1
00:00:00,400 --> 00:00:03,277
So the theory of great geniuses is that

2
00:00:03,360 --> 00:00:05,997
they understand one area extremely well

3
00:00:06,080 --> 00:00:08,957
and they're so brilliant the lady or man

4
00:00:09,040 --> 00:00:11,037
can then take their ideas and apply it

5
00:00:11,120 --> 00:00:14,237
to a completely different domain. If we

6
00:00:14,320 --> 00:00:16,877
can solve that problem then I think it's

7
00:00:16,960 --> 00:00:21,837
over. Then we get to AGI and then
it's a whole different world.

8
00:00:21,920 --> 00:00:25,757
>> So Eric Schmidt just gave another
eye openening interview about AGI. He

9
00:00:25,840 --> 00:00:30,397
unveiled the exact how we get to AGI and
also mapped the time we can achieve it.

10
00:00:30,480 --> 00:00:32,637
Let's listen to this before
we dive in deeper.

11
00:00:32,720 --> 00:00:35,677
>> It feels like there's
been um a pullback

12
00:00:35,760 --> 00:00:39,037
a little bit from the AGI narrative
which I think I think is actually a good

13
00:00:39,120 --> 00:00:43,437
thing. I think it's more conducive
to calm rational policym. What's your

14
00:00:43,520 --> 00:00:45,357
perception of AGI right now? Where where

15
00:00:45,440 --> 00:00:48,637
are we on that whole train? So, so, um,

16
00:00:48,720 --> 00:00:50,477
so first place, the speech that the

17
00:00:50,560 --> 00:00:55,757
president delivered about a month ago
about AI strategy, which I think you

18
00:00:55,840 --> 00:00:59,997
probably wouldn't say it, but you kind
of wrote it for him, was exactly right.

19
00:01:00,080 --> 00:01:03,437
Right. So, thank you.
>> David collaborated with an amazing

20
00:01:03,520 --> 00:01:06,797
leader who we all respect and admire so

21
00:01:06,880 --> 00:01:09,277
much, Eric.

22
00:01:09,360 --> 00:01:12,877
>> Yes. Uh, so, nevertheless,

23
00:01:12,960 --> 00:01:16,717
>> saying I wrote it was was way too
strong. I mean, actually, but anyway, If

24
00:01:16,800 --> 00:01:17,997
you didn't write it, then it must have

25
00:01:18,080 --> 00:01:21,437
been your twin. But in any case, um the

26
00:01:21,520 --> 00:01:24,077
you got you got the emphasis right,

27
00:01:24,160 --> 00:01:26,477
which was that investment in research,

28
00:01:26,560 --> 00:01:31,437
investment in the kind of stuff
that we do is really really important.

29
00:01:31,520 --> 00:01:35,597
I don't agree with you on this on this
AGI thing because there's this group

30
00:01:35,680 --> 00:01:38,397
which I call the San Francisco um

31
00:01:38,480 --> 00:01:42,637
narrative because they all live in San
Francisco and their narrative goes

32
00:01:42,720 --> 00:01:47,357
something like this. Um today we're
doing agents. Uh the agentic revolution

33
00:01:47,440 --> 00:01:51,837
will change businesses which I agree with.
Um that what happens is the

34
00:01:51,920 --> 00:01:56,077
systems will become
recursively self-intelligent.

35
00:01:56,160 --> 00:02:00,717
With recursive self-improvement as it's
called if you have a scale-free problem

36
00:02:00,800 --> 00:02:02,877
and a scale-free problem for example is

37
00:02:02,960 --> 00:02:04,957
programming or math where you can just

38
00:02:05,040 --> 00:02:07,757
keep doing it. You get these enormous

39
00:02:07,840 --> 00:02:12,237
fast gains if you buy enough hardware,
do enough software so forth and so on.

40
00:02:12,320 --> 00:02:14,637
that is still underway.

