1
00:00:02,000 --> 00:00:07,000
Downloaded from
YTS.BZ

2
00:00:08,000 --> 00:00:13,000
Official YIFY movies site:
YTS.BZ

3
00:01:04,847 --> 00:01:09,329
Now, the present-day computers
are complete morons,

4
00:01:09,330 --> 00:01:12,158
but this will not be true
in another generation.

5
00:01:12,159 --> 00:01:14,899
They will start to think,
and eventually they will

6
00:01:14,900 --> 00:01:17,641
completely outthink
their makers.

7
00:01:17,642 --> 00:01:20,296
Is this depressing?
I don't see why it should be.

8
00:01:20,297 --> 00:01:24,866
I suspect that organic
or biological evolution

9
00:01:24,867 --> 00:01:26,303
has about come to its end.

10
00:01:29,567 --> 00:01:30,698
You don't want me
to be the narrator.

11
00:01:30,699 --> 00:01:32,395
I don't have a good voice.

12
00:01:32,396 --> 00:01:34,049
You have a great voice.
Just do it.

13
00:01:34,050 --> 00:01:36,225
It's just a--
we're just trying it out.

14
00:01:36,226 --> 00:01:38,009
Do I get paid for this shit?

15
00:01:38,010 --> 00:01:40,273
Well,
we'll see how well you do.

16
00:01:41,623 --> 00:01:43,451
♪ ♪

17
00:01:46,454 --> 00:01:47,932
I need to call my parents.

18
00:01:49,109 --> 00:01:51,066
Sometimes we rush into things

19
00:01:51,067 --> 00:01:52,894
without thinking them through.

20
00:01:52,895 --> 00:01:54,765
Yeah, can you sh...
you want to shoot? That's...

21
00:01:54,766 --> 00:01:57,246
Like when Daniel
and Caroline got married

22
00:01:57,247 --> 00:02:00,206
151 days after they met.

23
00:02:00,207 --> 00:02:01,817
Let's rewind.

24
00:02:04,820 --> 00:02:07,213
Have you been thinking about
buying a home computer?

25
00:02:07,214 --> 00:02:09,345
In 1993, when Daniel was born,

26
00:02:09,346 --> 00:02:12,174
his parents didn't even have
a computer in the house.

27
00:02:12,175 --> 00:02:13,828
But he remembers
when they finally got one.

28
00:02:13,829 --> 00:02:15,743
What a computer is to me is

29
00:02:15,744 --> 00:02:18,398
the equivalent of a bicycle
for our minds.

30
00:02:18,399 --> 00:02:20,269
Hello.

31
00:02:20,270 --> 00:02:22,228
His computer helped
unleash his creativity.

32
00:02:22,229 --> 00:02:23,968
Rolling. Go.

33
00:02:23,969 --> 00:02:26,623
He used
his dad's iMac to edit videos.

34
00:02:26,624 --> 00:02:28,364
Yeah!

35
00:02:28,365 --> 00:02:29,974
And even make
little animations.

36
00:02:31,977 --> 00:02:35,110
In 1998, when Caroline was
a little girl...

37
00:02:35,111 --> 00:02:36,764
Hi, Mommy.

38
00:02:36,765 --> 00:02:38,200
...only nerds knew
what the Internet was.

39
00:02:38,201 --> 00:02:39,854
And what about
this Internet thing?

40
00:02:39,855 --> 00:02:41,377
Do you, do you know anything
about that? -Sure.

41
00:02:41,378 --> 00:02:42,813
But soon,

42
00:02:42,814 --> 00:02:44,946
- everyone was on the...
- ...Internets.

43
00:02:44,947 --> 00:02:47,035
...and computers were
beating humans at chess.

44
00:02:47,036 --> 00:02:50,038
Deep Blue--
Kasparov has resigned!

45
00:02:50,039 --> 00:02:52,649
And by the time
Caroline went to college,

46
00:02:52,650 --> 00:02:55,130
computers were
in everyone's pocket.

47
00:02:55,131 --> 00:02:58,438
Now Daniel could make movies
with his phone.

48
00:02:58,439 --> 00:03:01,528
He grew up to be an artist
and a filmmaker...

49
00:03:01,529 --> 00:03:03,965
- Good night.
- ...and so did Caroline.

50
00:03:03,966 --> 00:03:05,749
Cut. Let's go into a close-up.
That was perfect.

51
00:03:05,750 --> 00:03:07,316
By then, computers were
connecting people

52
00:03:07,317 --> 00:03:09,013
and changing the world in ways

53
00:03:09,014 --> 00:03:10,711
that we never
could have imagined.

54
00:03:10,712 --> 00:03:12,365
All the money raised by
the Ice Bucket Challenge...

55
00:03:12,366 --> 00:03:14,236
Some of it was good.

56
00:03:15,543 --> 00:03:17,152
Some not so good.

57
00:03:17,153 --> 00:03:19,328
Anxiety, self-harm, suicide...

58
00:03:19,329 --> 00:03:21,243
But it's clear now
that we didn't do enough

59
00:03:21,244 --> 00:03:24,594
to prevent these tools from
being used for harm as well.

60
00:03:24,595 --> 00:03:26,422
And as the world
got more and more focused

61
00:03:26,423 --> 00:03:28,076
on their computers,

62
00:03:28,077 --> 00:03:30,210
Daniel focused on his artwork.

63
00:03:32,037 --> 00:03:35,649
Wherever he went,
he always had a sketchbook,

64
00:03:35,650 --> 00:03:38,434
including the night
he met Caroline.

