 00:04:26,720
So generative AI is great for some tasks, it's like unparalleled, I like it, but I worry
77
00:04:26,720 --> 00:04:32,840
that there is no explaining things, there is no provenance, there is no trustworthiness.
78
00:04:32,840 --> 00:04:37,480
Like, okay, you're telling me this, but what is the source of information or have you fact
79
00:04:37,480 --> 00:04:38,720
checked it?
80
00:04:38,720 --> 00:04:44,720
So my worry is we are consuming it at such a great rate without stopping to think about
81
00:04:44,720 --> 00:04:47,120
how trusty this thing is.
82
00:04:47,120 --> 00:04:51,120
Or should we actually listen to that?
83
00:04:51,120 --> 00:04:56,840
And in a way, it might go on to kill critical thinking.
84
00:04:56,840 --> 00:05:02,200
I don't know anything, I'm just going to ask generative AI and it might give me an answer
85
00:05:02,200 --> 00:05:04,840
and I'm happy with it because it matches my worldview.
86
00:05:04,840 --> 00:05:08,000
I don't want to do any further research into that.
87
00:05:08,000 --> 00:05:09,000
Yeah, that's interesting.
88
00:05:09,000 --> 00:05:13,800
Let me pull that thread a little bit and ask, what are you worried most about in the current
89
00:05:13,800 --> 00:05:14,800
situation?
90
00:05:14,800 --> 00:05:20,720
Is it that critical thinking will be dead or are there other things that you worry about?
91
00:05:20,720 --> 00:05:25,600
So mostly, I think the thing at the top would be critical thinking, it's going to kill critical
92
00:05:25,600 --> 00:05:27,100
thinking.
93
00:05:27,100 --> 00:05:32,560
So I teach a class at USC and then it's like a very basic introduction to Python kind of
94
00:05:32,560 --> 00:05:33,560
class.
95
00:05:33,560 --> 00:05:40,620
And then I have to mention that you may use ChatGPT or any LLM, but do not use it to do
96
00:05:40,620 --> 00:05:44,720
the assignments because if you do that, then why are you taking this course?
97
00:05:44,720 --> 00:05:46,760
Just go talk to ChatGPT.
98
00:05:46,760 --> 00:05:52,680
So in a way, we should encourage use of AI, but with caution.
99
00:05:52,680 --> 00:05:56,440
Use it, but also don't lose your sense of critical thinking.
100
00:05:56,440 --> 00:06:02,160
Yeah, I heard someone recently opining on the need certainly to embrace AI in the classroom
101
00:06:02,160 --> 00:06:08,240
because it's going to be there, but losing sight of the fact that education is about
102
00:06:08,240 --> 00:06:09,600
the struggle, right?
103
00:06:09,600 --> 00:06:12,120
It's the struggle to understand that's actually the process.
104
00:06:12,120 --> 00:06:13,120
It's not the outcome.
105
00:06:13,120 --> 00:06:17,160
We grade and we measure because that's sort of how we have historically measured the process
106
00:06:17,160 --> 00:06:19,920
of education, but really it's about that struggle.
107
00:06:19,920 --> 00:06:23,960
And so this open question of how do we use AI to make the struggle more effective, not
108
00:06:23,960 --> 00:06:27,320
to remove the struggle, but to make it more effective?
109
00:06:27,320 --> 00:06:31,060
Are you optimistic about that, the ability to bring AI into education?
110
00:06:31,060 --> 00:06:32,840
What excites you most there?
111
00:06:32,840 --> 00:06:38,560
I'm weirdly optimistic, but again, as I said, it has to be...
112
00:06:38,560 --> 00:06:42,920
It's like everything should be enjoyed in moderation, so that applies to AI too.
113
00:06:42,920 --> 00:06:47,640
I would love for students to ask questions like, oh, what is this piece of code doing
114
00:06:47,640 --> 00:06:49,680
and why is it doing that?
115
00:06:49,680 --> 00:06:55,320
Just instead of asking for, give me the outcome of this particular assignment or just particular
116
00:06:55,320 --> 00:06:57,320
piece of code or any question.