41
00:02:14,720 --> 00:02:17,597
The collective of that says that in the

42
00:02:17,680 --> 00:02:20,637
next three-ish years they believe that

43
00:02:20,720 --> 00:02:23,357
we will get forms of super intelligence

44
00:02:23,440 --> 00:02:26,397
and the way they define it is basically

45
00:02:26,480 --> 00:02:29,277
a savant a chemist soant a physics soant

46
00:02:29,360 --> 00:02:33,517
a mathematician soant I don't agree
with the three years but I do agree that

47
00:02:33,600 --> 00:02:38,157
it'll be maybe six or seven years
>> but if it's a savant in you know a

48
00:02:38,240 --> 00:02:41,197
particular area
is that general intelligence

49
00:02:41,280 --> 00:02:46,317
>> it's not general intelligence Yet
general intelligence is when it can set

50
00:02:46,400 --> 00:02:48,637
its own objective function. >> Right?

51
00:02:48,720 --> 00:02:50,237
>> And >> there's no evidence of that.

52
00:02:50,320 --> 00:02:53,357
>> There's no evidence right now
of the ability to set your own objective

53
00:02:53,440 --> 00:02:55,917
function. Um the the the thinking and

54
00:02:56,000 --> 00:02:57,837
I'm writing a paper on this so I've been

55
00:02:57,920 --> 00:03:00,957
studying it is that the the the

56
00:03:01,040 --> 00:03:05,997
technical problem is non-stationerity
of mathematical proofs. And that what

57
00:03:06,080 --> 00:03:09,357
you're doing is you're trying to solve
against objective function but the

58
00:03:09,440 --> 00:03:13,517
objective function keeps changing which
is how humans operate. your goal changes

59
00:03:13,600 --> 00:03:15,117
every day. Whereas computers have

60
00:03:15,200 --> 00:03:17,677
trouble with that. As a math problem, we

61
00:03:17,760 --> 00:03:20,717
don't have an algorithm yet for LLMs

62
00:03:20,800 --> 00:03:22,557
that can do that. People are working on

63
00:03:22,640 --> 00:03:26,637
it. Um and the the test will be can you

64
00:03:26,720 --> 00:03:29,277
basically um using the information

65
00:03:29,360 --> 00:03:32,557
available in 1902, can you derive the

66
00:03:32,640 --> 00:03:34,157
same thing that Einstein did with

67
00:03:34,240 --> 00:03:36,397
special relativity followed by general

68
00:03:36,480 --> 00:03:39,997
relativity? We cannot do that today. Um

69
00:03:40,080 --> 00:03:42,797
and most people believe that the way

70
00:03:42,880 --> 00:03:45,357
this will be solved is through analogy.

71
00:03:45,440 --> 00:03:48,317
So the theory of great geniuses is that

72
00:03:48,400 --> 00:03:51,037
they understand one area extremely well

73
00:03:51,120 --> 00:03:53,997
and they're so brilliant the lady or man

74
00:03:54,080 --> 00:03:55,997
can then take their ideas and apply it

75
00:03:56,080 --> 00:03:59,277
to a completely different domain. If we

76
00:03:59,360 --> 00:04:01,837
can solve that problem then I think it's

77
00:04:01,920 --> 00:04:05,357
over. Then we get to AGI and then it's a

78
00:04:05,440 --> 00:04:07,037
whole different world.

79
00:04:07,120 --> 00:04:09,357
I think one of the reasons why it's hard

80
00:04:09,440 --> 00:04:13,437
to replace a human and you know JK and I

81
00:04:13,520 --> 00:04:17,757
debate this is that humans are end to end.
You know we can do the whole job.