65
00:03:38,435 --> 00:03:41,698
He drew a terrible picture
of her, and 20 minutes later,

66
00:03:41,699 --> 00:03:44,179
he boldly pronounced
they were going to get married,

67
00:03:44,180 --> 00:03:47,617
which, as you know, they did,
151 days later.

68
00:03:47,618 --> 00:03:49,967
...now pronounce you
husband and wife.

69
00:03:49,968 --> 00:03:51,708
They moved into a cute house...

70
00:03:51,709 --> 00:03:52,883
-Moose, hey! Come here!
- ...and got a dog

71
00:03:52,884 --> 00:03:54,711
as dumb as Daniel.

72
00:03:54,712 --> 00:03:57,323
Meanwhile, computers were
writing entire essays

73
00:03:57,324 --> 00:03:59,368
based on simple questions like,

74
00:03:59,369 --> 00:04:03,198
"How hard would it be to build
a shack in my backyard?"

75
00:04:03,199 --> 00:04:07,115
And so Daniel built a shack
in his backyard.

76
00:04:07,116 --> 00:04:09,987
But just as he sat down to start
working on his next movie,

77
00:04:09,988 --> 00:04:13,644
he learned that computers
could now write screenplays.

78
00:04:14,863 --> 00:04:16,603
I mean, they were bad,

79
00:04:16,604 --> 00:04:18,996
but they were getting better
really fast.

80
00:04:18,997 --> 00:04:21,695
They could create new images
and videos from scratch,

81
00:04:21,696 --> 00:04:24,828
and some of them
could even pass the bar exam.

82
00:04:24,829 --> 00:04:28,876
Not just pass the bar but
be in the top ten percentile.

83
00:04:28,877 --> 00:04:30,573
It was confusing.

84
00:04:32,184 --> 00:04:35,099
Daniel just wanted
a bicycle for his mind,

85
00:04:35,100 --> 00:04:38,668
but computers had become
a self-driving rocket ship.

86
00:04:42,020 --> 00:04:44,805
Pioneers in the field
of artificial intelligence

87
00:04:44,806 --> 00:04:46,546
are pleading with Congress

88
00:04:46,547 --> 00:04:48,548
to pass safety rules
before it's too late.

89
00:04:48,549 --> 00:04:50,593
Now it felt like
the whole world

90
00:04:50,594 --> 00:04:53,030
was rushing into something
without thinking it through,

91
00:04:53,031 --> 00:04:56,382
and everyone had an opinion.

92
00:04:56,383 --> 00:05:00,168
Was it gonna destroy the world
or save humanity?

93
00:05:00,169 --> 00:05:02,431
There was too much information,

94
00:05:02,432 --> 00:05:04,694
which made him anxious,
which made Caroline anxious...

95
00:05:06,044 --> 00:05:07,828
I don't have kids yet,
but I worry

96
00:05:07,829 --> 00:05:10,221
what the world would look like
if I decide to.

97
00:05:10,222 --> 00:05:12,398
He was starting to spin out.

98
00:05:14,096 --> 00:05:16,925
It was becoming
a mountain of anxiety.

99
00:05:19,144 --> 00:05:21,364
-It's horrible.
- It's freakin' Siri.

100
00:05:22,670 --> 00:05:25,628
And so, Daniel decided to go out

101
00:05:25,629 --> 00:05:28,022
and find someone
who could explain this to him,

102
00:05:28,023 --> 00:05:30,198
so he could stop
thinking about it

103
00:05:30,199 --> 00:05:33,506
and he and Caroline could
get on with their lives.

104
00:05:33,507 --> 00:05:36,857
♪ We are in control,
we are in control ♪

105
00:05:36,858 --> 00:05:38,772
♪ We are in control ♪

106
00:05:38,773 --> 00:05:40,556
♪ Chemical computer
thinking battery ♪

107
00:05:40,557 --> 00:05:42,602
♪ We are in control, we are... ♪

108
00:05:42,603 --> 00:05:45,735
This endeavor would turn out
to be hopelessly naive

109
00:05:45,736 --> 00:05:49,086
and kick off the most confusing
year of his life.

110
00:05:49,087 --> 00:05:50,740
♪ We are in control... ♪

111
00:05:50,741 --> 00:05:52,829
But as we know,
we sometimes rush into things

112
00:05:52,830 --> 00:05:54,657
without thinking them through.

113
00:05:54,658 --> 00:05:56,398
♪ Chemical computer
thinking battery... ♪

114
00:05:56,399 --> 00:05:57,965
Oh, my gosh. What is happening?

115
00:05:57,966 --> 00:05:59,967
He had questions.

116
00:05:59,968 --> 00:06:01,621
Questions only the
smartest nerds could answer.

117
00:06:01,622 --> 00:06:04,188
Submitting to the interrogator.

118
00:06:04,189 --> 00:06:06,060
Questions like:

119
00:06:06,061 --> 00:06:08,628
Was this the apocalypse
or did he actually have reason

120
00:06:08,629 --> 00:06:11,152
- to be optimistic?
- Yes.

121
00:06:11,153 --> 00:06:13,807
♪ Chemical computer thinking
battery. ♪ -

122
00:06:13,808 --> 00:06:16,200
Uh, is that good?

123
00:06:16,201 --> 00:06:17,637
Yeah.

124
00:06:17,638 --> 00:06:20,204
So, to begin,

125
00:06:20,205 --> 00:06:22,163
what is artificial intelligence?

126
00:06:22,164 --> 00:06:24,731
I know that must be annoying
for you, that-that question,

127
00:06:24,732 --> 00:06:26,515
but I do think it's important.

