Human Choice and the Ethics of AI

WEBVTT
1
00:00:00.720 --> 00:00:04.719
The Tech Policy Design Institute acknowledges and pays our respects
2
00:00:04.799 --> 00:00:09.759
to all First Nations people. We recognize and celebrate that,
3
00:00:09.880 --> 00:00:15.119
among many things, Indigenous people were Australia's first tech innovators.
4
00:00:18.839 --> 00:00:22.000
But I love that you've managed to link in their mortality,
5
00:00:22.199 --> 00:00:25.320
chips and craft to be out all in one respect.
6
00:00:25.440 --> 00:00:27.280
What would It's amazing.
7
00:00:29.760 --> 00:00:34.479
Too often we develop a technology without a clear sense
8
00:00:34.520 --> 00:00:38.079
of its purpose. But there's an old syllogism in ethics
9
00:00:38.159 --> 00:00:41.000
which asks, what is the purpose of a knife? Yeah,
10
00:00:41.000 --> 00:00:42.640
and so the purpose of a knife is to cut?
11
00:00:42.759 --> 00:00:44.560
What is a good knife? A good knife is a
12
00:00:44.600 --> 00:00:46.960
knife that cuts well, and the blade must be appropriately
13
00:00:47.000 --> 00:00:49.560
sharpened and shaped for the task it has to perform.
14
00:00:49.759 --> 00:00:52.679
It starts with that purpose when we come to the
15
00:00:52.679 --> 00:00:55.399
application of this technology. And even at the earlier stated
16
00:00:55.799 --> 00:01:00.159
let's not think about deploying things of power and utility
17
00:01:00.479 --> 00:01:03.920
in conditions of blindness where we just don't see what
18
00:01:03.960 --> 00:01:07.480
we're doing, like walking in a fog over a kind
19
00:01:07.480 --> 00:01:09.519
of the edge of a cliff into an ethic death trap.
20
00:01:23.959 --> 00:01:26.920
Hello, and welcome to tech Mirror. My name is Johannah
21
00:01:26.920 --> 00:01:29.519
Weaver and I'm your host, and this is the podcast
22
00:01:29.560 --> 00:01:32.959
where we talk about how technology is shaping our world
23
00:01:33.359 --> 00:01:37.239
and how we the humans can shape technology back. And
24
00:01:37.400 --> 00:01:41.239
today we have a really special guest, Simon Longstaff. Simon
25
00:01:41.439 --> 00:01:44.519
is the executive director of the Ethics Center and is
26
00:01:44.560 --> 00:01:48.920
widely recognized to be one of Australia's leading ethicists. Simon
27
00:01:48.920 --> 00:01:51.840
and I serve on an advisory board together, and he's
28
00:01:51.879 --> 00:01:54.519
one of these people that I've never had a boring
29
00:01:54.599 --> 00:01:57.239
conversation with. So Simon, welcome to the podcast.
30
00:01:57.439 --> 00:02:00.239
Thanks, No pressure, Urger, Hannah.
31
00:01:59.519 --> 00:02:03.079
No press it all. And actually, I mean the origin
32
00:02:03.159 --> 00:02:06.560
of this conversation Simon started. We're on an advisory board
33
00:02:06.640 --> 00:02:09.759
together and we were having a conversation in one of
34
00:02:09.800 --> 00:02:14.080
the breaks about how much angst and anxiety there is
35
00:02:14.159 --> 00:02:17.360
in the general population around artificial intelligence at the moment.
36
00:02:17.599 --> 00:02:21.199
And we were talking about this concept of social license
37
00:02:21.280 --> 00:02:23.919
and the fact that it seems to be, you know,
38
00:02:24.000 --> 00:02:27.000
the term of the hour at the moment that everyone's
39
00:02:27.039 --> 00:02:30.560
talking about the need for companies to have social license
40
00:02:30.639 --> 00:02:34.919
to make people feel comfortable in using artificial intelligence. And
41
00:02:34.960 --> 00:02:37.719
you challenged me and you said that you thought the
42
00:02:37.800 --> 00:02:42.199
social compact was a better phrase and that we should
43
00:02:42.199 --> 00:02:45.879
be aiming for more than social license. So let's start there,
44
00:02:46.080 --> 00:02:48.759
and can you explain what the difference between those two
45
00:02:48.840 --> 00:02:50.879
is and why you think one is better than the other.
46
00:02:51.560 --> 00:02:56.360
Yes. So the evolution of these concepts does begin, as
47
00:02:56.400 --> 00:03:00.599
you suggest, with the notion of a social license. Intuitive
48
00:03:01.039 --> 00:03:05.199
sense that if you are going to be coming as
49
00:03:05.240 --> 00:03:08.560
a legal person, which is the modern form of the corporation,
50
00:03:09.240 --> 00:03:12.639
into a community, you need a license in order to
51
00:03:12.719 --> 00:03:16.639
be able to pursue your interests. And the concept of
52
00:03:16.680 --> 00:03:20.039
a license is that, in a sense, you pay a
53
00:03:20.159 --> 00:03:23.000
fee in order to receive a license, or you pass
54
00:03:23.080 --> 00:03:26.319
some kind of test, and then you're given the liberty
55
00:03:26.360 --> 00:03:30.400
to go about pursuing your own ends. That's a very
56
00:03:30.560 --> 00:03:34.520
transactional notion, and it's a bit like you come into
57
00:03:34.639 --> 00:03:36.960
somebody's house and you pay rent. You say, well, now
58
00:03:36.960 --> 00:03:39.319
I've got my room, and if I want to make
59
00:03:39.319 --> 00:03:42.599
it into a stink hole of some kind, that's my room.
60
00:03:42.639 --> 00:03:45.199
I paid rent. I can do what I like in it.
61
00:03:45.479 --> 00:03:47.800
And as long as you continue to another transaction, you
62
00:03:47.919 --> 00:03:50.719
sort of go on your own separate way. There was
63
00:03:50.759 --> 00:03:53.240
an evolution in that thinking that then came about by
64
00:03:53.280 --> 00:03:57.080
two academics in Harvard, Porter and Kramer, around the notion
65
00:03:57.199 --> 00:04:01.599
of shared value. That is, rather than just seeing yourself
66
00:04:01.639 --> 00:04:03.840
as being an occupant of a space for which you
67
00:04:03.919 --> 00:04:08.639
plaid a license, you exchange some value back to your host.
68
00:04:09.479 --> 00:04:12.000
And now I may be doing them at disservice in
69
00:04:12.000 --> 00:04:14.560
the way I'll present it, but it has the slightly
70
00:04:15.439 --> 00:04:18.480
parasitic notion to it, that you have a host and
71
00:04:18.519 --> 00:04:20.800
you a parasite, and it's got some kind of symbiotic
72
00:04:21.519 --> 00:04:25.319
value of exchange there. And then there's this concept of
73
00:04:25.959 --> 00:04:30.279
social value which has emerged only very recently, only in
74
00:04:30.360 --> 00:04:34.560
limited context, but particularly has been championed and now given
75
00:04:34.839 --> 00:04:39.720
very specific practical application by BHP, which is the idea
76
00:04:39.759 --> 00:04:42.759
that rather than you being a parasite that lives within
77
00:04:42.839 --> 00:04:46.639
the host called society, you are in fact an organic
78
00:04:46.720 --> 00:04:50.000
part of it. In other words, it breaks down this
79
00:04:50.120 --> 00:04:54.839
distinction that sometimes is made between the state, civil society
80
00:04:54.959 --> 00:04:59.360
and business. Said, well, actually, isn't business part of civil society?
81
00:04:59.439 --> 00:05:02.879
And so the analogy, the organic analogy that I would
82
00:05:02.959 --> 00:05:06.480
use is to say, well, imagine a human body. It
83
00:05:06.519 --> 00:05:09.399
has a number of organs of which a corporation might
84
00:05:09.439 --> 00:05:13.120
be one. We'll say it's a heart for example. Now,
85
00:05:13.399 --> 00:05:17.199
its health is entirely dependent upon the health of the
86
00:05:17.240 --> 00:05:20.439
wider body of which it is a part, but likewise
87
00:05:20.480 --> 00:05:23.000
so is the health of that body. So instead of
88
00:05:23.040 --> 00:05:27.319
this being one which is based on transactions or just
89
00:05:27.360 --> 00:05:30.199
a kind of parasitic exchange of value, this is an
90
00:05:30.360 --> 00:05:34.399
organic whole which has different elements that perform different roles.
91
00:05:34.879 --> 00:05:37.079
But when you look at it in its application as
92
00:05:37.160 --> 00:05:41.360
some very profound implications. For example, the heart no longer
93
00:05:41.399 --> 00:05:45.839
claims any sense superior position visibly any other organ because
94
00:05:46.240 --> 00:05:49.120
all of them are important. And so you will see
95
00:05:49.279 --> 00:05:53.480
even a very powerful corporation like BHP voluntarily choose to
96
00:05:53.519 --> 00:05:57.000
subordinate its power and to sit down on the same
97
00:05:57.120 --> 00:06:00.160
side of the table as other parts of the community
98
00:06:00.439 --> 00:06:03.120
of which it is just one part of the larger hole.
99
00:06:03.839 --> 00:06:06.800
And that's quite a different model, both in terms of
100
00:06:07.360 --> 00:06:10.959
its qualitative aspects, but also in terms of what it
101
00:06:11.000 --> 00:06:13.720
then practically implies about what you do. And you can
102
00:06:13.759 --> 00:06:16.199
take that richer notion, as I would say it, and
103
00:06:16.199 --> 00:06:18.319
then apply it to a whole range of things, including
104
00:06:18.399 --> 00:06:19.480
the discussion about AI.
105
00:06:20.160 --> 00:06:23.079
I love that notion of looking at it as the
106
00:06:23.160 --> 00:06:26.759
human body and all parts needing to be functioning. Otherwise
107
00:06:27.040 --> 00:06:29.639
what's the point of being alive? Or you can't be alive.
108
00:06:30.519 --> 00:06:33.800
If we were to take that from the abstract to
109
00:06:34.680 --> 00:06:37.360
looking at what that would mean in terms of building
110
00:06:37.360 --> 00:06:41.639
a social compact for the Australian population. What would that
111
00:06:41.800 --> 00:06:46.480
look like in terms of the conversations that government, the
112
00:06:46.519 --> 00:06:51.079
tech companies, civil society and Australian citizens. What types of
113
00:06:51.079 --> 00:06:54.680
conversations would we need to be having and how would
114
00:06:54.680 --> 00:06:54.879
they be.