82
00:04:17,840 --> 00:04:20,477
You have sort of a complete
understanding. You can pivot very

83
00:04:20,560 --> 00:04:24,797
easily. AI at least as we know it
today is not end to end. It has to be

84
00:04:24,880 --> 00:04:28,637
prompted. You get an answer. That answer
has to be validated. Then you have to

85
00:04:28,720 --> 00:04:31,597
ask a new question because it never
gives you exactly what you want. You

86
00:04:31,680 --> 00:04:34,637
have to apply more context. You have
to go through an iterative loop. Finally,

87
00:04:34,720 --> 00:04:39,117
you get to an answer that has business
value. The way biology puts it is that

88
00:04:39,200 --> 00:04:41,037
AI is not endtoend. It's middle to

89
00:04:41,120 --> 00:04:43,197
middle. Humans are endtoend. And so, as

90
00:04:43,280 --> 00:04:46,157
a result of that, instead of AI

91
00:04:46,240 --> 00:04:48,797
replacing all of us, AI will be very

92
00:04:48,880 --> 00:04:53,277
synergistic with humans because we can
define the objective function. We do the

93
00:04:53,360 --> 00:04:57,357
prompting and we work with it to iterate
and it does a lot of the work in the

94
00:04:57,440 --> 00:05:00,237
middle. Um, that seems to me like a very

95
00:05:00,320 --> 00:05:03,437
optimistic, less doomeristic take on it.

96
00:05:03,520 --> 00:05:07,837
What you just said is exactly what's
going to happen for the next few years

97
00:05:07,920 --> 00:05:10,317
that each of us will have assistance

98
00:05:10,400 --> 00:05:12,797
which on our command and our prompting

99
00:05:12,880 --> 00:05:17,517
will be incredibly helpful to whatever
problem we have you know personal uh you

100
00:05:17,600 --> 00:05:20,477
have people who are using these things
for relationship advice for you know

101
00:05:20,560 --> 00:05:24,157
talking to their kids I mean it's
all crazy stuff um but the fact of the

102
00:05:24,240 --> 00:05:28,557
matter is that's it the to me the real
question is when does it cross over to

103
00:05:28,640 --> 00:05:32,557
having its own valition its own ability
to seek information and solve new

104
00:05:32,640 --> 00:05:34,317
problems s that's a different animal.

105
00:05:34,400 --> 00:05:37,677
>> But have we seen any evidence
of recursive self-improvement yet?

106
00:05:37,760 --> 00:05:40,717
>> Um, not yet. I'm I'm I've funded a

107
00:05:40,800 --> 00:05:44,957
number of startups which claim to be
close to it, but of course these are

108
00:05:45,040 --> 00:05:47,277
startups and you never know, which tells

109
00:05:47,360 --> 00:05:49,997
me it's 5 10 years.

110
00:05:50,080 --> 00:05:54,477
>> Okay, so just in 5 to 10 years, that
will be enormously fast for a new tech

111
00:05:54,560 --> 00:05:56,477
industry. And recursive self-improvement

112
00:05:56,560 --> 00:06:01,837
is the key element that define AGI
and ASI. Let's go deeper into this with

113
00:06:01,920 --> 00:06:04,477
Eric. >> People who are
now looking at how do you

114
00:06:04,560 --> 00:06:09,037
control a super intelligence with a dumb
intelligence, right? And there's math

115
00:06:09,120 --> 00:06:12,557
that I did not understand which says
that to some degree you can contain it,

116
00:06:12,640 --> 00:06:17,277
but we don't know. Um, but I think
this the first point to make is that

117
00:06:17,360 --> 00:06:21,037
recursive self-improvement or recursive
intelligence is where the system is

118
00:06:21,120 --> 00:06:23,837
learning itself and you're beginning to

119
00:06:23,920 --> 00:06:25,837
see agentic solutions that are doing

120
00:06:25,920 --> 00:06:28,957
self-learning. Um, I expect by the end

121
00:06:29,040 --> 00:06:33,277
of this year, so six months, you'll be
getting you you'll start using systems

122
00:06:33,360 --> 00:06:35,277
that are learning as you're using them.

123
00:06:35,360 --> 00:06:39,517
It's the natural progression. One
of the reasons to think that the hardware

124
00:06:39,600 --> 00:06:44,157
investments that you're seeing, which
are insanely large, make sense is that

125
00:06:44,240 --> 00:06:47,757
that those reasoning models are
infinitely more expensive than a Google

126
00:06:47,840 --> 00:06:50,397
search. Many many thousands of times

127
00:06:50,480 --> 00:06:52,477
more electricity queries and so forth

128
00:06:52,560 --> 00:06:57,677
and so on. And so you the recursive
stuff has already started.