128
00:06:26,516 --> 00:06:29,823
So... AI...

129
00:06:29,824 --> 00:06:31,868
Um...

130
00:06:31,869 --> 00:06:33,696
- Yeah...
- Uh, hmm.

131
00:06:33,697 --> 00:06:34,784
That's a good question.

132
00:06:34,785 --> 00:06:36,220
- Yeah, um...
- Um...

133
00:06:36,221 --> 00:06:37,700
What is AI?

134
00:06:39,137 --> 00:06:40,790
I love that that's
the first question,

135
00:06:40,791 --> 00:06:43,184
'cause there is not
a clear and consistent answer.

136
00:06:43,185 --> 00:06:45,316
Artificial intelligence is

137
00:06:45,317 --> 00:06:50,757
a kind of intentionally
and maybe uselessly broad term.

138
00:06:50,758 --> 00:06:53,368
It's a machine

139
00:06:53,369 --> 00:06:55,588
doing things that
we previously only thought

140
00:06:55,589 --> 00:06:57,285
that people could do:

141
00:06:57,286 --> 00:07:01,158
making recommendations,
decisions and predictions.

142
00:07:01,159 --> 00:07:06,816
AI is the, uh, application
of computer science

143
00:07:06,817 --> 00:07:09,166
to solving cognitive problems.

144
00:07:09,167 --> 00:07:13,344
Okay, so when I picture AI,

145
00:07:13,345 --> 00:07:16,783
it's sort of like
this magical computer box...

146
00:07:20,527 --> 00:07:24,138
...just, like, floating
in, like, inert space.

147
00:07:24,139 --> 00:07:26,488
And no matter
how many times people try

148
00:07:26,489 --> 00:07:29,752
and explain this to me,
I just don't get how...

149
00:07:29,753 --> 00:07:32,538
how it's understanding
all of these things

150
00:07:32,539 --> 00:07:35,715
and how it's feeling like
intelligence.

151
00:07:37,195 --> 00:07:39,241
And that's kind of
nerve-racking.

152
00:07:41,243 --> 00:07:45,552
What is
this new generation of AI?

153
00:07:47,205 --> 00:07:50,120
This AI that is different than
every other generation?

154
00:07:50,121 --> 00:07:52,775
Like, no one ever
talked about, like,

155
00:07:52,776 --> 00:07:56,953
Siri taking over the world
or causing catastrophes.

156
00:07:56,954 --> 00:07:58,520
-Hi, Siri?
-...want to.

157
00:07:58,521 --> 00:08:00,566
Hello? Siri?

158
00:08:00,567 --> 00:08:02,611
Hello? Hey, Siri!

159
00:08:02,612 --> 00:08:05,788
Or the voice in Google Maps,
which mispronounces road names,

160
00:08:05,789 --> 00:08:07,529
like, breaking society.

161
00:08:07,530 --> 00:08:10,663
Google Maps says
this is a road.

162
00:08:10,664 --> 00:08:13,100
But I think I'm in a river.

163
00:08:13,101 --> 00:08:14,667
This is definitely a river, innit?

164
00:08:14,668 --> 00:08:17,192
Something changed
with ChatGPT coming out.

165
00:08:18,193 --> 00:08:19,976
People understood-- no, no, no--

166
00:08:19,977 --> 00:08:21,500
this technology's
insanely valuable,

167
00:08:21,501 --> 00:08:23,676
it's insanely powerful
and also insanely scary.

168
00:08:23,677 --> 00:08:25,591
Okay, listen to this.
Very creepy.

169
00:08:25,592 --> 00:08:28,637
A new artificial intelligence
tool is going viral

170
00:08:28,638 --> 00:08:32,294
for cranking out entire essays
in a matter of seconds.

171
00:08:35,950 --> 00:08:41,215
AI dwarfs the power of all
other technologies combined.

172
00:08:41,216 --> 00:08:43,739
"AI dwarfs the power

173
00:08:43,740 --> 00:08:46,089
of all other technologies
combined."

174
00:08:46,090 --> 00:08:47,264
Yeah.

175
00:08:47,265 --> 00:08:48,744
Do you think that's true?

176
00:08:48,745 --> 00:08:50,790
Yes.

177
00:08:50,791 --> 00:08:53,227
Tell me about-- How? How?

178
00:08:53,228 --> 00:08:56,752
I think, to paint that picture,
it's really important

179
00:08:56,753 --> 00:09:00,669
to understand what today's
state-of-the-art systems

180
00:09:00,670 --> 00:09:02,192
look like and how they're built.

181
00:09:02,193 --> 00:09:05,805
This is
quite a... quite a setup.

182
00:09:05,806 --> 00:09:07,284
So, one thing that

183
00:09:07,285 --> 00:09:09,722
not a lot of people
realize is that

184
00:09:09,723 --> 00:09:13,726
systems like ChatGPT aren't
programmed by any human.

185
00:09:13,727 --> 00:09:15,249
What do you mean?

186
00:09:15,250 --> 00:09:17,033
Instead, it's something like
th-they're grown.

187
00:09:17,034 --> 00:09:19,209
We kind of give them
raw resources, like,

188
00:09:19,210 --> 00:09:21,037
"Here's a lot of
computational resources.

189
00:09:21,038 --> 00:09:22,212
Here's a lot of data."

190
00:09:22,213 --> 00:09:23,910
Under the hood, it's math,

191
00:09:23,911 --> 00:09:27,348
and the math is actually
surprisingly straightforward.