115
00:06:55.759 --> 00:06:57.839
What it would happen. First of all, you would start
116
00:06:57.920 --> 00:07:03.720
to reposition the whole conversation where everybody, whether you are
117
00:07:03.720 --> 00:07:06.560
a government which is of course a big gorilla in
118
00:07:06.600 --> 00:07:10.480
the room in any state like Australia, or the largest
119
00:07:10.519 --> 00:07:14.560
corporation or the smallest, or just a citizen, and you'd say, actually,
120
00:07:14.759 --> 00:07:19.399
rather than breaking us into fractions that therefore compete for
121
00:07:19.519 --> 00:07:24.319
power or opportunity, that instead we see our competition, which
122
00:07:24.360 --> 00:07:27.399
will always be there, as being something which is directed
123
00:07:27.439 --> 00:07:31.639
towards a larger common good. In fact, this harks back
124
00:07:31.680 --> 00:07:34.439
to the earliest arguments that were put forward by people
125
00:07:34.480 --> 00:07:37.920
like Adam Smith for the very notion of a free market,
126
00:07:38.000 --> 00:07:41.199
which in his view had no intrinsic value. It was
127
00:07:41.240 --> 00:07:44.399
simply a mechanism by which, if it was instantiated, it
128
00:07:44.399 --> 00:07:46.720
would lead to an increase in the stock of common good.
129
00:07:46.800 --> 00:07:49.639
That is, everybody would be better off because we're all
130
00:07:49.680 --> 00:07:53.920
in the same proverbial boat together. So you would start
131
00:07:53.920 --> 00:07:56.439
off with that as an idea, and then practically you
132
00:07:56.480 --> 00:08:00.879
would start to see even the most powerful of the
133
00:08:01.240 --> 00:08:06.480
tech companies actually choosing to say, we won't sit across
134
00:08:06.519 --> 00:08:11.319
the table as an antagonist against the wishes of government
135
00:08:11.399 --> 00:08:14.000
or a people, where we wrestle them to a position
136
00:08:14.040 --> 00:08:17.600
which we can live with. Rather, you would say, actually,
137
00:08:17.680 --> 00:08:21.439
we're part of this society, even if we are based overseas.
138
00:08:21.480 --> 00:08:23.560
We know that in the way we present ourselves, say
139
00:08:23.600 --> 00:08:27.319
in a country, we are part of the Australian story,
140
00:08:27.120 --> 00:08:30.759
the Australia political and social and economic conditions. How do
141
00:08:30.920 --> 00:08:34.559
we all work together in a way that ensures that
142
00:08:34.919 --> 00:08:39.759
we are as a whole actually not merely being sustained
143
00:08:39.759 --> 00:08:43.440
but flourishing. Which is not to say that an economic
144
00:08:43.519 --> 00:08:46.960
actor wouldn't be pursuing profit, or that there wouldn't be
145
00:08:47.000 --> 00:08:49.480
a regard by governments or the social welfare. It's community
146
00:08:49.480 --> 00:08:51.559
and a whole lot of other things. But you don't
147
00:08:51.639 --> 00:08:55.120
start down on the basis of an arm wrestle to
148
00:08:55.240 --> 00:08:58.039
try and see what you can extract from a relationship
149
00:08:58.639 --> 00:09:01.759
rather that you have that practical conversation. Now, it would
150
00:09:01.799 --> 00:09:04.399
be a very different set of conversations we would be
151
00:09:04.399 --> 00:09:06.840
having if that was the starting point. Then what is
152
00:09:06.879 --> 00:09:07.919
currently the situation?
153
00:09:08.519 --> 00:09:11.039
And how do we change the starting point? Because I
154
00:09:11.039 --> 00:09:14.919
think many people listening to that will say that sounds excellent,
155
00:09:15.039 --> 00:09:17.919
Yes we should do that, but how do we get there?
156
00:09:17.960 --> 00:09:20.720
That doesn't reflect the power dynamics of the modern world.
157
00:09:21.240 --> 00:09:22.559
So how do you respond to that.
158
00:09:23.000 --> 00:09:26.480
It's going to sound like a rather abstract answer, because
159
00:09:26.519 --> 00:09:28.600
I know people will say that's all very good, it's
160
00:09:28.679 --> 00:09:32.000
kind of high in the sky. But I think we
161
00:09:32.080 --> 00:09:35.559
underestimate the power of ideas in these things, and the
162
00:09:35.600 --> 00:09:37.240
power of language, and the fact that you can have
163
00:09:37.320 --> 00:09:41.000
a concept if you've grown up in a world which
164
00:09:41.039 --> 00:09:45.919
assumes that various actors are in a sense competing for
165
00:09:46.159 --> 00:09:49.759
their own interest, and that somehow, and this is the
166
00:09:49.879 --> 00:09:52.840
kind of the logical part of the market, that it
167
00:09:53.080 --> 00:09:56.879
presumes that, for example, creatures within the marketplace are not
168
00:09:56.960 --> 00:09:59.480
at all concerned about the general well being, that they're
169
00:09:59.519 --> 00:10:02.360
just pursuing their self interest, but that ultimately that will
170
00:10:02.399 --> 00:10:04.799
lead to that increasing stock in that stock of common
171
00:10:04.840 --> 00:10:06.960
good he said, actually, no, we're going to have a
172
00:10:07.000 --> 00:10:09.320
bit more of a nuanced view of the kind that
173
00:10:09.360 --> 00:10:12.559
actually Adam Smith does put forward in his wider writings.
174
00:10:13.440 --> 00:10:16.480
And you think that's what life could be like, not
175
00:10:17.039 --> 00:10:20.000
just could be, but it should be like that, because
176
00:10:20.000 --> 00:10:22.360
that's going to lead to a society in which we
177
00:10:22.440 --> 00:10:26.840
all flourish. Then the fact that you can imagine this, now,
178
00:10:27.480 --> 00:10:29.879
that you can actually invoke that language, that you can
179
00:10:29.919 --> 00:10:34.000
actually start to insist on institutional arrangements in which that
180
00:10:34.120 --> 00:10:37.799
become the starting point rather than treated as an exception
181
00:10:37.879 --> 00:10:42.039
that could never be realized, then I think that's the beginning. Now.
182
00:10:42.559 --> 00:10:46.879
I know it's not as firm as creating an institutional
183
00:10:46.960 --> 00:10:51.240
framework or some kind of tangible structure, but I think
184
00:10:51.279 --> 00:10:55.519
those things follow from ideas about what is possible, and
185
00:10:55.559 --> 00:10:57.840
once you get enough people saying, oh, yes, I'd like
186
00:10:57.919 --> 00:10:59.840
to be part of that, how do I do it?
187
00:11:00.320 --> 00:11:03.240
Then you start to co create the environment which that
188
00:11:03.279 --> 00:11:04.120
becomes true.
189
00:11:04.559 --> 00:11:07.799
I couldn't agree more that it is the power or
190
00:11:08.080 --> 00:11:11.519
in the power of ideas, and what we're seeing at
191
00:11:11.519 --> 00:11:15.759
the moment is very clearly an awakening of the Australian population.
192
00:11:15.840 --> 00:11:19.639
But I also think globally of concern about what AI
193
00:11:19.799 --> 00:11:21.399
is going to do, the impact that it's going to
194
00:11:21.440 --> 00:11:23.480
have on their lives, on the environment, on the people
195
00:11:23.480 --> 00:11:26.519
they love. And so I think there's real power to
196
00:11:26.600 --> 00:11:29.559
what you're saying right in this very moment, that this
197
00:11:29.639 --> 00:11:33.600
is a moment where we can shape the future. But
198
00:11:33.679 --> 00:11:35.799
we actually do have to recognize that we have the
199
00:11:35.840 --> 00:11:38.159
power to shape the future, because if we don't step
200
00:11:38.240 --> 00:11:40.360
up and recognize that, then it's just going to continue
201
00:11:40.360 --> 00:11:41.120
on the trajectory.
202
00:11:41.840 --> 00:11:45.320
It's the greatest source of oppression of human beings is
203
00:11:45.360 --> 00:11:48.360
not that we are powerless, it's that we are led
204
00:11:48.399 --> 00:11:53.480
to believe that we are powerless. Now, it's an incredible
205
00:11:53.480 --> 00:11:56.919
thing to realize that the most powerful force on this planet,
206
00:11:57.480 --> 00:12:00.000
limited only by the laws of physics, is human choice.
207
00:12:00.919 --> 00:12:03.279
We often think, oh, that that's only the choice of
208
00:12:03.320 --> 00:12:08.840
the people who have formal power, institutional power, when in fact,
209
00:12:08.840 --> 00:12:13.320
what we know is that when people make very tiny
210
00:12:13.399 --> 00:12:17.159
decisions in large numbers, that can tilt the world on
211
00:12:17.200 --> 00:12:23.960
its axis. So one look, the world celebrates heroic individuals
212
00:12:24.000 --> 00:12:28.919
who shape the narratives of history because of their singular actions,
213
00:12:29.759 --> 00:12:32.960
but that isn't actually what drives most of the change
214
00:12:33.399 --> 00:12:37.039
in the world. It's people just falling a little bit,
215
00:12:37.240 --> 00:12:39.600
just a tiny bit on the right side of any question.
216
00:12:40.480 --> 00:12:43.840
But if you are told that you count for nothing,
217
00:12:43.879 --> 00:12:47.360
that your choices don't matter, that you are powerless, and
218
00:12:47.480 --> 00:12:50.000
you come to believe that because when you look at
219
00:12:50.240 --> 00:12:53.399
the size of the challenge they seem so overwhelming in
220
00:12:53.480 --> 00:12:56.919
their size and complexity, then you can lay down that
221
00:12:57.000 --> 00:13:00.120
opportunity to be the person who makes the choice that
222
00:13:00.200 --> 00:13:03.639
might have been like the straw on the proverbial camel's back,
223
00:13:03.720 --> 00:13:06.759
it's that last straw break that you just made, that
224
00:13:06.799 --> 00:13:10.320
final decision that tilted the world. And so I think
225
00:13:10.519 --> 00:13:15.399
it's incredibly important that we understand that with AI, that
226
00:13:15.440 --> 00:13:20.399
there's nothing inevitable about it. It is a product of choice.