129
00:06:57,760 --> 00:06:59,277
The general consensus there are

130
00:06:59,360 --> 00:07:03,437
different definitions of AGI is that AGI

131
00:07:03,520 --> 00:07:07,837
means it's generally intelligent. Now
generally intelligent means that you get

132
00:07:07,920 --> 00:07:11,997
up there in the morning and you have
free will. Uh I think largely invented

133
00:07:12,080 --> 00:07:16,317
here in France and you you have free
will and you basically have this notion

134
00:07:16,400 --> 00:07:20,797
of I get up, I do what I want, I learn
what I want, I seek what I want. That

135
00:07:20,880 --> 00:07:25,117
kind of strategic intelligence is not
yet present in these systems. The

136
00:07:25,200 --> 00:07:27,757
consensus is that occurs my own dates

137
00:07:27,840 --> 00:07:29,917
are between you know four to six years

138
00:07:30,000 --> 00:07:33,597
relatively soon. The step after that is

139
00:07:33,680 --> 00:07:38,157
when the system the general term there
is super intelligence. It's where the

140
00:07:38,240 --> 00:07:40,317
system is smarter than the sum of all

141
00:07:40,400 --> 00:07:43,437
humans. And the test for for super

142
00:07:43,520 --> 00:07:46,477
intelligence is pretty simple. It can

143
00:07:46,560 --> 00:07:49,357
prove something that we know to be true

144
00:07:49,440 --> 00:07:52,477
but we cannot understand the proof. We

145
00:07:52,560 --> 00:07:54,877
humans no human can understand it. Even

146
00:07:54,960 --> 00:07:56,877
all of us together cannot understand it

147
00:07:56,960 --> 00:07:59,677
but we know it's true. Henry would say

148
00:07:59,760 --> 00:08:03,837
what is that? Is that magic?

149
00:08:03,920 --> 00:08:05,757
He also would say and we said in the

150
00:08:05,840 --> 00:08:08,397
book that when people see that they may

151
00:08:08,480 --> 00:08:11,037
take up arms against it because it'll be

152
00:08:11,120 --> 00:08:13,997
so frightening to humans. A number of us

153
00:08:14,080 --> 00:08:18,637
wrote a a paper on super intelligence
where we speculated how do you contain

154
00:08:18,720 --> 00:08:23,277
super intelligence? I do a lot
of national security work and how do you

155
00:08:23,360 --> 00:08:28,077
deal with the fact that your competitor
might be ahead of you? Well, in network

156
00:08:28,160 --> 00:08:31,437
effect businesses, which is what I've
lived in for my whole life, didn't have

157
00:08:31,520 --> 00:08:35,197
a name when I started, but there
understand that all the most valuable

158
00:08:35,280 --> 00:08:37,597
companies in the world are essentially
network effect businesses because

159
00:08:37,680 --> 00:08:40,157
network effects work. What happens when

160
00:08:40,240 --> 00:08:44,637
you've got two countries where one
is ahead of the other? And I'll give you a

161
00:08:44,720 --> 00:08:47,997
simple example. This is you and me.
We have our little tech startup. We have a

162
00:08:48,080 --> 00:08:50,317
thousand people so forth. And then one

163
00:08:50,400 --> 00:08:52,957
day you and I decide that we're going to

164
00:08:53,040 --> 00:08:57,997
have AI researchers. That is computers
doing AI research. And how many do we

165
00:08:58,080 --> 00:09:02,477
have? Well, we have a million because we
we have enough money. And all they need

166
00:09:02,560 --> 00:09:06,397
is electricity. You know, they don't
eat pizza, they don't sleep, whatever. So

167
00:09:06,480 --> 00:09:08,157
now your slope of innovation goes like

168
00:09:08,240 --> 00:09:11,197
that. this in a network effect business.