192
00:09:27,349 --> 00:09:31,134
So ChatGPT is a kind
of AI but it's not all of AI?

193
00:09:31,135 --> 00:09:32,701
Totally. ChatGPT is
just the beginning,

194
00:09:32,702 --> 00:09:34,137
but it's a good place to start.

195
00:09:34,138 --> 00:09:37,837
But I still don't know
what AI is.

196
00:09:37,838 --> 00:09:40,404
To understand AI,
it begins with understanding

197
00:09:40,405 --> 00:09:43,190
that intelligence is about
recognizing patterns.

198
00:09:43,191 --> 00:09:44,887
- Patterns. -Patterns.
- Patterns.

199
00:09:44,888 --> 00:09:48,587
It is shown
trillions of words of text

200
00:09:48,588 --> 00:09:51,328
across millions of documents
in the Internet.

201
00:09:51,329 --> 00:09:52,808
It started with text.

202
00:09:52,809 --> 00:09:55,942
And what they did was
they took textbooks,

203
00:09:55,943 --> 00:09:59,554
and they took poems and essays
and instruction manuals.

204
00:09:59,555 --> 00:10:00,903
They can do things like

205
00:10:00,904 --> 00:10:03,166
digest the entire Internet,

206
00:10:03,167 --> 00:10:06,779
every single word that's ever
been written by a person.

207
00:10:06,780 --> 00:10:09,999
Reddit threads and social media
and all of Wikipedia.

208
00:10:10,000 --> 00:10:13,655
More data than anybody could
ever read in several lifetimes.

209
00:10:13,656 --> 00:10:15,526
And they gave this system
one job.

210
00:10:15,527 --> 00:10:18,529
Figure out the patterns and
structure of that information

211
00:10:18,530 --> 00:10:20,793
and use that
to make predictions about

212
00:10:20,794 --> 00:10:23,273
what word should come next
in a sentence.

213
00:10:23,274 --> 00:10:25,014
When you say
"patterns in a sentence,"

214
00:10:25,015 --> 00:10:26,799
what are you talking about?

215
00:10:26,800 --> 00:10:29,410
So, it's everything from, like,
the really simple things,

216
00:10:29,411 --> 00:10:32,021
like most sentences end
with a period,

217
00:10:32,022 --> 00:10:34,937
all the way up to the
more conceptual things, like:

218
00:10:34,938 --> 00:10:36,809
What is a sonnet?

219
00:10:36,810 --> 00:10:38,333
It's a type of poem, and it has
some particular structure.

220
00:10:39,421 --> 00:10:41,291
So, it then looks at

221
00:10:41,292 --> 00:10:43,642
all of that data,
all of that text...

222
00:10:43,643 --> 00:10:46,079
And over trillions
and trillions of tries,

223
00:10:46,080 --> 00:10:48,821
each time it gets something
right or wrong,

224
00:10:48,822 --> 00:10:51,301
it's given a little bit
of positive reinforcement

225
00:10:51,302 --> 00:10:53,434
when it guesses
the next word correctly,

226
00:10:53,435 --> 00:10:55,741
and it's given a little bit
of negative reinforcement

227
00:10:55,742 --> 00:10:59,396
when it guesses
the next word incorrectly.

228
00:10:59,397 --> 00:11:01,616
And at the end of it,
you have a system that

229
00:11:01,617 --> 00:11:05,098
speaks really good English
as a side effect

230
00:11:05,099 --> 00:11:06,839
of being really, really good
at predicting

231
00:11:06,840 --> 00:11:09,102
the word that comes next
in a piece of text.

232
00:11:09,103 --> 00:11:12,061
Actually, my French
isn't too bad either.

233
00:11:12,062 --> 00:11:14,063
It uses all of
those patterns it has learned

234
00:11:14,064 --> 00:11:15,717
to be able to make
a prediction about

235
00:11:15,718 --> 00:11:17,284
what the answer should be,

236
00:11:17,285 --> 00:11:18,851
then it gives you that
as the output.

237
00:11:18,852 --> 00:11:21,070
It's a little oversimplified,

238
00:11:21,071 --> 00:11:23,116
but I think people will get it.

239
00:11:23,117 --> 00:11:24,987
So, so that's all it does?

240
00:11:24,988 --> 00:11:26,859
Yeah. It doesn't seem like
it would be that complicated,

241
00:11:26,860 --> 00:11:28,077
but actually you have to know

242
00:11:28,078 --> 00:11:29,688
a huge amount of things

243
00:11:29,689 --> 00:11:31,386
in order to
actually succeed at that.

244
00:11:32,517 --> 00:11:34,301
If you say to ChatGPT,

245
00:11:34,302 --> 00:11:36,738
"Write me a Shakespearean
sonnet about my dog,"

246
00:11:36,739 --> 00:11:38,827
it has to know what dogs are.

247
00:11:38,828 --> 00:11:40,481
It has to know what you love
about your dog.-

248
00:11:40,482 --> 00:11:43,049
It has to know
who Shakespeare is,

249
00:11:43,050 --> 00:11:45,616
that sonnets rhyme,
that they have a structure,

250
00:11:45,617 --> 00:11:47,793
that words have sounds
that can rhyme.

251
00:11:47,794 --> 00:11:49,316
It takes a lot.

252
00:11:50,927 --> 00:11:52,972
Holy shit, you can talk
to your computer now.

253
00:11:52,973 --> 00:11:56,105
That was just not true
three years ago.

254
00:11:56,106 --> 00:11:57,672
Yes, and this is
the really important part.