227
00:13:20.480 --> 00:13:22.919
And yes, you might think that it's the great tech
228
00:13:22.960 --> 00:13:27.480
bros whose choices ultimately matter. No, they are susceptible as
229
00:13:27.559 --> 00:13:32.360
anyone has been through emperors, kings, presidents and anyone else,
230
00:13:32.840 --> 00:13:36.000
to the large scale opinion of humanity as a whole.
231
00:13:36.159 --> 00:13:39.240
And so when we're talking about the tech bros, one
232
00:13:39.240 --> 00:13:42.080
of the things that they're really really good at talking
233
00:13:42.080 --> 00:13:45.240
about at the moment is the potential of AI. So
234
00:13:45.519 --> 00:13:48.480
can we do this with AI and the potential for
235
00:13:48.559 --> 00:13:51.679
it to have this incredibly transformative effect, which we are
236
00:13:51.720 --> 00:13:53.960
already seeing. So it's not something that is just in
237
00:13:54.000 --> 00:13:59.360
the future. You often talk about should we and injecting
238
00:13:59.480 --> 00:14:02.559
that frame into the conversation. So for people listening to
239
00:14:02.559 --> 00:14:06.039
what you're saying, Okay, you've convinced me, Simon, I've got
240
00:14:06.080 --> 00:14:09.960
some power. What's some framing that you can give them
241
00:14:10.000 --> 00:14:13.879
in terms of ways to think about whether AI is
242
00:14:14.399 --> 00:14:17.159
going to bring a benefit. So not just the can it,
243
00:14:17.279 --> 00:14:17.919
but should we?
244
00:14:19.080 --> 00:14:22.480
So you've already answered one of the most important things
245
00:14:22.480 --> 00:14:25.080
as a principle, and that is that can does not
246
00:14:25.159 --> 00:14:28.679
imply ought. In the fact that you can do something
247
00:14:28.799 --> 00:14:31.279
doesn't mean you should do it. And we know intuitively
248
00:14:31.360 --> 00:14:33.799
this is the case. There's lots and lots of things
249
00:14:33.840 --> 00:14:36.279
we could do that we choose not to do. And
250
00:14:36.360 --> 00:14:38.720
the fact that we spend in most of our lives
251
00:14:39.000 --> 00:14:41.320
we wander around the streets of wherever we're happen to live.
252
00:14:41.360 --> 00:14:43.919
We're not worried about being mugged for most of the
253
00:14:44.000 --> 00:14:46.519
time or threatened by someone. It's not because there's a
254
00:14:46.559 --> 00:14:50.679
policeman on every corner or surveillance. It's because we just
255
00:14:50.879 --> 00:14:53.159
carry within our selves the fact that, yes, we could
256
00:14:53.200 --> 00:14:54.879
do all sorts of terrible things to each other, but
257
00:14:54.919 --> 00:14:57.080
we choose not to do so now for fear of
258
00:14:57.159 --> 00:14:59.840
being caught, but because that would be, as we understand it,
259
00:15:00.000 --> 00:15:03.879
simply the wrong thing to do. And so that sense
260
00:15:03.919 --> 00:15:07.840
of self regulation can be brought to dare in relation
261
00:15:08.000 --> 00:15:13.360
to the application of this extraordinary technology. Now, the second
262
00:15:13.360 --> 00:15:15.600
thing about that we can do as a way of frame,
263
00:15:15.679 --> 00:15:17.960
and this is to ask when we come to the
264
00:15:18.000 --> 00:15:20.600
application of this technology, and even at the earlier state
265
00:15:20.960 --> 00:15:23.720
when it's first being developed, is what is the purpose
266
00:15:24.159 --> 00:15:27.799
that it is to serve? Because it's only if you
267
00:15:27.879 --> 00:15:31.000
understand its purpose that you can understand whether or not
268
00:15:31.080 --> 00:15:34.399
it is a good expression of the realization of that purpose.
269
00:15:34.440 --> 00:15:38.159
So there's an old syllogism in ethics which asks, what
270
00:15:38.279 --> 00:15:40.639
is the purpose of a knife? Yeah, and so the
271
00:15:40.679 --> 00:15:42.440
purpose of a knife is to cut. What is a
272
00:15:42.440 --> 00:15:44.559
good knife? A good knife is a knife that cuts
273
00:15:44.559 --> 00:15:47.200
well and entailed in that is the idea that the
274
00:15:47.200 --> 00:15:50.120
blade must be appropriately sharpened and shaped for the task
275
00:15:50.159 --> 00:15:52.440
it has to perform. But that's the point was having
276
00:15:52.480 --> 00:15:54.960
a blade which is perfect in that respect if it
277
00:15:54.960 --> 00:15:57.039
can't be used because it's attached to a handle the
278
00:15:57.080 --> 00:15:58.799
size of an elephant. You know, I could pick it up,
279
00:15:58.840 --> 00:16:01.360
no one could use it. There's something about fitting in
280
00:16:01.360 --> 00:16:04.080
the hand to user, but it starts with that purpose.
281
00:16:04.919 --> 00:16:09.639
Too often we develop a technology without a clear sense
282
00:16:09.679 --> 00:16:13.240
of its purpose. So, for example, in AI, there's a
283
00:16:13.399 --> 00:16:19.200
massive difference between developing an AI stack, which is simply
284
00:16:19.279 --> 00:16:24.320
about bringing about greater efficiency, where you seek to reduce
285
00:16:24.360 --> 00:16:26.840
the unit cost of producing whatever it happens to be.
286
00:16:27.320 --> 00:16:29.840
If that is your purpose, and that is your only purpose,
287
00:16:29.879 --> 00:16:33.039
then you invest that idea into the thing you're making.
288
00:16:33.679 --> 00:16:36.200
That's radically different than if you have a purpose, which
289
00:16:36.240 --> 00:16:40.200
is better to realize the mission for which your organization
290
00:16:40.399 --> 00:16:44.159
has been established. So that's the second element to it.
291
00:16:44.200 --> 00:16:47.120
To understand purpose the third part of a framework which
292
00:16:47.159 --> 00:16:50.240
any of us can INPT. So I'm thinking here now
293
00:16:50.559 --> 00:16:52.360
an ordinary person saying, oh, what do I think about
294
00:16:52.399 --> 00:16:55.200
this bit of AI? I'd be asking, was not just
295
00:16:55.240 --> 00:16:57.200
something that's been done because it can be done? Can
296
00:16:57.240 --> 00:16:59.960
they explain these things? What do they say? Is it purpose?
297
00:17:00.080 --> 00:17:02.679
Can I see that being realized? The next stage would
298
00:17:02.679 --> 00:17:05.839
be to what extent have they been explicit about the
299
00:17:05.880 --> 00:17:10.319
core values and principles which are informing how this has
300
00:17:10.319 --> 00:17:13.319
been developed and how it's supposed to operate in the world. Again,
301
00:17:13.359 --> 00:17:16.480
this isn't an abstract question. Good software engineers should have
302
00:17:16.480 --> 00:17:19.920
things called markdown files. When they say this is the
303
00:17:19.960 --> 00:17:23.079
framework within which the AI has been developed, you should
304
00:17:23.079 --> 00:17:24.599
be able to see that, and you should be able
305
00:17:24.599 --> 00:17:27.519
to know that the system is transparent enough such that
306
00:17:27.640 --> 00:17:30.799
an expert and disinterested person could actually look at and say, yeah,
307
00:17:30.839 --> 00:17:33.640
I can see those values, I can see those principles
308
00:17:33.680 --> 00:17:36.119
being given effect through the way of which they've at
309
00:17:36.200 --> 00:17:38.839
least tried to make this work, and then we'll come
310
00:17:38.839 --> 00:17:41.640
back perhaps some of the limitations in this. The next
311
00:17:41.680 --> 00:17:44.480
thing you would ask is have they been clear about
312
00:17:44.519 --> 00:17:48.680
the affordances which they have created through the establishment of
313
00:17:48.720 --> 00:17:52.599
this technology this bit of AI. An affordance is a
314
00:17:52.680 --> 00:17:57.119
capacity to do something which is either deliberately brought into existence,
315
00:17:57.200 --> 00:17:59.640
for example, in pursuit of a purpose, or you can
316
00:17:59.680 --> 00:18:04.599
create unintended affordances which create new opportunities which might be
317
00:18:04.680 --> 00:18:09.200
dangerous and damaging. Or you might create affordances or you
318
00:18:09.240 --> 00:18:12.440
may fail to rule them out. So classic example is,
319
00:18:12.480 --> 00:18:16.759
say a handgun, where they're prevalent in a country like
320
00:18:16.799 --> 00:18:19.000
the United states, a person can walk in and pick
321
00:18:19.039 --> 00:18:22.119
up a handgun and they can just pointed it and
322
00:18:22.160 --> 00:18:24.440
if it's loaded, they could shoot anyone in your respective
323
00:18:24.480 --> 00:18:28.160
of who they are. Now that's an affordance which exists
324
00:18:28.160 --> 00:18:30.599
within the handgun because there's no control over who gets
325
00:18:30.599 --> 00:18:33.519
to use it, whether they're licensed or otherwise. Whereas if
326
00:18:33.519 --> 00:18:37.079
you built into the GRIP a finger print recognition system
327
00:18:37.240 --> 00:18:40.400
so that it could only be activated and used by
328
00:18:40.480 --> 00:18:43.960
the person who is the license holder of that immediately
329
00:18:44.000 --> 00:18:46.839
you shrink the possibility of its being used for purposes
330
00:18:46.839 --> 00:18:50.880
other than rich it was created and licensed. So these
331
00:18:50.920 --> 00:18:54.440
are practical things that you can do to both think
332
00:18:54.440 --> 00:18:56.960
about how you create those technologies and as a person
333
00:18:57.000 --> 00:18:59.359
who is curious about them, to decide you approve or
334
00:18:59.359 --> 00:19:01.839
not approve what the framework within which you should assess them.