169
00:09:11,280 --> 00:09:13,597
This is likely to produce slopes of

170
00:09:13,680 --> 00:09:17,277
gains at this level. The opponent may

171
00:09:17,360 --> 00:09:21,517
decide that once you get there,
they'll never catch up and then you can do

172
00:09:21,600 --> 00:09:23,517
really bad things to them. Then you have

173
00:09:23,600 --> 00:09:25,837
an issue essentially a race condition of

174
00:09:25,920 --> 00:09:28,797
preeemption. So these are just the

175
00:09:28,880 --> 00:09:31,437
beginnings of these ideas. But if you

176
00:09:31,520 --> 00:09:34,477
believe that AI is this powerful, then

177
00:09:34,560 --> 00:09:37,117
every aspect of human experience will be

178
00:09:37,200 --> 00:09:41,837
touched by it, including national
security, politicians, democracies, and

179
00:09:41,920 --> 00:09:44,077
so forth. I spend lots of time in

180
00:09:44,160 --> 00:09:46,157
conversations with people who say, will

181
00:09:46,240 --> 00:09:51,037
democracy uh survive AGI, right? I mean,

182
00:09:51,120 --> 00:09:54,397
these are really the core questions
about humanity that are being raised by

183
00:09:54,480 --> 00:09:56,637
the arrival of this intelligence.

184
00:09:56,720 --> 00:10:00,397
>> Pretty wild interviews so far. So
in the upcoming part, let's see what Dr.

185
00:10:00,480 --> 00:10:03,837
Schmidt talks about the AI bubble and
the historically important discovery for

186
00:10:03,920 --> 00:10:07,197
our mankind. >> If you ask
most of the executives in the

187
00:10:07,280 --> 00:10:09,997
industry, they will say the following.

188
00:10:10,080 --> 00:10:13,837
They'll say that we're
in a period of overbuilding.

189
00:10:13,920 --> 00:10:15,757
They'll say that there will be over

190
00:10:15,840 --> 00:10:19,437
capacity in two or three years.

191
00:10:19,520 --> 00:10:23,357
And when you ask them, they'll say,
"But I'll be fine and the other guys are

192
00:10:23,440 --> 00:10:25,437
going to lose all their money." So

193
00:10:25,520 --> 00:10:28,717
that's a classic bubble, right?

194
00:10:28,800 --> 00:10:32,317
What would mean what would it be? What
would need to be true for that not to be

195
00:10:32,400 --> 00:10:35,997
a bubble? If you believe the San Francisco

196
00:10:36,080 --> 00:10:40,237
consensus and you believe that reasoning,

197
00:10:40,320 --> 00:10:44,717
which is done through inference
and through these back and forth um

198
00:10:44,800 --> 00:10:50,077
essentially reinforcement learning
chains, if you believe that those are

199
00:10:50,160 --> 00:10:54,877
going to be the defining aspects
of humanity, then it's underhyped and we

200
00:10:54,960 --> 00:10:58,237
need even more. I know that sounds insane,

201
00:10:58,320 --> 00:11:03,117
right? But that's the basis of the
reasoning. I personally don't know. Um

202
00:11:03,200 --> 00:11:07,597
I've looked at this pretty hard because
you have these massive data centers and

203
00:11:07,680 --> 00:11:12,477
Nvidia is quite happy to sell them all
the chips, you know, and uh the real

204
00:11:12,560 --> 00:11:13,997
estate developers are used to borrowing

205
00:11:14,080 --> 00:11:16,877
money to build these things. I've never

206
00:11:16,960 --> 00:11:20,637
seen a situation where hardware capacity

207
00:11:20,720 --> 00:11:23,677
was not taken up by software. There was

208
00:11:23,760 --> 00:11:27,117
an old saying when I was young that was

209
00:11:27,200 --> 00:11:31,917
grove giveth and gates taketh away,

210
00:11:32,000 --> 00:11:36,637
right? Referring to Intel. So you the
old-timers will remember that the Intel

211
00:11:36,720 --> 00:11:40,717
chips would get better, but your
computer never got faster. And we used

212
00:11:40,800 --> 00:11:42,317
to say, why does it not get faster?