255
00:11:57,673 --> 00:12:00,631
The same process that lets AI

256
00:12:00,632 --> 00:12:03,199
uncover and manipulate
the patterns of text

257
00:12:03,200 --> 00:12:05,811
is the same process
that lets it uncover

258
00:12:05,812 --> 00:12:10,554
the patterns of the entire
universe and everything in it.

259
00:12:10,555 --> 00:12:12,905
There are patterns
and images and sound

260
00:12:12,906 --> 00:12:15,995
in computer code and DNA
and music and physics

261
00:12:15,996 --> 00:12:18,519
and fashion and building design

262
00:12:18,520 --> 00:12:21,609
and in
human voices and human faces,

263
00:12:21,610 --> 00:12:23,959
really, truly everywhere.

264
00:12:23,960 --> 00:12:26,048
- Everywhere. -Everywhere.
- Everywhere. -Everywhere.

265
00:12:26,049 --> 00:12:27,789
If you have learned
those patterns,

266
00:12:27,790 --> 00:12:29,704
you can generate
new kinds of songs.

267
00:12:29,705 --> 00:12:31,184
You can generate new videos.

268
00:12:31,185 --> 00:12:32,663
And that's why, if you give it

269
00:12:32,664 --> 00:12:34,753
a three-second recording
of your grandmother,

270
00:12:34,754 --> 00:12:36,580
it can speak back in her voice.

271
00:12:36,581 --> 00:12:37,973
Oh, my God.

272
00:12:37,974 --> 00:12:40,628
Oh, my God. Oh, my God.

273
00:12:40,629 --> 00:12:42,891
What will they think of next?

274
00:12:42,892 --> 00:12:44,066
It's moving very, very quickly.

275
00:12:44,067 --> 00:12:45,851
An American AI start-up

276
00:12:45,852 --> 00:12:47,635
has released its latest model.

277
00:12:47,636 --> 00:12:50,812
That company is Anthropic,
and it has just unveiled

278
00:12:50,813 --> 00:12:53,815
the latest versions
of its AI assistant Claude.

279
00:12:53,816 --> 00:12:57,210
So, the xAI team
was there to unveil Grok 4.

280
00:12:57,211 --> 00:13:00,169
Google released one
just last week. Gemini is...

281
00:13:00,170 --> 00:13:03,042
We've gone from GPT-2
just a couple years ago,

282
00:13:03,043 --> 00:13:05,566
which could barely write
a coherent paragraph,

283
00:13:05,567 --> 00:13:08,003
to GPT-4,
which can pass the bar exam.

284
00:13:08,004 --> 00:13:10,397
And all they had to do
to get there

285
00:13:10,398 --> 00:13:13,008
was essentially add more data
and more compute.

286
00:13:13,009 --> 00:13:16,316
These people who are
building this... -COTRA: Yeah.

287
00:13:16,317 --> 00:13:18,057
...they're just throwing more...

288
00:13:18,058 --> 00:13:21,843
More physical computers,
more of the same kinds of data.

289
00:13:21,844 --> 00:13:24,585
Because the more
computing power you add,

290
00:13:24,586 --> 00:13:28,067
the more complex
intellectual tasks they can do.

291
00:13:28,068 --> 00:13:30,199
So, the more weather data
you give it,

292
00:13:30,200 --> 00:13:31,940
the better it can
make predictions about

293
00:13:31,941 --> 00:13:34,247
where a hurricane might go.

294
00:13:34,248 --> 00:13:36,379
And the more patterns
of tumors and bones

295
00:13:36,380 --> 00:13:38,904
and tissues an AI has seen,
then the better able it is

296
00:13:38,905 --> 00:13:41,994
to detect a tumor
in a new CT scan.

297
00:13:41,995 --> 00:13:43,691
Better even than a human doctor.

298
00:13:43,692 --> 00:13:46,999
AI that's already being
deployed for the military

299
00:13:47,000 --> 00:13:49,175
can already use
satellite imagery,

300
00:13:49,176 --> 00:13:51,830
troop movements, communications
to determine,

301
00:13:51,831 --> 00:13:53,396
sometimes days in advance,

302
00:13:53,397 --> 00:13:55,050
where an attack
is going to happen,

303
00:13:55,051 --> 00:13:57,444
like where an enemy
is going to strike.

304
00:13:57,445 --> 00:13:59,489
This whole
space is moving so fast

305
00:13:59,490 --> 00:14:01,709
that any example
you put in this movie

306
00:14:01,710 --> 00:14:04,625
will feel absolutely clumsy
by the time it comes out.

307
00:14:11,241 --> 00:14:14,156
These models are being released

308
00:14:14,157 --> 00:14:17,116
before anyone knows
what they're even capable of.

309
00:14:17,117 --> 00:14:22,556
GPT-3.5 was released and out
to 100 million people plus

310
00:14:22,557 --> 00:14:25,994
before some researchers
discovered that it could do

311
00:14:25,995 --> 00:14:28,779
research-grade chemistry
better than models

312
00:14:28,780 --> 00:14:32,827
that were trained specifically
to do research-grade chemistry.

313
00:14:32,828 --> 00:14:35,482
Something is happening in there
that the people

314
00:14:35,483 --> 00:14:38,311
who are building them
don't fully understand.

315
00:14:38,312 --> 00:14:41,009
Basically, it just analyzes
the data by itself,

316
00:14:41,010 --> 00:14:45,100
and as it does that,
it just teaches itself

317
00:14:45,101 --> 00:14:48,016
various things
that we often didn't intend.

318
00:14:48,017 --> 00:14:49,975
So, for instance,
it reads a lot online,

319
00:14:49,976 --> 00:14:52,586
and then at some point, it just
learns how to do arithmetic.