335
00:19:02.119 --> 00:19:05.279
One of the other things that I really love is
336
00:19:05.319 --> 00:19:09.799
your explanation of the difference between ethics, law, and morals,
337
00:19:10.000 --> 00:19:13.480
which are often used to not like the synonyms, but
338
00:19:13.640 --> 00:19:16.920
which you really clearly articulate the difference between them, and
339
00:19:16.920 --> 00:19:19.599
I think that's important when we're talking about, particularly around
340
00:19:19.640 --> 00:19:23.039
governing AI and the frameworks in which we're going to
341
00:19:23.079 --> 00:19:26.359
allow or maybe not allow this technology to be shaping
342
00:19:26.359 --> 00:19:28.480
our society. So can you talk that through for us?
343
00:19:29.000 --> 00:19:32.359
Yeah, So let's just start with law. Law, of course,
344
00:19:32.480 --> 00:19:36.359
is a series of constraints which are typically agreed upon,
345
00:19:36.599 --> 00:19:40.799
certainly in the democratic society, sometimes imposed in more authoritarian societies,
346
00:19:41.880 --> 00:19:44.960
that in their more usual form tell you what you
347
00:19:45.119 --> 00:19:49.880
cannot do. So, and that certainly liberal democracies. The law
348
00:19:50.400 --> 00:19:53.160
tries to allow the largest amount of freedom for a
349
00:19:53.200 --> 00:19:55.720
person to exercise their own judgment, live their own life,
350
00:19:55.720 --> 00:19:58.240
and so it will say, but you cannot do this,
351
00:19:58.559 --> 00:20:01.200
and then it will constrain. There are exceptions to that,
352
00:20:01.279 --> 00:20:05.880
of course, but it's a rule which is typically imposed
353
00:20:05.960 --> 00:20:09.000
by some legitimate authority or something claiming authority, whether it's
354
00:20:09.079 --> 00:20:12.920
legitimate or otherwise, and which is imposed to the extent
355
00:20:12.960 --> 00:20:16.000
that if you violate the rule then it has certain
356
00:20:16.039 --> 00:20:19.279
penalties often attached to it finds, imprisonment and other forms
357
00:20:19.279 --> 00:20:24.279
of punishment. But the thing about law is that law
358
00:20:24.720 --> 00:20:27.960
is indifferent to whether it is just or unjust. So
359
00:20:28.680 --> 00:20:32.319
in Australia, just as existed in South Africa. In Victoria
360
00:20:32.720 --> 00:20:35.559
in the nineteenth century, there are anti misagenation laws that
361
00:20:35.680 --> 00:20:40.359
said it was wrong for a person who was a
362
00:20:40.400 --> 00:20:44.039
white person a man to be in a relationship with
363
00:20:44.119 --> 00:20:47.119
an Aboriginal woman. So it was basically an apartheid type
364
00:20:47.119 --> 00:20:51.400
of law which existed in Australia. And today we would
365
00:20:51.440 --> 00:20:54.400
look at that and we would say that was wrong,
366
00:20:54.519 --> 00:20:56.400
that was an unjust law. And there have been other
367
00:20:56.759 --> 00:20:59.680
laws at that kind around things like slavery. And it's
368
00:20:59.680 --> 00:21:02.759
implet here to understand that there were people at the
369
00:21:02.799 --> 00:21:06.440
time who knew that that was wrong, and nonetheless these
370
00:21:06.480 --> 00:21:09.680
things persistent because they had that status of war. So
371
00:21:09.720 --> 00:21:12.000
it's a contingent fact about the world if you like
372
00:21:12.119 --> 00:21:14.880
that such laws exist, and they can be amended, they
373
00:21:14.920 --> 00:21:19.480
can be repealed, and so forth. Now below that, the
374
00:21:19.559 --> 00:21:21.839
reason that we're able to say that something is an
375
00:21:21.880 --> 00:21:24.039
unjust or adjust law is that there's some kind of
376
00:21:24.079 --> 00:21:27.359
other standard which seems to be more permanent that we
377
00:21:27.400 --> 00:21:32.039
can appeal to. When you ask a person what should
378
00:21:32.079 --> 00:21:38.359
one do, including what laws should we have, typically what
379
00:21:38.400 --> 00:21:42.240
they would go to is to a moral framework. And
380
00:21:42.279 --> 00:21:46.720
so the overarching question about what should one do is
381
00:21:46.759 --> 00:21:51.160
answered in multiple ways throughout history by people who present
382
00:21:51.720 --> 00:21:56.319
to communities and to individuals little packages which have got
383
00:21:56.640 --> 00:22:00.640
If you're like managine a little box which has got
384
00:22:00.640 --> 00:22:02.680
the label morality on it, and you open up the
385
00:22:02.720 --> 00:22:07.079
box and inside you can see some values, some principles
386
00:22:07.519 --> 00:22:10.839
in the case of religions, revealed truths, exemplary lives of
387
00:22:11.000 --> 00:22:15.960
saints and gurus and others. And somebody says, when to
388
00:22:16.000 --> 00:22:18.359
a person who's just recently be born into a community,
389
00:22:18.400 --> 00:22:20.759
they say, now, when you get presented with the question
390
00:22:20.759 --> 00:22:23.519
of what do want to do, here's the box. Look
391
00:22:23.559 --> 00:22:25.440
in the box. This will tell you what you ought
392
00:22:25.480 --> 00:22:29.279
to do. Now. The thing about a moral life is
393
00:22:29.279 --> 00:22:31.920
that it is possible to be inducted into that to
394
00:22:32.039 --> 00:22:35.640
learn how to apply those values and principles without ever
395
00:22:35.720 --> 00:22:40.039
really questioning it. And in fact, one can become habitually moral.
396
00:22:40.160 --> 00:22:45.400
So you might be habitually kind, habitually truthful. Why because
397
00:22:45.640 --> 00:22:48.720
that's the box I go to. So we've now got
398
00:22:48.799 --> 00:22:52.440
law morality, and then we come to the deeper level,
399
00:22:53.160 --> 00:22:57.960
and that is ethics. In ethics, particularly in its expression
400
00:22:58.000 --> 00:23:01.880
in the West, going back to Socrates, is by definition
401
00:23:03.160 --> 00:23:06.359
an examined life. It goes back to socrates claim that
402
00:23:06.400 --> 00:23:10.319
the unexamined life is not worth living. So although you
403
00:23:10.400 --> 00:23:13.440
may have the same values and principles as a person
404
00:23:13.480 --> 00:23:16.839
who's living the purely moral life. Is you're living an
405
00:23:16.880 --> 00:23:19.400
ethical life, then what you will do is you will
406
00:23:19.440 --> 00:23:22.599
stop and you will reflect. You will think about the
407
00:23:22.599 --> 00:23:25.799
circumstances in whi you're applying. You will ask yourself whether
408
00:23:25.880 --> 00:23:29.799
these values and these principles are the ones that ought
409
00:23:29.839 --> 00:23:31.799
to be implied. In fact, what you will have done
410
00:23:32.240 --> 00:23:34.920
is you will have taken something that was derivative, that
411
00:23:35.039 --> 00:23:38.359
is merely inherited from your culture or your religion, and
412
00:23:38.400 --> 00:23:40.720
you will have made it your own. And the kind
413
00:23:40.759 --> 00:23:44.000
of the psychological story of the difference is that when
414
00:23:44.000 --> 00:23:46.039
you're very young and you're growing up and somebody says, well,
415
00:23:46.079 --> 00:23:48.039
what should we do? The voice you hear in your
416
00:23:48.039 --> 00:23:51.039
head is that of your mum, your dad, or some
417
00:23:51.200 --> 00:23:55.759
community leader or influential figure. And then when you reach
418
00:23:55.839 --> 00:23:58.960
that point of ethical maturity, even though the answers might
419
00:23:59.000 --> 00:24:02.880
be the same your voice that you hear, I understand it.
420
00:24:03.079 --> 00:24:06.160
But it's essentially the examined life. And the claim made
421
00:24:06.160 --> 00:24:08.480
by philosophers, at least I would certainly make it and
422
00:24:08.599 --> 00:24:11.359
not being good company, is that this is the kind
423
00:24:11.400 --> 00:24:13.359
of life which is fit for the form of being
424
00:24:13.400 --> 00:24:18.799
in which we participate, namely human being. Because unlike other animals,
425
00:24:18.799 --> 00:24:21.960
we're definitely animals. We have instinct and desires, but we're
426
00:24:22.000 --> 00:24:24.680
not captured by them in the same way. We can
427
00:24:25.119 --> 00:24:27.000
go beyond what they might drive us to do and
428
00:24:27.039 --> 00:24:28.000
make conscious choices.
429
00:24:29.319 --> 00:24:32.000
And so what I love about that description is, I
430
00:24:32.039 --> 00:24:37.640
think most commonly when I, at least from a lawyer's perspective,
431
00:24:38.039 --> 00:24:41.640
when people talk about law, moral and ethics, it's in
432
00:24:41.680 --> 00:24:45.039
a hierarchy, so people would we lawyers put law at
433
00:24:45.039 --> 00:24:46.680
the top. And what I love about this is you're
434
00:24:46.759 --> 00:24:49.559
kind of flipping it. I say, you are, but a
435
00:24:49.599 --> 00:24:56.440
long tradition or philosophy tradition. When we look at governing AI,
436
00:24:56.519 --> 00:24:59.960
there is a much deeper tradition of ethics around AI
437
00:25:00.200 --> 00:25:03.599
then there is around hard law and regulation. What do
438
00:25:03.640 --> 00:25:08.359
you say in defense of ethics when people say that
439
00:25:08.400 --> 00:25:13.240
ethics aren't enough the artificial intelligence that we're currently experiencing.
440
00:25:13.559 --> 00:25:16.200
There are ethical frameworks that are in place, but the
441
00:25:17.000 --> 00:25:20.319
artificial intelligence that we're experiencing in society isn't ethical.
442
00:25:21.000 --> 00:25:23.440
So I would never say that we should not have
443
00:25:23.480 --> 00:25:27.359
any laws. What I worry about is that this goes
444
00:25:27.440 --> 00:25:31.000
right back to Aristotle, that you can be in a
445
00:25:31.039 --> 00:25:35.960
world where many many things are going wrong, not because
446
00:25:36.000 --> 00:25:40.880
of any malicious intent on part of any individual, but
447
00:25:40.960 --> 00:25:43.519
because good people are making bad decisions.
448
00:25:44.440 --> 00:25:44.519
It.