117
00:06:57,320 --> 00:07:02,960
Just instead of telling me what the answer is, tell me why it's that and how did you
118
00:07:02,960 --> 00:07:08,200
come to that or what other factors did you consider while answering this question or
119
00:07:08,200 --> 00:07:09,200
things like that?
120
00:07:09,200 --> 00:07:14,440
So with introduction of that, I think it would be a great tool.
121
00:07:14,440 --> 00:07:18,600
The fact that we were having this conversation about how to incorporate something this powerful
122
00:07:18,600 --> 00:07:22,360
in the classroom has taken a lot of people by surprise, obviously in the last four or
123
00:07:22,360 --> 00:07:23,880
five years.
124
00:07:23,880 --> 00:07:27,720
What has surprised you most, both about the current state of AI, but also with your own
125
00:07:27,720 --> 00:07:28,720
research?
126
00:07:28,720 --> 00:07:34,640
When you've been doing this for 10 years at least at USC, what have you found in the way
127
00:07:34,640 --> 00:07:37,160
of your results that you thought, oh, that's counterintuitive?
128
00:07:37,160 --> 00:07:41,280
I didn't see that coming or what has surprised you the most?
129
00:07:41,280 --> 00:07:49,200
So I think the thing that surprised me most is how fast it can consume and generate something.
130
00:07:49,200 --> 00:07:52,280
I remember working on a project called DIG.
131
00:07:52,280 --> 00:07:56,640
It was domain specific insight graphs and one of the domains we worked on was human
132
00:07:56,640 --> 00:07:58,000
trafficking.
133
00:07:58,000 --> 00:08:05,640
And then we spent so much time writing some algorithms to extract phone numbers or emails
134
00:08:05,640 --> 00:08:11,320
from the text posted on website, like absolute garbage, HTML.
135
00:08:11,320 --> 00:08:16,400
And if we had generative AI at that point, our life would be so much easier and we would
136
00:08:16,400
--> 00:08:19,240
be so much more productive with that project.
137
00:08:19,240 --> 00:08:24,600
I mean, eventually it was a successful project, but I wonder we could have used that to fight
138
00:08:24,600 --> 00:08:33,640
human trafficking a lot better.
139
00:08:33,640 --> 00:08:38,400
The teams that you've been a part of here at ISI, what makes teams also very good is
140
00:08:38,400 --> 00:08:40,720
everybody brings kind of their own superpower.
141
00:08:40,720 --> 00:08:43,080
I call it a secret superpower to the team.
142
00:08:43,080 --> 00:08:46,520
What do you think your secret superpower is?
143
00:08:46,520 --> 00:08:52,280
So I would say I am very fast at adapting new things.
144
00:08:52,280 --> 00:08:58,320
So if you throw a problem at me, I will give you a solution fairly quickly.
145
00:08:58,320 --> 00:09:02,560
I am able to resist and handle pressure well.
146
00:09:02,560 --> 00:09:09,160
I would say I'm not easy to break down in when something adverse happens.
147
00:09:09,160 --> 00:09:11,200
So I think those are my superpowers.
148
00:09:11,200 --> 00:09:14,080
I could use some of that.
149
00:09:14,080 --> 00:09:15,080
Being unflappable.
150
00:09:15,080 --> 00:09:21,440
Given this sort of, I think frankly, like I said, informed confusion, this sort of infomusion
151
00:09:21,440 --> 00:09:26,400
about the future and where we sit right now, if you could have dinner, let's say with any
152
00:09:26,400 --> 00:09:31,880
person living or dead and spend some time with them to get their advice about how to
153
00:09:31,880 --> 00:09:33,760
think about the future.
154
00:09:33,760 --> 00:09:36,240
Who would that be and why?
155
00:09:36,240 --> 00:09:38,760
So this is not an AI related person.
156
00:09:38,760 --> 00:09:39,760
It could be anybody.
157
00:09:39,760 --> 00:09:40,760
Yeah.
158
00:09:40,760 --> 00:09:43,520
Anybody you think who could be useful for helping us kind of anticipate what's coming
159
00:09:43,520 --> 00:09:44,920
or how to think about it.