213
00:11:42,400 --> 00:11:44,397
Well, because in this case, Microsoft

214
00:11:44,480 --> 00:11:46,797
would add all these extra features. So I

215
00:11:46,880 --> 00:11:49,997
think it's it's unlikely based on my

216
00:11:50,080 --> 00:11:53,117
experience that this is a bubble. It's

217
00:11:53,200 --> 00:11:57,197
much more likely that you're seeing
an whole new industrial structure. So one

218
00:11:57,280 --> 00:11:59,837
way to think about this is the discovery

219
00:11:59,920 --> 00:12:03,197
of LLMs and foundation models which

220
00:12:03,280 --> 00:12:05,037
we're properly going to give credit to

221
00:12:05,120 --> 00:12:08,317
the team at OpenAI at the time is a

222
00:12:08,400 --> 00:12:11,757
historically important discovery. it

223
00:12:11,840 --> 00:12:15,837
began right to give them some credit
although in fairness quite a few of

224
00:12:15,920 --> 00:12:19,757
those people used to work for me uh we
won't bring that up but we'll give them

225
00:12:19,840 --> 00:12:22,797
credit that historical event

226
00:12:22,880 --> 00:12:26,797
is a very big deal overall AI is no

227
00:12:26,880 --> 00:12:29,597
longer just a tool it is a force

228
00:12:29,680 --> 00:12:32,717
reshaping business society and culture

229
00:12:32,800 --> 00:12:37,197
no one knows how this is all going
to work yet but preparation matters let's

230
00:12:37,280 --> 00:12:41,357
wrap things up with this powerful speech
from the former open AAI chief chief

231
00:12:41,440 --> 00:12:42,477
scientist.

232
00:12:42,560 --> 00:12:43,917
>> There is something a little different

233
00:12:44,000 --> 00:12:48,557
going on right now. You all leave, we

234
00:12:48,640 --> 00:12:51,997
all leave in the most unusual time ever.

235
00:12:52,080 --> 00:12:55,037
And this is something that people might
say often, but I think it's actually

236
00:12:55,120 --> 00:12:57,117
true this time. And the reason it's true

237
00:12:57,200 --> 00:13:01,357
this time is because of AI, right?

238
00:13:01,440 --> 00:13:06,157
Obviously, I mean, from what I hear,

239
00:13:06,240 --> 00:13:08,637
the AI of today has already changed what

240
00:13:08,720 --> 00:13:12,237
it means to be a student by a pretty

241
00:13:12,320 --> 00:13:14,957
considerable degree.

242
00:13:15,040 --> 00:13:16,797
That's

243
00:13:16,880 --> 00:13:20,797
I uh especially I that's what I I sense

244
00:13:20,880 --> 00:13:23,357
and I think it's true. But of course,

245
00:13:23,440 --> 00:13:27,357
the impact of AI goes beyond that.

246
00:13:27,440 --> 00:13:29,597
What happens to the kind of work we do?

247
00:13:29,680 --> 00:13:34,237
Well, it's starting to change a little bit
in some unknown and unpredictable ways.

248
00:13:34,320 --> 00:13:39,037
And some some some
work may feel it sooner,

249
00:13:39,120 --> 00:13:43,437
some work might feel it later.
With today's AI, you can go on on uh on

250
00:13:43,520 --> 00:13:47,517
Twitter and you can look at what AI
can do and what people say and you might

251
00:13:47,600 --> 00:13:50,797
feel a little bit of that. You wonder,
hey, which skills are useful? Which ones

252
00:13:50,880 --> 00:13:52,877
will be less useful? So, you got these

253
00:13:52,960 --> 00:13:55,197
questions going on. And so you can say

254
00:13:55,280 --> 00:14:01,117
that the current level of challenge is

255
00:14:01,200 --> 00:14:04,077
how will it affect

256
00:14:04,160 --> 00:14:07,277
work and our careers.