449
00:25:44.519 --> 00:25:47.279
I spend over thirty or around thirty five years now
450
00:25:47.319 --> 00:25:50.759
talking to people particularly where that's happened, and when you
451
00:25:50.799 --> 00:25:53.160
ask people to tell you about how they came to
452
00:25:53.200 --> 00:25:58.279
be responsible for making that choice, they say, with total sincerity,
453
00:25:59.119 --> 00:26:02.240
I just didn't say see it at the time. They
454
00:26:02.279 --> 00:26:04.200
can see it now, but they couldn't. And why couldn't
455
00:26:04.240 --> 00:26:07.160
they see it? The most common answer is because everybody
456
00:26:07.240 --> 00:26:09.680
was doing it. Well, that's just the way it was done.
457
00:26:09.839 --> 00:26:14.000
But one subspecies of that are people who say, when
458
00:26:14.039 --> 00:26:17.319
I was thinking about I was told if it's not illegal,
459
00:26:17.440 --> 00:26:22.079
it's not wrong. Now that move to say, well, if
460
00:26:22.079 --> 00:26:25.000
it's not illegal, it's not wrong is effectively to invite
461
00:26:25.039 --> 00:26:29.079
a kind of conditioned blindness where you learn not to
462
00:26:29.160 --> 00:26:32.599
look at a richer range of considerations because they simply
463
00:26:32.640 --> 00:26:36.599
fall outside the boundary of what the law contemplates. So
464
00:26:36.799 --> 00:26:38.720
I would want to have a world in which there
465
00:26:38.759 --> 00:26:43.559
is a strong ethos, that is ethical foundation which people
466
00:26:43.559 --> 00:26:47.240
are constantly wondering about, why we're doing this? Is that
467
00:26:47.319 --> 00:26:51.319
a legitimate purpose? What are the values and principles that
468
00:26:51.400 --> 00:26:54.559
really ought to apply for us at this time. They
469
00:26:54.559 --> 00:26:57.119
may be very stable, but they'd be constantly being thought
470
00:26:57.119 --> 00:26:59.839
about and reviewed, so that you're looking around, you're not
471
00:27:00.119 --> 00:27:04.640
being blind to certain dimensions of the world which are
472
00:27:04.680 --> 00:27:08.160
simply you're oblivious to, and therefore you find yourself like
473
00:27:08.240 --> 00:27:10.000
walking in a fog over a kind of the edge
474
00:27:10.039 --> 00:27:13.440
of a cliff, into an ethical death trap. And given
475
00:27:13.519 --> 00:27:16.240
that work, then I would say, of course, we want
476
00:27:16.279 --> 00:27:19.440
to use the resources of the law to be able
477
00:27:19.480 --> 00:27:23.440
to set some hard boundaries where necessary, and we've got
478
00:27:23.440 --> 00:27:27.640
to have the same approach to conscious institutional design to
479
00:27:27.680 --> 00:27:30.680
make sure that that's there. But I would say, let's
480
00:27:30.880 --> 00:27:36.400
not think about deploying things of power and utility in
481
00:27:36.440 --> 00:27:40.319
conditions of blindness where we just don't see what we're doing.
482
00:27:40.759 --> 00:27:45.119
And the best antidote I know to seeing more is
483
00:27:45.319 --> 00:27:46.440
what ethics has to offer.
484
00:27:46.720 --> 00:27:49.559
Yeah, I agree, but I didn't before I met you
485
00:27:49.640 --> 00:27:51.160
some and so that you go, you've got a convert
486
00:27:51.200 --> 00:27:55.200
you're exactly because I think you know, so many times
487
00:27:55.240 --> 00:27:57.839
I've heard people say, well, ask the lawyers is it legal?
488
00:27:58.079 --> 00:28:01.519
And the answer comes back, and that can't be the
489
00:28:01.640 --> 00:28:06.200
end of the conversation. You mentioned also earlier about affordance.
490
00:28:07.000 --> 00:28:10.519
Another a word that is used so often when we're
491
00:28:10.519 --> 00:28:14.279
talking about artificial intelligence is accountability, which is not quite
492
00:28:14.480 --> 00:28:17.960
not the same as affordance. But when we're talking about
493
00:28:18.160 --> 00:28:23.160
AI systems that are becoming more and more capable, what
494
00:28:23.240 --> 00:28:28.279
affordances do you think are minimum affordances that we need
495
00:28:28.319 --> 00:28:33.240
to build into systems? And then how do you see
496
00:28:33.279 --> 00:28:36.000
the difference between affordance and accountability.
497
00:28:37.359 --> 00:28:41.200
Okay, so I've already mentioned some of the affordances that
498
00:28:41.240 --> 00:28:44.440
I think we have. Rich The most important is the
499
00:28:44.480 --> 00:28:50.079
ability to monitor, whether using blockchain or other systems. For this,
500
00:28:51.119 --> 00:28:55.799
the extent to which an AI stack continues to preserve
501
00:28:55.960 --> 00:28:59.440
the original design intent in terms of its ethical content,
502
00:28:59.519 --> 00:29:03.319
it's perpose, its values and principles. Of course, I'm presuming
503
00:29:03.319 --> 00:29:05.400
in this answer that we have people who are building
504
00:29:05.440 --> 00:29:08.599
these systems who understand precisely what I'm getting at and
505
00:29:08.640 --> 00:29:11.759
the distinction between those three things, and that in their
506
00:29:11.799 --> 00:29:15.319
markdown files it's all there and it's published, and they've
507
00:29:15.319 --> 00:29:17.000
had the ability which I know because I've done the
508
00:29:17.000 --> 00:29:19.240
work on it, that it's technically possible to give these
509
00:29:19.279 --> 00:29:22.799
things effect. There's no point if what you get is
510
00:29:22.839 --> 00:29:27.599
a kind of subtle mutation in the underlying code that
511
00:29:27.920 --> 00:29:31.599
takes it away from those express purposes, values and principles.
512
00:29:31.599 --> 00:29:34.880
So a disinterested but expert third party has got to
513
00:29:34.880 --> 00:29:37.160
be able to auder this and see that it's still preserved.
514
00:29:37.319 --> 00:29:41.519
That's a critical affordance. Now. That then, though, runs into
515
00:29:41.720 --> 00:29:48.960
the roadblock which brings about the difference between affordances and accountability,
516
00:29:49.559 --> 00:29:52.480
and that is the fact that although we can explain
517
00:29:52.559 --> 00:29:56.400
what was intended, and we can explain what we sought
518
00:29:56.519 --> 00:30:00.000
to bring about, and I'm talking about particular large language models,
519
00:30:00.119 --> 00:30:03.279
new orsis of that kind, what we can't explain is
520
00:30:03.359 --> 00:30:08.839
precisely then what happened inside the black box. And because
521
00:30:09.000 --> 00:30:15.480
we can't do that, accountability in one sense is almost impossible,
522
00:30:15.920 --> 00:30:19.480
maybe literally impossible. And so what we need to do
523
00:30:19.960 --> 00:30:24.279
is to rely upon what I described as necessary fictions,
524
00:30:25.240 --> 00:30:27.319
of a kind that we've had to do in other places.
525
00:30:27.359 --> 00:30:32.599
So you take a liberal democratic society like Australia, we
526
00:30:32.839 --> 00:30:37.720
have still intact, although somewhat weakened. We rely on the
527
00:30:37.720 --> 00:30:40.960
doctrine of minister or responsibility, which is based on a
528
00:30:41.079 --> 00:30:45.079
necessary fiction, and that is that a minister in the
529
00:30:45.160 --> 00:30:51.039
executive actually knows about and is accountable to the Parliament
530
00:30:51.160 --> 00:30:54.599
for everything that is done in a department, no matter
531
00:30:54.640 --> 00:30:59.319
how sprawling and massive and complex it is. We know
532
00:30:59.480 --> 00:31:02.880
that it is impossible for any single individual to know
533
00:31:03.000 --> 00:31:07.240
every decision, to see within every nook and cranny. But
534
00:31:07.319 --> 00:31:11.720
it's a central to our democratic system that we pretend
535
00:31:12.359 --> 00:31:16.319
that they do. And anybody who volunteers to be a minister,
536
00:31:16.960 --> 00:31:18.839
when they put up their hand and say I will
537
00:31:18.880 --> 00:31:22.839
accept that commission, they should know that there is that
538
00:31:23.000 --> 00:31:27.519
necessary fiction that, despite it being literally impossible, nonetheless the
539
00:31:27.599 --> 00:31:30.200
conventions require that we pretend that it is true that
540
00:31:30.240 --> 00:31:33.119
they knew, so that when they stand up before Parliament
541
00:31:33.119 --> 00:31:37.200
they can be held accountable. In that sense, I think
542
00:31:37.240 --> 00:31:40.559
we need the equivalent necessary fiction in relation to AI,
543
00:31:41.240 --> 00:31:43.960
where at least with the state of the technology as
544
00:31:44.000 --> 00:31:47.759
it is now, in the most powerful models, we know
545
00:31:48.000 --> 00:31:50.960
that nobody, not even the person who created the system,
546
00:31:51.319 --> 00:31:54.839
can actually give you a step by step account of
547
00:31:54.880 --> 00:31:56.559
what has happened inside the black pot. It's not like
548
00:31:56.599 --> 00:31:58.720
Buolean logic, where a leads to be, leads to say
549
00:31:58.720 --> 00:32:01.400
and so forth. But it's a necessary fiction that the
550
00:32:01.400 --> 00:32:05.559
people in charge of it must nonetheless be accountable. They
551
00:32:05.599 --> 00:32:09.240
are responsible in that sense, even though they can't literally
552
00:32:09.359 --> 00:32:12.000
know for the operation of these systems, and that's where
553
00:32:12.039 --> 00:32:15.519
I think we've yet to map out how that actually works.
554
00:32:15.559 --> 00:32:17.880
I don't think we've even agreed on the fact that
555
00:32:17.920 --> 00:32:21.359
necessary fiction is required. But that's what I think is
556
00:32:21.400 --> 00:32:25.480
system as the technology is as of today, it requires
557
00:32:25.519 --> 00:32:27.559
and without that we can't have any accountability.
558
00:32:27.680 --> 00:32:30.079
Yeah, I definitely don't think we've reached the point of
559
00:32:30.319 --> 00:32:34.680
agreeing that fiction is required. It's persuasive that it is.