160
00:09:44,920 --> 00:09:45,920
Yeah.
161
00:09:45,920 --> 00:09:51,760
While I was growing, so I saw the 90s India, which was on the development stage.
162
00:09:51,760 --> 00:09:55,640
It's still in development, but like very beginning stages.
163
00:09:55,640 --> 00:10:02,160
And then I saw this scientist who shaped up the future of Indian missile system and nuclear
164
00:10:02,160 --> 00:10:07,360
tests and he eventually became the president of the country.
165
00:10:07,360 --> 00:10:09,760
His name is APJ Abdul Kalam.
166
00:10:09,760 --> 00:10:13,040
So I would like to have dinner with him.
167
00:10:13,040 --> 00:10:14,040
He's dead now.
168
00:10:14,040 --> 00:10:15,220
He died in 2015.
169
00:10:15,220 --> 00:10:22,480
But if I would like to hear his views about how he came up with all those ideas or what
170
00:10:22,480 --> 00:10:25,440
was his work ethic.
171
00:10:25,440 --> 00:10:32,480
So he has, I mean, reading about him has introduced me to a term
called syncretism.
172
00:10:32,480 --> 00:10:37,400
And syncretism means a practice of combining different methodologies and different practices
173
00:10:37,400 --> 00:10:38,880
into your day to day life.
174
00:10:38,880 --> 00:10:46,320
So for example, he was a Muslim guy, but he read the Hindu holy books, he read Bhagavad
175
00:10:46,320 --> 00:10:52,120
Gita and then he played music, for example.
176
00:10:52,120 --> 00:10:58,920
And then so he was able to combine all those mindsets and traditions into being successful,
177
00:10:58,920 --> 00:11:00,000
I would say.
178
00:11:00,000 --> 00:11:04,720
So I would like to hear his thought about like, what's a good way of humanity to proceed?
179
00:11:04,720 --> 00:11:07,120
Like, you know, how do we not self-destruct?
180
00:11:07,120 --> 00:11:09,200
Yeah, that's interesting.
181
00:11:09,200 --> 00:11:12,600
I'm proud to say that I think it's an anthropological term.
182
00:11:12,600 --> 00:11:16,040
I think syncretism was something that came out of early anthropology, but certainly that's
183
00:11:16,040 --> 00:11:17,040
where I heard it first.
184
00:11:17,040 --> 00:11:23,520
But yeah, this notion that you can blend things and you can bring on a lot of different lenses.
185
00:11:23,520 --> 00:11:27,440
And the value is you can look at the world through those different lenses and see things
186
00:11:27,440 --> 00:11:31,520
collectively that anybody with just one lens won't see.
187
00:11:31,520 --> 00:11:35,880
Do you think AI needs to be regulated more than it is now?
188
00:11:35,880 --> 00:11:39,680
I realize this is a very controversial question, but just curious in your thought.
189
00:11:39,680 --> 00:11:44,160
Yes, I think that 100% it should be regulated.
190
00:11:44,160 --> 00:11:50,320
So there is internet now is full of AI generated content, like articles written by AI, tweets
191
00:11:50,320 --> 00:11:53,720
by AI, deepfakes by AI.
192
00:11:53,720 --> 00:11:59,720
And then like making very credible videos, for example, of someone saying something which
193
00:11:59,720 --> 00:12:01,640
they have never said.
194
00:12:01,640 --> 00:12:04,880
And it's just unregulated.
195
00:12:04,880 --> 00:12:07,480
The commercial companies say, oh, we don't do that.
196
00:12:07,480 --> 00:12:10,040
We cannot create this image for you.
197
00:12:10,040 --> 00:12:13,760
But you can always take that model training to do whatever you want it to do.
198
00:12:13,760 --> 00:12:15,840
And it's already happening.
199
00:12:15,840 --> 00:12:19,920
So yeah, definitely I think it needs to be regulated.
200
00:12:19,920 --> 00:12:25,920
We need to have a trustworthiness index.
201
00:12:25,920 --> 00:12:29,120
And also I think most important is provenance.
202
00:12:29,120 --> 00:12:30,720
Where are you getting this from?