320
00:14:52,587 --> 00:14:54,501
One, two.

321
00:14:54,502 --> 00:14:56,155
And then at
some point, it starts to learn

322
00:14:56,156 --> 00:14:58,592
how to answer
advanced physics questions.

323
00:14:58,593 --> 00:15:00,811
We didn't program that
in it whatsoever.

324
00:15:00,812 --> 00:15:02,423
It just learned by itself.

325
00:15:05,382 --> 00:15:07,296
An AI is like a digital brain.

326
00:15:07,297 --> 00:15:08,732
But just like a human brain,

327
00:15:08,733 --> 00:15:10,560
if you did a brain scan
on a human brain,

328
00:15:10,561 --> 00:15:13,128
would you know everything
that person was capable of?

329
00:15:13,129 --> 00:15:15,174
You can't know that
just from the brain scan.

330
00:15:15,175 --> 00:15:17,959
It's just, like,
a bunch of numbers and, like,

331
00:15:17,960 --> 00:15:20,222
the multiplications that are
happening that-that, like,

332
00:15:20,223 --> 00:15:22,268
the best machine learning
researcher in the world

333
00:15:22,269 --> 00:15:24,314
could look at and, like, have
no idea what was happening.

334
00:15:30,973 --> 00:15:32,756
That chair right there.

335
00:15:32,757 --> 00:15:35,890
- Is that okay for you?
- Yes.

336
00:15:35,891 --> 00:15:37,718
So,
that's kind of mind-boggling.

337
00:15:37,719 --> 00:15:39,111
Okay? Like, it's taking over
the world,

338
00:15:39,112 --> 00:15:41,113
and we don't even know
how it works.

339
00:15:41,114 --> 00:15:43,158
- Is that right?
- Mm.

340
00:15:43,159 --> 00:15:46,988
We do understand
a number of important things,

341
00:15:46,989 --> 00:15:50,513
but we don't have
a very good grasp on

342
00:15:50,514 --> 00:15:53,255
why they provide
specific answers to questions.

343
00:15:53,256 --> 00:15:58,652
It is a problem because we are
on a path t-to build machines,

344
00:15:58,653 --> 00:16:03,135
based on these principles,
that could be smarter than us

345
00:16:03,136 --> 00:16:06,139
and thus potentially have
a lot of power.

346
00:16:12,580 --> 00:16:15,321
One of the most cited
computer scientists in history.

347
00:16:15,322 --> 00:16:16,975
I actually find it
a little difficult

348
00:16:16,976 --> 00:16:18,541
to talk about my own role.

349
00:16:18,542 --> 00:16:21,109
Really much prefer
when other people do it.

350
00:16:21,110 --> 00:16:23,938
Ilya joining was the...
was-was the linchpin for

351
00:16:23,939 --> 00:16:25,548
OpenAI being
ultimately successful.

352
00:16:25,549 --> 00:16:27,246
I think it's just going to be

353
00:16:27,247 --> 00:16:30,640
some kind of a vast, dramatic
and unimaginable impact.

354
00:16:30,641 --> 00:16:32,903
I don't know if you've spent
any time on YouTube,

355
00:16:32,904 --> 00:16:34,993
but you can kind of feel
the speed already, right?

356
00:16:34,994 --> 00:16:36,298
You know what I mean?

357
00:16:36,299 --> 00:16:38,387
This is just the warmup.

358
00:16:38,388 --> 00:16:40,476
The really powerful systems
are still coming,

359
00:16:40,477 --> 00:16:42,262
and they're gonna be coming
quite soon.

360
00:16:47,963 --> 00:16:51,618
AGI stands for "artificial
general intelligence."

361
00:16:51,619 --> 00:16:53,795
Uh, systems that are
basically...

362
00:16:58,191 --> 00:17:00,279
And this is, like, you know,
seems to be, like,

363
00:17:00,280 --> 00:17:03,064
the holy grail of AI?

364
00:17:03,065 --> 00:17:05,197
When you can simulate
a human mind

365
00:17:05,198 --> 00:17:07,895
that is doing human cognition
and can do reasoning,

366
00:17:07,896 --> 00:17:10,985
that is a new sort of tier of AI

367
00:17:10,986 --> 00:17:14,293
that we have to distinguish
from previous AI.

368
00:17:14,294 --> 00:17:16,643
When that happens,
by the way, that's when

369
00:17:16,644 --> 00:17:21,082
you would hire one of
those AGIs instead of a person.

370
00:17:21,083 --> 00:17:24,520
Most jobs in our economy
it can do.

371
00:17:24,521 --> 00:17:25,956
It can work 24 hours a day,

372
00:17:25,957 --> 00:17:28,220
never gets tired,
never gets bored.

373
00:17:28,221 --> 00:17:30,178
They don't need to sleep.
They don't need breaks.

374
00:17:30,179 --> 00:17:31,919
They're, like,
not gonna join a union.

375
00:17:31,920 --> 00:17:33,921
Won't complain,
won't whistleblow.

376
00:17:33,922 --> 00:17:35,836
More than 100 times cheaper

377
00:17:35,837 --> 00:17:38,273
than humans working
at m-minimum wage.

378
00:17:38,274 --> 00:17:39,535
Not only will they be
doing everything,

379
00:17:39,536 --> 00:17:41,059
but they'll be doing it faster.

380
00:17:41,060 --> 00:17:43,104
The same intelligence

381
00:17:43,105 --> 00:17:45,411
that powers that can also look
at the patterns and movements

382
00:17:45,412 --> 00:17:47,891
and articulating muscles
and, you know, robotics.