560
00:32:34.759 --> 00:32:36.880
I think all of the conversations that I'm having are
561
00:32:36.920 --> 00:32:39.960
about the fact that actually that cause of the chain
562
00:32:40.039 --> 00:32:44.279
of causation can't be established. Therefore we're in a challenging situation.
563
00:32:44.400 --> 00:32:46.640
So I like that framing.
564
00:32:47.200 --> 00:32:50.480
Nobody who is in charge of these systems is required
565
00:32:50.519 --> 00:32:53.039
to do so other than because they chose to place
566
00:32:53.079 --> 00:32:57.079
themselves in that position. And you got to understand that
567
00:32:57.119 --> 00:33:01.720
when you make that choice, it will almost certainly, if
568
00:33:01.759 --> 00:33:05.200
not now, sometime in the future, require the invocation of
569
00:33:05.200 --> 00:33:07.839
that necessary fiction, and you will be held accountable for
570
00:33:07.920 --> 00:33:10.559
what he's done, even though you cannot know. And pleading
571
00:33:10.559 --> 00:33:11.880
that as a defense way ambiotoqua.
572
00:33:12.039 --> 00:33:15.079
Yeah, I can certainly see that coming down the future,
573
00:33:15.079 --> 00:33:18.400
and I think that the expectations of the public in
574
00:33:18.440 --> 00:33:21.480
our new social compact that we were talking about will
575
00:33:21.559 --> 00:33:23.839
lean towards that. And this is a big part of
576
00:33:23.960 --> 00:33:26.240
how having people engaged in One of the things we
577
00:33:26.279 --> 00:33:28.000
try and do in Tech Mirrors to say to people
578
00:33:28.079 --> 00:33:31.240
get engaged in these convertations. Tell your politicians that you
579
00:33:31.319 --> 00:33:34.720
want this type of accountability because these are the decisions
580
00:33:34.720 --> 00:33:39.440
that are being made right now. Simon, Another thing that
581
00:33:39.720 --> 00:33:45.000
I always really enjoy talking about with you is what
582
00:33:45.119 --> 00:33:47.880
AI is forcing us to learn about what it is
583
00:33:47.920 --> 00:33:51.640
to be human. You have said, for example, that it's
584
00:33:51.680 --> 00:33:55.599
an AI is not just a technological revolution, but that
585
00:33:55.640 --> 00:33:58.920
it's a mirror forcing us to rediscover what it means
586
00:33:58.920 --> 00:34:02.200
to be human. Of them, we're on the Tech Mirror podcast,
587
00:34:02.279 --> 00:34:05.559
which is actually a play on exactly that concept. I
588
00:34:05.599 --> 00:34:09.480
couldn't let that go without asking you, So, what do
589
00:34:09.519 --> 00:34:12.880
you think that maybe we've forgotten about being human that
590
00:34:13.039 --> 00:34:16.480
AI is now making impossible to ignore? Or is that
591
00:34:16.559 --> 00:34:19.639
I saw you frowning there? Maybe challenge the challenge the
592
00:34:19.679 --> 00:34:20.639
premise of the question.
593
00:34:21.280 --> 00:34:23.039
No, I was just going to say, you're just a
594
00:34:23.079 --> 00:34:26.880
remarkably good listener. Most of what I say people just away,
595
00:34:26.920 --> 00:34:30.199
no one ever remembers it. I was feeling very impressed
596
00:34:30.199 --> 00:34:33.239
by the fact that you remember. Look, it's it is
597
00:34:33.280 --> 00:34:35.800
far more than technical. I mean, people, it's a commonplace
598
00:34:35.840 --> 00:34:38.960
to talk about it being sociod technical revolution, and it
599
00:34:39.000 --> 00:34:41.679
is a revolution in that sense. But the point I've
600
00:34:41.679 --> 00:34:44.039
made is that it comes back to some work I
601
00:34:44.079 --> 00:34:48.159
was doing for orthopedic surgeons around the world where I
602
00:34:48.199 --> 00:34:50.039
was asked in a number of different places here the
603
00:34:50.119 --> 00:34:52.199
US and lator in New Zealand to talk about the
604
00:34:52.320 --> 00:34:56.360
end of the surgeon, because I mean, it's not science
605
00:34:56.440 --> 00:34:59.800
fiction to imagine a world in which a surgeon walks
606
00:34:59.800 --> 00:35:04.519
in and operative and draped over the unethetized body of
607
00:35:04.559 --> 00:35:09.039
a human is a digital twin. And well, I'm taking
608
00:35:09.039 --> 00:35:11.280
a step. You might get all of the cues with
609
00:35:11.360 --> 00:35:14.519
an augmented reality to know where to cut, how to
610
00:35:14.519 --> 00:35:16.159
deap to cut a whole lot of things. And then
611
00:35:16.519 --> 00:35:18.320
somebody says, well, why do we need the human in
612
00:35:18.360 --> 00:35:20.840
the loop? Surely, if you're a person who's having an
613
00:35:20.880 --> 00:35:23.480
operation to remove a cancer, you want something which has
614
00:35:23.559 --> 00:35:26.679
got machine drade vision that can see beyond the spectrum
615
00:35:26.679 --> 00:35:29.920
that we can recognize. It's got the precision and the
616
00:35:30.039 --> 00:35:32.559
site to be able to distinguish between healthy flesh and
617
00:35:32.639 --> 00:35:36.559
disease first, so that only the minimum is taken with
618
00:35:36.800 --> 00:35:40.320
the slightest and say, well, why would you where's the
619
00:35:40.440 --> 00:35:42.679
human in this? And that's a kind of a paradigm
620
00:35:42.760 --> 00:35:47.280
case where that sort of question arises just as it
621
00:35:47.320 --> 00:35:50.440
does in something for example, in terms of diagnostics, where
622
00:35:51.199 --> 00:35:56.840
already AI is superior in its ability to look at
623
00:35:56.880 --> 00:36:02.679
pathology and histology and identify a particular disease condition. Now
624
00:36:02.719 --> 00:36:05.920
the difference is this, there is no machine on earth,
625
00:36:06.519 --> 00:36:10.880
no robot, no way I stack that can be in
626
00:36:10.960 --> 00:36:14.119
possession of them. I would say that you, the patient
627
00:36:14.960 --> 00:36:17.280
only has a matter of months to live. Who can
628
00:36:17.559 --> 00:36:20.880
tell you this in a way that another human being
629
00:36:20.920 --> 00:36:24.360
can where they reach out and they have a consoling
630
00:36:24.440 --> 00:36:27.599
touch on your shoulder and they say, I've got some
631
00:36:27.719 --> 00:36:31.719
terrible news. And as they tell you that news, you
632
00:36:31.920 --> 00:36:34.599
know that they know what it means to be mortal,
633
00:36:35.360 --> 00:36:39.440
that is, to contemplate the prospect of an unsought for
634
00:36:39.679 --> 00:36:44.320
and early death and the kind of empathy. I was
635
00:36:44.360 --> 00:36:48.239
in another discussion recently where there was a discussion about
636
00:36:48.280 --> 00:36:53.440
the use of AI to improve safety conditions in a
637
00:36:53.440 --> 00:36:57.159
really dangerous working environment, where of course you could have
638
00:36:57.239 --> 00:37:02.679
cameras and senses and automotive systems that could stop an
639
00:37:02.719 --> 00:37:05.440
at work, at risk worker from putting the hands eye
640
00:37:05.440 --> 00:37:08.760
into a dangerous conveyor belt and it could just stop
641
00:37:08.800 --> 00:37:13.039
it on the sad flashing light stop danger, danger. But
642
00:37:13.079 --> 00:37:16.400
that is so different from another human being who says
643
00:37:16.880 --> 00:37:19.760
stop watch out, where they convey to you a sense
644
00:37:19.840 --> 00:37:24.480
of their care for you. Now, these are distinctively human attributes.
645
00:37:24.519 --> 00:37:29.760
And the question that I've had really it. It started
646
00:37:29.760 --> 00:37:31.719
to come to my mind when I saw people writing
647
00:37:31.760 --> 00:37:36.960
on menus that they were hand cut chips. I thought,
648
00:37:37.599 --> 00:37:39.920
why would you want to tell me that the potatoes
649
00:37:39.960 --> 00:37:43.639
have been cut by hand? And obviously there's a sense
650
00:37:43.800 --> 00:37:46.320
that the person is making the chips and selling the
651
00:37:46.360 --> 00:37:48.639
chips things. You're going to value this that you would
652
00:37:48.719 --> 00:37:50.840
rather have a hand cut chip than a chip cut
653
00:37:50.880 --> 00:37:55.079
by And this of course predated the emergence of artisanal
654
00:37:55.239 --> 00:37:59.599
breweries and bakeries. And there's something about that which bears
655
00:37:59.599 --> 00:38:03.719
the mark it's maker, which is value. There's something about
656
00:38:04.039 --> 00:38:08.039
the human sense of empathy, about knowledge of our mortality.
657
00:38:08.159 --> 00:38:12.400
So there are just three examples of things which you excellence.
658
00:38:12.679 --> 00:38:16.039
A technical excellence of what the machine can do forces
659
00:38:16.039 --> 00:38:18.719
you back to, so, well, what do I bring that's
660
00:38:18.760 --> 00:38:21.840
distinctive about my humanity that simply can't be replicated.
661
00:38:21.840 --> 00:38:24.760
I love that you've managed to link in their mortality,
662
00:38:24.960 --> 00:38:28.559
chips and craft beer all in one. All in one
663
00:38:28.679 --> 00:38:33.639
response wouldn't exactly, it's amazing. So someone I'm conscious of
664
00:38:33.840 --> 00:38:37.159
coming up to times, I've got two questions left for you.
665
00:38:37.400 --> 00:38:40.000
The first one is about trust. This is something that
666
00:38:40.000 --> 00:38:43.199
we've been focused on a lot in many different conversations
667
00:38:43.199 --> 00:38:45.920
that we've been having, but also that society is having
668
00:38:45.920 --> 00:38:48.760
at the moment in terms of do we trust AI
669
00:38:49.199 --> 00:38:51.119
and should we trust AI? I think there's a very
670
00:38:51.199 --> 00:38:53.440
large question about whether there should be a presumption that
671
00:38:53.480 --> 00:38:55.840
we want to trust AI, or whether not trusting it
672
00:38:55.880 --> 00:38:59.199
in and of itself is a good thing. From your perspective,
673
00:38:59.760 --> 00:39:03.360
what are the conditions for trust? How do you build trust?