203
00:12:30,720 --> 00:12:32,800
Or what are your sources?
204
00:12:32,800 --> 00:12:42,520
And why did you show this to me?
205
00:12:42,520 --> 00:12:47,720
What is your favorite sort of AI relevant movie or book?
206
00:12:47,720 --> 00:12:52,040
I don't know if you read science fiction or what do you think is an interesting story
207
00:12:52,040 --> 00:12:54,240
that resonates with you?
208
00:12:54,240 --> 00:12:58,880
So I would go from movies and books to games.
209
00:12:58,880 --> 00:13:00,560
So I play video games.
210
00:13:00,560 --> 00:13:02,920
And then recently this game came out.
211
00:13:02,920 --> 00:13:05,640
It's called Horizon Zero Dawn.
212
00:13:05,640 --> 00:13:09,640
And it's a game about how the humans have self-destructed.
213
00:13:09,640 --> 00:13:11,840
It's like thousand years in future.
214
00:13:11,840 --> 00:13:13,760
And they are now primitive tribes.
215
00:13:13,760 --> 00:13:18,720
But there are machines roaming the world which are very advanced.
216
00:13:18,720 --> 00:13:22,720
So how they came into picture, like who is making these machines and so on.
217
00:13:22,720 --> 00:13:27,300
So somebody made a robot which works on biomass.
218
00:13:27,300 --> 00:13:28,740
So something went out of hand.
219
00:13:28,740 --> 00:13:33,840
And then those robots or those machines start killing every human being because they are
220
00:13:33,840 --> 00:13:35,760
apparently biomass.
221
00:13:35,760 --> 00:13:39,040
So then there was a scientist who wanted to prevent all this.
222
00:13:39,040 --> 00:13:46,120
So they created an AI which could self-regulate, which could create terraforming systems, which
223
00:13:46,120 --> 00:13:48,000
could generate machines.
224
00:13:48,000 --> 00:13:52,160
So for example, they don't have cows in the future, but they have a machine which does
225
00:13:52,160 --> 00:13:53,640
what cows is doing.
226
00:13:53,640 --> 00:13:58,800
So they produce grass and they eat the grass and then all these things.
227
00:13:58,800 --> 00:14:04,200
So self-sustained terraforming system is what that whole game is about.
228
00:14:04,200 --> 00:14:09,920
And then finally, as a player, you find what's the origin, what happened and how we came
229
00:14:09,920 --> 00:14:10,920
up and so on.
230
00:14:10,920 --> 00:14:16,880
So I think that's my favorite recent games about AI.
231
00:14:16,880 --> 00:14:21,440
I think that's maybe the first time someone's mentioned a game on the podcast.
232
00:14:21,440 --> 00:14:24,240
I think it's really informative.
233
00:14:24,240 --> 00:14:28,560
I think that's actually because it is a good story, but it's one you're playing as well.
234
00:14:28,560 --> 00:14:29,560
Okay.
235
00:14:29,560 --> 00:14:30,560
Got your story.
236
00:14:30,560 --> 00:14:33,000
Talk to your 12-year-old self.
237
00:14:33,000 --> 00:14:37,480
Talk now to not your 12-year-old self, but a younger person nonetheless, who's interested
238
00:14:37,480 --> 00:14:39,080
in getting this area.
239
00:14:39,080 --> 00:14:43,760
What advice would you offer to, let's say, a young person, won't define the age, but
240
00:14:43,760 --> 00:14:48,400
a young person
who's interested in this general area, given how fast technology is moving,
241
00:14:48,400 --> 00:14:52,720
given how confusing the future looks, what would you advise them to do?
242
00:14:52,720 --> 00:14:58,200
I would definitely advise them to pursue this, pursue the AI field.
243
00:14:58,200 --> 00:14:59,200
It is the future.
244
00:14:59,200 --> 00:15:00,260
It is interesting.
245
00:15:00,260 --> 00:15:01,840
It is here.
246
00:15:01,840 --> 00:15:10,840
But I would tell them to work on how do you actually make someone believe that this is
247
00:15:10,840 --> 00:15:12,480
for your benefit.