257
00:14:07,360 --> 00:14:11,517
But the thing the real challenge with AI
is that is really unprecedented and

258
00:14:11,600 --> 00:14:15,757
really extreme and it's going to be very
different in the future compared to the

259
00:14:15,840 --> 00:14:19,917
way it is today. Like you know we've all
seen AI. We've all spoken to a computer

260
00:14:20,000 --> 00:14:24,477
and a computer has spoken back to us
which is a new thing. Computers would

261
00:14:24,560 --> 00:14:28,717
not do this in the past, but now they do.
So, you speak to a computer and it

262
00:14:28,800 --> 00:14:32,397
understands you and it speaks back
to you and it also does it in voice and it

263
00:14:32,480 --> 00:14:35,197
write some code. It's it's pretty crazy,

264
00:14:35,280 --> 00:14:39,437
but there are so many things it cannot
do as well and it's so efficient. So,

265
00:14:39,520 --> 00:14:41,117
you can say it still needs to catch up

266
00:14:41,200 --> 00:14:44,957
on a lot of things, but

267
00:14:45,040 --> 00:14:47,037
it's evocative.

268
00:14:47,120 --> 00:14:48,957
It's good enough that you can ask

269
00:14:49,040 --> 00:14:54,237
yourself, you could imagine, okay, fine,
in some number of years, some people say

270
00:14:54,320 --> 00:14:58,397
it's in three, some people say it's
in five, 10, numbers are being thrown

271
00:14:58,480 --> 00:15:03,197
around. It's a bit hard
to predict the future, but [snorts]

272
00:15:03,280 --> 00:15:07,437
slowly but surely or maybe not so slowly,
AI will keep getting better. And

273
00:15:07,520 --> 00:15:12,637
the day will come when AI will do all
of our all the things that we can do. Not

274
00:15:12,720 --> 00:15:13,997
just some of them, but all of them.

275
00:15:14,080 --> 00:15:15,837
Anything which I can learn, anything

276
00:15:15,920 --> 00:15:18,317
which any any one of you can learn, the

277
00:15:18,400 --> 00:15:21,117
AI could do as well. How do we know

278
00:15:21,200 --> 00:15:23,037
this? By the way, how can I be so sure?

279
00:15:23,120 --> 00:15:25,837
How can I be so sure of that? The reason

280
00:15:25,920 --> 00:15:27,437
is

281
00:15:27,520 --> 00:15:30,637
that all of us have a brain and the

282
00:15:30,720 --> 00:15:33,757
brain is a biological computer.

283
00:15:33,840 --> 00:15:36,397
That's why we have a brain, the brain is

284
00:15:36,480 --> 00:15:41,277
a biological computer. So why can't
a digital computer, a digital brain do the

285
00:15:41,360 --> 00:15:44,957
same things? This is the one sentence
summary for why AI will be able to do

286
00:15:45,040 --> 00:15:49,357
all those things because we have a brain
and the brain is a biological computer.

287
00:15:49,440 --> 00:15:53,357
And so you can start asking yourselves
what's going to happen? What's going to

288
00:15:53,440 --> 00:15:56,557
happen when computers can do all
of our jobs? Right? Those are really big

289
00:15:56,640 --> 00:15:58,637
questions. Those are dramatic questions.

290
00:15:58,720 --> 00:16:01,357
And right now like you start thinking
about it a little bit. You go, gosh,

291
00:16:01,440 --> 00:16:02,957
that's a little intense. But it's

292
00:16:03,040 --> 00:16:05,037
actually only part of the intensity

293
00:16:05,120 --> 00:16:08,717
because what's going to happen?

294
00:16:08,800 --> 00:16:11,757
What what will we the collective V want

295
00:16:11,840 --> 00:16:16,157
to use these AIs for? Do more work, grow

296
00:16:16,240 --> 00:16:20,397
the economy, do R&D, do AI research. So

297
00:16:20,480 --> 00:16:24,717
then the rate of progress will become
really extremely fast for some time at

298
00:16:24,800 --> 00:16:27,117
least. These are such extreme things.