674
00:39:03.719 --> 00:39:07.320
Specifically looking at it in the context of AI.
675
00:39:08.519 --> 00:39:13.719
Two things are required for trust. Firstly, you must declare
676
00:39:14.320 --> 00:39:16.559
that is the person if you desire to be trusted.
677
00:39:16.599 --> 00:39:19.800
You must declare the standard by which you are prepared
678
00:39:19.840 --> 00:39:23.519
to be judged, and that standard will be in the
679
00:39:23.519 --> 00:39:26.000
form of your values and principles. So these are the
680
00:39:26.119 --> 00:39:28.679
things if you watch me, These are the choices I
681
00:39:28.719 --> 00:39:31.159
will make. These are the good things I will pursue
682
00:39:31.440 --> 00:39:34.880
by the means conditioned by the principles which I declare.
683
00:39:35.480 --> 00:39:37.679
That's the first step, because if you don't declare that
684
00:39:38.039 --> 00:39:40.519
up front, no one can know how to judge what
685
00:39:40.559 --> 00:39:43.840
you do. It'll just be there's no point of calibration,
686
00:39:43.920 --> 00:39:46.760
there's no way of seeing how far you deviate from
687
00:39:46.880 --> 00:39:51.039
or accord with some standard for judgment. So it's essential
688
00:39:51.119 --> 00:39:54.400
you declare the core values and principles which are going
689
00:39:54.400 --> 00:39:56.079
to be the basis for which you are prepared to
690
00:39:56.079 --> 00:39:59.000
be judged. And of course the second thing is extent
691
00:39:59.079 --> 00:40:02.760
to which you act in accordance with these things. So
692
00:40:02.800 --> 00:40:05.880
they are the two obvious parts. Then there's the third part,
693
00:40:06.159 --> 00:40:10.199
which is often overlooked in that people have not for
694
00:40:10.239 --> 00:40:13.320
a long time, if ever, but certainly not for a
695
00:40:13.360 --> 00:40:16.800
long time. Simply judge whether a person can be trusted
696
00:40:17.440 --> 00:40:20.199
as a result of their own direct experience. So, Johannah,
697
00:40:20.239 --> 00:40:22.239
if you said to me, these are my values and principles,
698
00:40:22.280 --> 00:40:23.960
and simon, I'm going to treat you like this, And
699
00:40:24.400 --> 00:40:26.719
if one hundred percent of the time that's perfectly aligned.
700
00:40:27.239 --> 00:40:30.360
Then I would say that's all well and good, but
701
00:40:31.480 --> 00:40:34.440
how do you treat other people? People who are not
702
00:40:34.599 --> 00:40:39.000
made so for corporations, for example, they might think if
703
00:40:39.000 --> 00:40:42.400
there be retail of example, as hell, we've told people
704
00:40:42.400 --> 00:40:45.159
who am we treat our customers like this every day.
705
00:40:45.199 --> 00:40:48.480
We're relentless. But they need to realize that their customers
706
00:40:48.480 --> 00:40:50.760
is going to be asking, well, well, that's very good,
707
00:40:50.800 --> 00:40:52.559
but how do you trust how do you treat each
708
00:40:52.559 --> 00:40:56.519
other inside that company? How do you treat your suppliers?
709
00:40:57.000 --> 00:40:59.639
Because if we can see a gap between what you
710
00:40:59.719 --> 00:41:04.320
say you stand for and how you treat anybody, then
711
00:41:04.360 --> 00:41:08.639
because we're skeptical and sometimes even we're cynical, so aha,
712
00:41:09.119 --> 00:41:13.400
there's the exception that proves the rule. So it's no
713
00:41:13.559 --> 00:41:18.320
easy thing to merit the trust of others because you've
714
00:41:18.320 --> 00:41:20.320
got to be clear about that standard. But then you've
715
00:41:20.320 --> 00:41:22.519
got to be consistent in the way you apply those
716
00:41:22.599 --> 00:41:27.000
values and principles across the board. And of course there
717
00:41:27.000 --> 00:41:29.159
can be tensions within these values and principles. I wish
718
00:41:29.199 --> 00:41:31.559
I could say we lived in a universe where everything
719
00:41:31.599 --> 00:41:34.159
that's good lined up and there was never attention. The
720
00:41:34.199 --> 00:41:36.920
truth is that most of the challenging ethical questions we
721
00:41:36.960 --> 00:41:38.960
face in our lives are not between good and bad,
722
00:41:39.320 --> 00:41:42.039
right and wrong. It's good versus good, it's right versus right.
723
00:41:43.440 --> 00:41:46.000
That's and that's why people don't like ethics so much.
724
00:41:46.039 --> 00:41:48.599
So much prefer the self kind of the certainty that
725
00:41:48.599 --> 00:41:51.119
a fundamental is to one kind or another, political, religious
726
00:41:51.159 --> 00:41:53.079
or otherwise, and give order to escape. You are kind
727
00:41:53.079 --> 00:41:55.199
of a head and nestic haze, and that someone else
728
00:41:55.239 --> 00:41:57.239
sorts it out. But if you're prepared to stay in
729
00:41:57.239 --> 00:42:00.960
that space, it is messy. But that means then that
730
00:42:01.000 --> 00:42:03.599
if you are, as a person wanting to be trusted,
731
00:42:04.000 --> 00:42:07.719
perhaps having to compromise between truth and compassion or some
732
00:42:08.119 --> 00:42:10.440
other set of competitor, you should be open about that.
733
00:42:10.559 --> 00:42:14.400
You shouldn't pretend, oh, well, you know, this is a
734
00:42:14.440 --> 00:42:17.480
good outcome, so it excuses me doing that. You could say, no, actually,
735
00:42:17.519 --> 00:42:20.039
I wrestle with this, and I violated one of my
736
00:42:20.079 --> 00:42:23.280
own values, the ones I've told you about. But this
737
00:42:23.360 --> 00:42:24.960
is why I did it, and this is the way
738
00:42:25.000 --> 00:42:27.599
I tried to preserve that one to the greatest extent possible.
739
00:42:28.000 --> 00:42:31.079
And typically in public life, we don't hear those conversations.
740
00:42:31.079 --> 00:42:33.360
We hear people saying more often, oh, well I just
741
00:42:33.440 --> 00:42:35.119
made this is the right choice. Ultimate we will be
742
00:42:35.159 --> 00:42:38.719
proven to be correct by some index of judgment, and
743
00:42:38.760 --> 00:42:42.360
so we lose some of that fine granularity that we want,
744
00:42:43.119 --> 00:42:45.840
and in losing that, that's how trust begins to it away,
745
00:42:45.880 --> 00:42:48.199
because people say, oh, you didn't you didn't notice that
746
00:42:48.239 --> 00:42:50.280
you violated one of your own values or principles, or
747
00:42:50.280 --> 00:42:52.639
you didn't feel like you had to tell me why,
748
00:42:52.760 --> 00:42:56.039
or you didn't say how you balanced it, and people say, well,
749
00:42:56.320 --> 00:42:57.960
you're a hypocrite. I can't trust you.
750
00:42:58.119 --> 00:43:01.079
Yeah, And I think you see that really prevalent in
751
00:43:01.159 --> 00:43:03.400
many of the conversations where you have tech companies who
752
00:43:03.480 --> 00:43:06.800
say things like no regulate us, we want to be regulated,
753
00:43:06.880 --> 00:43:10.239
and then their actions are quite different or they'll push
754
00:43:10.360 --> 00:43:13.880
back on enforcement. And so I think when you're describing
755
00:43:13.880 --> 00:43:15.920
it as these are the standards we live by and
756
00:43:16.039 --> 00:43:19.360
then actually living by those stands, demonstrating that they are
757
00:43:19.519 --> 00:43:23.360
actually putting those standards walking the walk, that's often I
758
00:43:23.360 --> 00:43:28.320
think where there is this mismatch in trust with large companies.
759
00:43:29.599 --> 00:43:32.440
And it's not just that they say one thing and
760
00:43:32.480 --> 00:43:37.199
then say do something explicitly which is at odds with that.
761
00:43:37.920 --> 00:43:42.559
What they also do without necessarily realizing this is that
762
00:43:42.599 --> 00:43:48.960
they build systems, policies, and structures which conveying messages that
763
00:43:49.039 --> 00:43:51.960
are completely at odds with what they say. So we've
764
00:43:51.960 --> 00:43:54.760
seen this, for example in banking and financial services for
765
00:43:54.800 --> 00:43:56.519
a period of time, where they say, not what customers
766
00:43:56.519 --> 00:43:59.280
are really important, and yet all of their KPI has
767
00:43:59.320 --> 00:44:01.400
had nothing to do with the well being of the customer.
768
00:44:01.440 --> 00:44:04.320
It was all about profitability. And people said, well, if
769
00:44:04.320 --> 00:44:08.440
you say that, but you've created systems which drive people
770
00:44:08.480 --> 00:44:11.280
to do something else, then you are hypocrites. And that
771
00:44:11.920 --> 00:44:17.239
charge of hypocrisy is lethal. It just creates cynicism, It
772
00:44:17.679 --> 00:44:22.320
destroys the bonds of association and community. And so often
773
00:44:22.880 --> 00:44:25.519
it's not a product of any kind of deliberate intent
774
00:44:25.599 --> 00:44:27.760
on the part of people who are CEOs or otherwise.
775
00:44:27.920 --> 00:44:30.960
They just don't think. They're just part of that state
776
00:44:31.000 --> 00:44:36.000
of conditioned blindness we talked about before, and they destroy trust.
777
00:44:36.559 --> 00:44:38.519
And of course, just on your point about them saying
778
00:44:38.519 --> 00:44:41.639
we like regulation, it's really interesting. I've seen this now
779
00:44:42.079 --> 00:44:43.800
over a long period of time. If you give a
780
00:44:43.920 --> 00:44:50.000
choice to a company between voluntarily choosing to apply a
781
00:44:50.079 --> 00:44:54.280
high ethical standard or instead being forced to apply that
782
00:44:54.360 --> 00:44:58.400
standard by regulation. So often they will say please regulate
783
00:44:58.440 --> 00:45:03.000
me because to crave some kind of illusory sense of certainty,
784
00:45:03.519 --> 00:45:05.880
or they think it will force their competitors to adopt
785
00:45:05.880 --> 00:45:09.639
the same standard, because they have this hunch that perhaps
786
00:45:09.679 --> 00:45:11.719
if they were to adopt the higher standard, it may
787
00:45:11.760 --> 00:45:15.199
actually work against them in a competitive marketplace. In fact,
788
00:45:15.199 --> 00:45:17.000
there's as much evidence to show that in fact, it
789
00:45:17.000 --> 00:45:18.440
could be to their advantage.