248
00:15:12,480 --> 00:15:18,080
And the goal is, you achieve the goal, but by what means?
249
00:15:18,080 --> 00:15:23,880
So if you're a lawyer, you want to win a case, but you were breaking into houses to gather
250
00:15:23,880 --> 00:15:25,080
some evidence or whatnot.
251
00:15:25,080 --> 00:15:27,420
So let's say you're not very ethical lawyer.
252
00:15:27,420 --> 00:15:34,920
So ethics of it, I would tell that person to pursue this, but also try to build all
253
00:15:34,920 --> 00:15:44,160
these supporting systems, which help the general population have the sense of relief, or they
254
00:15:44,160 --> 00:15:49,840
should not be scared or worried of using something, because they know where it came from or how
255
00:15:49,840 --> 00:15:50,960
it's doing its things.
256
00:15:50,960 --> 00:15:56,360
So yeah, pursue this, but also build systems to keep it ethical.
257
00:15:56,360 --> 00:15:57,360
Right.
258
00:15:57,360 --> 00:16:02,400
So think about it in the context of building this technology, but really always have trustworthiness
259
00:16:02,400 --> 00:16:06,960
in mind and think about the ends don't justify the means at some point.
260
00:16:06,960 --> 00:16:08,720
Yeah, those means matter too.
261
00:16:08,720 --> 00:16:10,680
Yeah, I think that's fair.
262
00:16:10,680 --> 00:16:14,760
You mentioned your concerns to a degree reasonable about things like deep fakes.
263
00:16:14,760 --> 00:16:23,000
I do want to ask if you could have any deep fake voice on this podcast, what voice would
264
00:16:23,000 --> 00:16:24,000
you choose?
265
00:16:24,000 --> 00:16:28,160
Oh, I will give you an interesting answer again.
266
00:16:28,160 --> 00:16:30,600
There's this very famous Bollywood actor.
267
00:16:30,600 --> 00:16:35,560
His name is Amitabh Bachchan, very deep commanding voice.
268
00:16:35,560 --> 00:16:40,200
When they made, I think it's called Bruce Almighty or something.
269
00:16:40,200 --> 00:16:41,600
So he was the voice of the God.
270
00:16:41,600 --> 00:16:46,160
So he's like very powerful, very powerful commanding voice.
271
00:16:46,160 --> 00:16:48,080
So that would be my choice.
272
00:16:48,080 --> 00:16:49,080
I like it.
273
00:16:49,080 --> 00:16:54,960
I'd have to go with James Earl Jones who just died, unfortunately.
274
00:16:54,960 --> 00:16:58,080
Having apparently signed off the rights to his voice.
275
00:16:58,080 --> 00:17:01,520
So before he died, that's going to be interesting.
276
00:17:01,520 --> 00:17:04,760
Yeah, I think that is happening a lot now.
277
00:17:04,760 --> 00:17:08,760
So a lot of actors they have signed off or their families are signing off and then so
278
00:17:08,760 --> 00:17:12,320
that they can continue their voice in games.
279
00:17:12,320 --> 00:17:14,120
So Lance Reddick is another example.
280
00:17:14,120 --> 00:17:15,880
So he died.
281
00:17:15,880 --> 00:17:17,480
He was in one of the John Wick movies.
282
00:17:17,480 --> 00:17:18,480
He's in like many different.
283
00:17:18,480 --> 00:17:22,280
He's also main character in Horizon Zero Dawn.
284
00:17:22,280 --> 00:17:26,640
So now they want to make SQL solid and then they're still able to use his voice.
285
00:17:26,640 --> 00:17:27,640
So that is.
286
00:17:27,640 --> 00:17:28,640
Man.
287
00:17:28,640 --> 00:17:29,640
Yeah.
288
00:17:29,640 --> 00:17:31,680
I mean, it's going to be interesting.
289
00:17:31,680 --> 00:17:37,080
I guess two other questions then.
290
00:17:37,080 --> 00:17:40,520
Given how fast the technology is moving, how do you keep up?
291
00:17:40,520 --> 00:17:44,360
So the term I'm using right is informed confusion.