299
00:16:27,200 --> 00:16:30,797
These are such unimaginable things.
So right now I'm trying to pull you into

300
00:16:30,880 --> 00:16:33,357
that a little bit into this headsp space

301
00:16:33,440 --> 00:16:38,637
of this really extreme and radical
future that AI creates. But it's also

302
00:16:38,720 --> 00:16:41,517
very difficult to imagine. It's very
very difficult to imagine. It's very

303
00:16:41,600 --> 00:16:45,037
difficult to internalize and to really
believe on an emotional level. Even I

304
00:16:45,120 --> 00:16:47,917
struggle with it. And yet the logic

305
00:16:48,000 --> 00:16:51,917
seems to dictate that this very likely

306
00:16:52,000 --> 00:16:54,237
should happen.

307
00:16:54,320 --> 00:16:56,877
So

308
00:16:56,960 --> 00:17:00,077
what does one do in such a world? You

309
00:17:00,160 --> 00:17:02,717
know there is a quote which is like this

310
00:17:02,800 --> 00:17:04,637
uh

311
00:17:04,720 --> 00:17:07,917
uh which goes like this. It says you may

312
00:17:08,000 --> 00:17:09,917
not take interest in politics but

313
00:17:10,000 --> 00:17:12,717
politics will take interest in you. So

314
00:17:12,800 --> 00:17:15,437
the same applies to AI many times over.

315
00:17:15,520 --> 00:17:20,237
And in particular I think that by simply

316
00:17:20,320 --> 00:17:23,037
using AI and looking at what the best AI

317
00:17:23,120 --> 00:17:26,317
of today can do, you get an intuition.

318
00:17:26,400 --> 00:17:31,037
You get an intuition. And as AI
continues to improve in one year, in two

319
00:17:31,120 --> 00:17:33,117
years, in three years, the intuition

320
00:17:33,200 --> 00:17:34,877
will become stronger. And a lot of the

321
00:17:34,960 --> 00:17:38,237
things that we are talking about now,

322
00:17:38,320 --> 00:17:40,877
they will become much more real. They'll

323
00:17:40,960 --> 00:17:42,957
become less imaginary. In the end of the

324
00:17:43,040 --> 00:17:47,997
day, no amount of essays
and and explanations can

325
00:17:48,080 --> 00:17:49,917
can compete with what we see with our

326
00:17:50,000 --> 00:17:52,477
own senses, with our own two eyes. And

327
00:17:52,560 --> 00:17:55,357
especially with AI, the very smart,

328
00:17:55,440 --> 00:18:00,557
super intelligent AI in the future,
there will be very profound issues about

329
00:18:00,640 --> 00:18:04,717
making sure that they are they say what
they say and not pretend to be something

330
00:18:04,800 --> 00:18:07,357
else. And I'm really condensing a lot

331
00:18:07,440 --> 00:18:09,757
into a small amount of information here

332
00:18:09,840 --> 00:18:13,917
in time here. But overall by simply

333
00:18:14,000 --> 00:18:16,797
looking at what AI can do not ignoring

334
00:18:16,880 --> 00:18:19,437
it when the time comes that will

335
00:18:19,520 --> 00:18:23,917
generate the energy that's required to
overcome the huge challenge that AI will

336
00:18:24,000 --> 00:18:26,317
pose and the challenge that AI poses in

337
00:18:26,400 --> 00:18:29,197
some sense is the greatest challenge of

338
00:18:29,280 --> 00:18:33,117
humanity ever and

339
00:18:33,200 --> 00:18:35,197
overcoming it will also have the will

340
00:18:35,280 --> 00:18:37,757
also bring the greatest reward

341
00:18:37,840 --> 00:18:40,797
and in some sense whether you like it or

342
00:18:40,880 --> 00:18:45,437
not your life is going to be affected
by AI to a great extent and so looking at

343
00:18:45,520 --> 00:18:47,757
it paying attention and then generating

344
00:18:47,840 --> 00:18:50,237
the energy to solve the problems that

345
00:18:50,320 --> 00:18:54,880
will come up that's going to be the main