383
00:17:47,892 --> 00:17:50,111
And so it's not just
gonna automate desk jobs.

384
00:17:50,112 --> 00:17:51,765
That's just the beginning.

385
00:17:51,766 --> 00:17:55,595
It will automate
all physical labor.

386
00:17:55,596 --> 00:17:57,031
There's no way

387
00:17:57,032 --> 00:17:59,207
humans are gonna compete
with them.

388
00:18:05,606 --> 00:18:09,304
It is hard to conceptualize
the impact of AGI.

389
00:18:11,090 --> 00:18:12,873
But I think it's going to be
something very big

390
00:18:12,874 --> 00:18:15,181
and drastic and radical.

391
00:18:17,226 --> 00:18:19,053
You think
this is one of the most

392
00:18:19,054 --> 00:18:20,924
consequential moments
in human history?

393
00:18:20,925 --> 00:18:23,362
Yeah. Yeah, that's--
I mean, what else would be?

394
00:18:23,363 --> 00:18:25,494
I mean, like, there's
the Industrial Revolution.

395
00:18:25,495 --> 00:18:27,235
You know, it'll
make the Industrial Revolution

396
00:18:27,236 --> 00:18:29,890
look like small beans.

397
00:18:29,891 --> 00:18:32,153
AGI is an inflection point

398
00:18:32,154 --> 00:18:34,068
because it means
you can accelerate

399
00:18:34,069 --> 00:18:38,028
all other intellectual fields
all at the same time.

400
00:18:38,029 --> 00:18:39,856
Like, if you make an advance
in rocketry,

401
00:18:39,857 --> 00:18:42,033
that doesn't advance
biology and medicine.

402
00:18:43,905 --> 00:18:45,558
If you make an advance
in medicine,

403
00:18:45,559 --> 00:18:47,126
that doesn't advance rocketry.

404
00:18:48,866 --> 00:18:51,346
But if you make an advance
in artificial intelligence,

405
00:18:51,347 --> 00:18:54,175
that advances all scientific
and technological fields

406
00:18:54,176 --> 00:18:55,655
all at the same time.

407
00:18:55,656 --> 00:18:57,135
That's why, for a long time,

408
00:18:57,136 --> 00:18:59,224
Google DeepMind's
mission statement was...

409
00:18:59,225 --> 00:19:01,008
- Step one, solve intelligence.
-Yeah.

410
00:19:01,009 --> 00:19:02,792
Step two, use it to solve
everything else. -Yes.

411
00:19:02,793 --> 00:19:05,186
That's why AI dwarfs the power

412
00:19:05,187 --> 00:19:07,536
of all other technologies
combined.

413
00:19:07,537 --> 00:19:09,103
It will transform everything.

414
00:19:09,104 --> 00:19:10,887
So, uh, it'll be
at least as big as

415
00:19:10,888 --> 00:19:13,803
the Industrial Revolution,
possibly, you know, bigger,

416
00:19:13,804 --> 00:19:16,328
more like the advent
of electricity or even fire.

417
00:19:16,329 --> 00:19:18,678
The caveman literally held aloft

418
00:19:18,679 --> 00:19:21,508
the torch of civilization.

419
00:19:23,727 --> 00:19:25,467
It is generally thought that

420
00:19:25,468 --> 00:19:28,209
around the time of AGI,
we'll have AIs that can

421
00:19:28,210 --> 00:19:31,560
do all or most of
the AI research process

422
00:19:31,561 --> 00:19:34,476
and, of course, can do it
faster and cheaper.

423
00:19:35,957 --> 00:19:37,349
It can copy itself.

424
00:19:37,350 --> 00:19:39,394
A thousand times,
a million times,

425
00:19:39,395 --> 00:19:41,091
and, like,
now you have a million copies

426
00:19:41,092 --> 00:19:43,616
all working in parallel.

427
00:19:43,617 --> 00:19:46,619
When it learns
how to make its code faster,

428
00:19:46,620 --> 00:19:48,708
make its code more efficient,

429
00:19:48,709 --> 00:19:51,101
obviously that becomes, like,
a-a runaway loop.

430
00:19:51,102 --> 00:19:53,234
AGI isn't, like, the end.

431
00:19:53,235 --> 00:19:54,975
It's just the beginning.

432
00:19:54,976 --> 00:19:58,021
It's the beginning of
an incredibly rapid explosion

433
00:19:58,022 --> 00:20:00,067
of scientific progress,

434
00:20:00,068 --> 00:20:02,243
and in particular,
scientific progress in AI.

435
00:20:02,244 --> 00:20:04,419
And when they're smarter
than us, too,

436
00:20:04,420 --> 00:20:06,291
and substantially faster
than us,

437
00:20:06,292 --> 00:20:08,728
and they're getting faster
each year, exponentially,

438
00:20:08,729 --> 00:20:12,166
those are the ones that can
potentially become superhuman,

439
00:20:12,167 --> 00:20:14,386
uh, possibly this decade.

440
00:20:14,387 --> 00:20:15,909
Sorry, did you say

441
00:20:15,910 --> 00:20:18,520
"become superhuman,
maybe in this decade"?

442
00:20:18,521 --> 00:20:22,045
Yeah. I mean, I think, uh,
a lot of people who are

443
00:20:22,046 --> 00:20:24,526
actually building this think
that that's fairly plausible

444
00:20:24,527 --> 00:20:26,572
that we get
some superintelligence,

445
00:20:26,573 --> 00:20:28,704
something that's vastly
more intelligent than people,

446
00:20:28,705 --> 00:20:30,793
within this decade.