292
00:17:44,360 --> 00:17:47,860
How do you stay informed in this area?
293
00:17:47,860 --> 00:17:51,320
So I did follow this podcast on your recommendation.
294
00:17:51,320 --> 00:17:53,920
I think it's called Last Week in AI or something.
295
00:17:53,920 --> 00:17:55,480
So I'm following that.
296
00:17:55,480 --> 00:17:56,480
Pretty solid.
297
00:17:56,480 --> 00:17:57,480
Yeah.
298
00:17:57,480 --> 00:18:03,520
I try to keep up with stuff happening, articles online, and then in general, just talking
299
00:18:03,520 --> 00:18:09,800
to other researchers at ISI because I think that is the most productive way of me to learn
300
00:18:09,800 --> 00:18:10,800
about things.
301
00:18:10,800 --> 00:18:16,680
So just talk to like Fred, for example, or Gleb who I work with and just talk like what's
302
00:18:16,680 --> 00:18:17,680
new?
303
00:18:17,680 --> 00:18:18,680
What are you doing?
304
00:18:18,680 --> 00:18:19,680
How are you using it?
305
00:18:19,680 --> 00:18:20,680
Stuff like that.
306
00:18:20,680 --> 00:18:21,680
So.
307
00:18:21,680 --> 00:18:22,680
What a radical concept.
308
00:18:22,680 --> 00:18:24,440
Talk to other human beings and learn from them.
309
00:18:24,440 --> 00:18:28,200
Anthropologists will tell you that, yeah, we are over fixated on our own intelligence as
310
00:18:28,200 --> 00:18:29,800
something that happens between our ears.
311
00:18:29,800 --> 00:18:33,120
And actually human intelligence is what happens between people.
312
00:18:33,120 --> 00:18:37,640
That's sort of the distributed cognition in that regard.
313
00:18:37,640 --> 00:18:38,640
Makes sense to me.
314
00:18:38,640 --> 00:18:43,120
And in this pursuit, Amandeep, I'm really thankful that you're part of the team.
315
00:18:43,120 --> 00:18:45,120
So thanks again for coming on board.
316
00:18:45,120 --> 00:18:48,800
Thanks for being so thoughtful in your responses.
317
00:18:48,800 --> 00:18:51,360
And I think again, a lot of people think that sort of AI is this magic.
318
00:18:51,360 --> 00:18:52,360
It is not.
319
00:18:52,360 --> 00:18:54,120
It is people making this happen.
320
00:18:54,120 --> 00:18:58,920
You are a perfect example of how this happens and by people who at the age of 12 never would
321
00:18:58,920 --> 00:19:00,740
have considered this possible.
322
00:19:00,740 --> 00:19:03,480
So yeah, some of the small hero for me.
323
00:19:03,480 --> 00:19:04,480
Thank you, Amandeep.
324
00:19:04,480 --> 00:19:06,920
Thanks for having me.
325
00:19:06,920 --> 00:19:07,920
Thank you.
326
00:19:07,920 --> 00:19:08,920
All right.
327
00:19:08,920 --> 00:19:12,840
Well, if you enjoy these short podcasts, even if you don't, please let us know.
328
00:19:12,840 --> 00:19:13,840
We love feedback.
329
00:19:13,840 --> 00:19:19,000
So you can rate us, grade us, subscribe to us, hit me or ISI up on LinkedIn, send over
330
00:19:19,000 --> 00:19:20,640
a carrier pigeon, whatever.
331
00:19:20,640 --> 00:19:25,480
But of course, most of all, we hope you'll keep listening and learning and that you'll
332
00:19:25,480 --> 00:19:29,880
join us again for another episode of AI Insiders where we'll keep exploring and keep learning
333
00:19:29,880 --> 00:19:34,000
and keep humanizing AI and hopefully help navigate our way through this weird, weird
334
00:19:34,000 --> 00:19:38,960
world trying to do what humans can do best when they face these kinds of challenges.
335
00:19:38,960 --> 00:19:44,440
Working together as if all our lives depend on each other because they do.
336
00:19:44,440 --> 00:20:13,000
And for now, for the future, bite on.