790
00:45:18.639 --> 00:45:21.320
So Simon, I think that's absolutely right in terms of
791
00:45:21.360 --> 00:45:25.840
what you're saying of the power of the potential commercial
792
00:45:25.880 --> 00:45:30.000
benefit that could come from companies that recognize that actually
793
00:45:30.079 --> 00:45:33.480
designing and making ethical AI is a competitive advantage, and
794
00:45:33.519 --> 00:45:36.360
you're starting to see that in the market. Not a lot,
795
00:45:36.400 --> 00:45:37.920
but you're starting to see that in that look.
796
00:45:38.000 --> 00:45:41.360
I think the story of Anthropic recently is possibly an
797
00:45:41.400 --> 00:45:44.159
example of that. It turns out to be a good story.
798
00:45:44.400 --> 00:45:49.440
It could go bright the way, but let's end. Let's
799
00:45:49.519 --> 00:45:53.119
keep that positive tone to end on for us. Can
800
00:45:53.159 --> 00:45:55.239
I ask you to help us to paint a picture
801
00:45:55.559 --> 00:45:59.519
of what the world could look like in ten years
802
00:45:59.599 --> 00:46:04.239
time if we have more people waking up to the
803
00:46:04.280 --> 00:46:08.239
power of well, the fact that this is not technology
804
00:46:08.280 --> 00:46:11.320
is not evolving in any particular way, inevitably that it
805
00:46:11.360 --> 00:46:14.159
can evolve in many different ways, that we have the
806
00:46:14.159 --> 00:46:18.119
power to shape it, but also the power of ethics
807
00:46:18.440 --> 00:46:22.239
to help us to frame the way that AI may
808
00:46:22.280 --> 00:46:25.800
be unfolding into our society. What could the world look
809
00:46:25.920 --> 00:46:27.719
like in a positive way?
810
00:46:28.559 --> 00:46:34.360
Okay, so let's go back to ancient Athens and Aristotle's
811
00:46:34.480 --> 00:46:38.679
view that ethics and politics were different sides of the
812
00:46:38.760 --> 00:46:42.400
same coin, one being the good life of the individual
813
00:46:42.440 --> 00:46:44.519
and the other the good life for a whole community
814
00:46:44.840 --> 00:46:48.960
or society. Now, I think what we should do is
815
00:46:49.000 --> 00:46:53.159
to think in those terms of not just as a
816
00:46:53.199 --> 00:46:57.400
technical question to be resolved or informed by ethics. It
817
00:46:57.559 --> 00:46:59.880
is a much larger system about the kind of society
818
00:47:00.039 --> 00:47:02.320
we want to be. So we're going to have to
819
00:47:02.320 --> 00:47:05.800
think about a world which I'm really optimistic about what's possible,
820
00:47:06.599 --> 00:47:12.280
in which all of the dangerous, dirty and drudge like
821
00:47:12.400 --> 00:47:16.719
work is taken over by systems that can do it
822
00:47:16.800 --> 00:47:21.039
without any exposure to pain or distress, and it alleviates
823
00:47:21.239 --> 00:47:23.880
us of those burdens in a way that has not
824
00:47:24.000 --> 00:47:27.639
been true in the past. But then at the same time,
825
00:47:27.719 --> 00:47:30.119
which causes us to ask, well, what is a good
826
00:47:30.239 --> 00:47:33.599
life for a person living in such a society, and
827
00:47:33.639 --> 00:47:37.599
what are the economic and political adjustments, And they're going
828
00:47:37.639 --> 00:47:40.280
to be profound that we need to make in order
829
00:47:40.320 --> 00:47:44.039
to sustain such a society in which everybody can flourish.
830
00:47:44.280 --> 00:47:47.800
I would love to live it in that world where
831
00:47:48.480 --> 00:47:52.760
all that is beneficial can be harnessed, with the risks
832
00:47:52.800 --> 00:47:56.920
well contained, and where it serves the purpose of lifting
833
00:47:57.000 --> 00:48:01.840
up a society rather than simply enslaving it to a
834
00:48:01.880 --> 00:48:06.519
particular model or making human beings commodities who work in
835
00:48:06.639 --> 00:48:10.480
service of the machine. We've seen Fritz Lang's film Metropolis.
836
00:48:10.639 --> 00:48:13.440
If anybody's seen that, that's the vision. It's actually an
837
00:48:13.440 --> 00:48:18.639
incredibly dystopian view of what technology could become if it's
838
00:48:18.719 --> 00:48:21.719
given or its logic is allowed to unfold in a
839
00:48:21.719 --> 00:48:24.440
way that's indifferent to the well being and people. And
840
00:48:24.480 --> 00:48:26.719
I'd encourage anybody to look at that just to see
841
00:48:27.119 --> 00:48:30.519
what it anticipates could happen. I don't think that is
842
00:48:30.559 --> 00:48:33.320
in any sense inevitable. I think it could be a
843
00:48:33.400 --> 00:48:36.599
brilliant future, particularly for a country like Australia, which has
844
00:48:36.639 --> 00:48:39.840
got all the natural resources and the people and the
845
00:48:39.840 --> 00:48:44.280
institutional arrangements to create an extraordinary period of which could
846
00:48:44.280 --> 00:48:47.400
be the most just and prosperous society the world's even known.
847
00:48:47.639 --> 00:48:49.880
And in doing that it could be aided by a
848
00:48:49.920 --> 00:48:52.639
combination of AI, robotics and other things, so I can
849
00:48:52.679 --> 00:48:53.679
realize that potential.
850
00:48:54.239 --> 00:48:56.000
And if you could make if you could have a
851
00:48:56.039 --> 00:48:59.280
magic wand and change one thing about the way people
852
00:48:59.360 --> 00:49:02.400
think about an ethics to help make that future come true,
853
00:49:02.400 --> 00:49:03.840
what would it be simon.
854
00:49:04.840 --> 00:49:08.320
Comparably to get people to be curious about why things
855
00:49:08.320 --> 00:49:12.639
that have unleash your curiosity and ask is what's been
856
00:49:12.719 --> 00:49:15.800
proposed in a form which allows you to identify the
857
00:49:15.840 --> 00:49:17.960
purpose which is going to be served? Is it its
858
00:49:18.000 --> 00:49:22.320
purpose that you can approve? And then those the value
859
00:49:22.360 --> 00:49:24.519
do the values and principles which they're saying they're trying
860
00:49:24.559 --> 00:49:26.920
to embody align with yours? And if not, what are
861
00:49:26.960 --> 00:49:28.760
you going to do about it? Rather than washing your
862
00:49:28.760 --> 00:49:32.599
hands and being indifferent. And it may simply be just
863
00:49:32.760 --> 00:49:35.719
what small purchase you make where you exercise a little
864
00:49:35.719 --> 00:49:38.239
bit of economic sovereignty, Or it could be the way
865
00:49:38.280 --> 00:49:40.199
you vote, or the letting you write to your MP,
866
00:49:40.440 --> 00:49:43.079
or the fact that you juggle meeting or attend something
867
00:49:43.079 --> 00:49:45.719
of interest. Any of those things can collectively make a
868
00:49:45.719 --> 00:49:46.360
big difference.
869
00:49:48.159 --> 00:49:50.960
Simon, I always note when I get emails from you,
870
00:49:51.039 --> 00:49:53.800
this that you have in your signature that an unexamined
871
00:49:53.880 --> 00:49:56.719
life is not worth living from Socrates, which you've referred
872
00:49:56.719 --> 00:49:59.599
to many times. But also thank you so much for
873
00:49:59.679 --> 00:50:02.679
chilling challenging us to also think about that and to
874
00:50:02.719 --> 00:50:06.639
apply those concepts. I've really enjoyed this conversation. Thank you
875
00:50:06.679 --> 00:50:08.599
for joining us, and we hope to talk to you
876
00:50:08.679 --> 00:50:09.519
again sometime soon.
877
00:50:09.920 --> 00:50:12.239
I've really enjoyed it too, and thanks for having me on.
878
00:50:13.000 --> 00:50:22.880
Thanks saman Well, that's it for this episode of Tech Mirror,
879
00:50:22.920 --> 00:50:26.039
which is brought to you by the Tech Policy Design Institute.
880
00:50:26.440 --> 00:50:28.679
We're based here in Canberra on the lands of the
881
00:50:28.719 --> 00:50:32.760
nun Wall Nambri people. If you found today's conversation useful
882
00:50:32.880 --> 00:50:35.760
or thought provoking, please do share it with a friend
883
00:50:35.840 --> 00:50:39.000
or a colleague, or leave a review and subscribe wherever
884
00:50:39.039 --> 00:50:42.599
you get your podcasts. If you're watching, please don't forget
885
00:50:42.639 --> 00:50:45.639
to like and follow. For show notes, you can visit
886
00:50:45.679 --> 00:50:49.800
tech Policy dot Au Forward Slash podcast and this podcast
887
00:50:49.920 --> 00:50:53.840
was made possible with thanks to the generous contributions from government,
888
00:50:54.000 --> 00:50:58.079
industry and philanthropy to the Tech Policy Design Fund. The
889
00:50:58.119 --> 00:51:01.559
full details of which are available on our website. The
890
00:51:01.639 --> 00:51:04.719
team at Audiocraft produced this pod on the lands of
891
00:51:04.760 --> 00:51:09.000
the Gadigul people of the Eora nation. Music is Vitali escopellis.
892
00:51:09.559 --> 00:51:11.920
A big thank you also to the team at the
893
00:51:11.960 --> 00:51:15.559
Tech Policy Design Institute, without whom this pod would not
894
00:51:15.639 --> 00:51:19.559
be possible. Thank you for joining us, and as always,
895
00:51:19.800 --> 00:51:21.679
get in touch and get involved.