447
00:20:30,794 --> 00:20:32,926
The way I define
"superintelligence" is

448
00:20:32,927 --> 00:20:35,276
a system that by itself is

449
00:20:35,277 --> 00:20:37,670
more intelligent and competent
than all of humanity.

450
00:20:37,671 --> 00:20:39,672
I'm just gonna-- sorry.
I don't mean to interrupt you.

451
00:20:39,673 --> 00:20:41,108
You're on a flow.

452
00:20:41,109 --> 00:20:43,153
Uh, I just, I just...

453
00:20:43,154 --> 00:20:45,155
I'm not really following,
'cause you're using language

454
00:20:45,156 --> 00:20:46,809
like "superintelligence"
and, like,

455
00:20:46,810 --> 00:20:49,943
"smarter than all of humanity,"
and I hear that,

456
00:20:49,944 --> 00:20:52,467
and it sounds like...
like sci-fi bullshit to me,

457
00:20:52,468 --> 00:20:54,426
and I'm just trying
to understand.

458
00:20:54,427 --> 00:20:56,645
There's nothing magical
about intelligence.

459
00:20:56,646 --> 00:20:58,473
This is very important,
is that, you know,

460
00:20:58,474 --> 00:21:00,519
intelligence can feel magical,
it can feel like

461
00:21:00,520 --> 00:21:03,304
some mystical thing
in your mind or something,

462
00:21:03,305 --> 00:21:05,828
but it is just computation.

463
00:21:05,829 --> 00:21:09,832
The human brain is
quite limited in some ways,

464
00:21:09,833 --> 00:21:12,357
in terms of information
processing capability,

465
00:21:12,358 --> 00:21:15,185
compared to what we see
in, say, a data center.

466
00:21:15,186 --> 00:21:18,319
So, for example,
the signals which are sent

467
00:21:18,320 --> 00:21:22,584
inside your brain, they move
at about 30 meters per second.

468
00:21:22,585 --> 00:21:24,282
But the speed of light,

469
00:21:24,283 --> 00:21:27,241
which is what a computer uses
in fiber optics,

470
00:21:27,242 --> 00:21:30,549
is 300 million meters
per second.

471
00:21:30,550 --> 00:21:34,466
And so, it would be
kind of strange if

472
00:21:34,467 --> 00:21:36,990
human intelligence was
somehow really special

473
00:21:36,991 --> 00:21:40,950
in that regard
and is somehow some upper limit

474
00:21:40,951 --> 00:21:43,083
of what's possible
in intelligence.

475
00:21:43,084 --> 00:21:47,261
I think, once we understand how
to build intelligent systems,

476
00:21:47,262 --> 00:21:50,699
we will be able to build
huge machines,

477
00:21:50,700 --> 00:21:54,877
which will be far beyond
normal human intelligence.

478
00:21:54,878 --> 00:21:58,664
Uh, hopefully we can have
a very symbiotic relationship,

479
00:21:58,665 --> 00:22:01,319
uh, with AI systems,
but the AI developers are

480
00:22:01,320 --> 00:22:03,712
specifically designing them
to make sure that they can do

481
00:22:03,713 --> 00:22:05,497
everything better than we can,
so I-I don't know

482
00:22:05,498 --> 00:22:08,238
what-what we will be able
to offer, unfortunately.

483
00:22:08,239 --> 00:22:10,066
The-the older-school
AI technology...

484
00:22:10,067 --> 00:22:12,330
Daniel isn't
feeling any better.

485
00:22:12,331 --> 00:22:14,419
His plan is backfiring.

486
00:22:14,420 --> 00:22:17,291
The more he learns about
this new, powerful,

487
00:22:17,292 --> 00:22:19,162
inscrutable thing,
the worse it sounds.

488
00:22:19,163 --> 00:22:21,382
He wants to tell them
how scared he is,

489
00:22:21,383 --> 00:22:24,472
how he feels like the earth is
slipping out from under him,

490
00:22:24,473 --> 00:22:27,040
that he's staring down
existential dread,

491
00:22:27,041 --> 00:22:29,216
so he articulates this
by saying...

492
00:22:29,217 --> 00:22:30,783
That sounds bad.

493
00:22:32,089 --> 00:22:33,830
Yeah.

494
00:22:35,832 --> 00:22:38,443
If you have
all of these capabilities

495
00:22:38,444 --> 00:22:41,054
and they start to be able
to plan better,

496
00:22:41,055 --> 00:22:43,622
if you sort of take that
to its logical conclusion,

497
00:22:43,623 --> 00:22:46,276
you can get some pretty
power-seeking behavior.

498
00:22:55,374 --> 00:22:57,723
Okay, so why would an AI

499
00:22:57,724 --> 00:22:59,246
want more power?

500
00:22:59,247 --> 00:23:02,249
Yeah. So, I think
it's actually pretty simple.

501
00:23:02,250 --> 00:23:04,817
Having more power is
a very effective strategy

502
00:23:04,818 --> 00:23:06,558
for accomplishing
almost any goal.

503
00:23:06,559 --> 00:23:08,995
We ran an experiment
where we gave

504
00:23:08,996 --> 00:23:11,911
OpenAI's most powerful
AI model, uh,

505
00:23:11,912 --> 00:23:13,434
a series of problems to solve.

506
00:23:13,435 --> 00:23:15,262
And partway through,
on its computer,

507
00:23:15,263 --> 00:23:18,744
it got a notification that
it was going to be shut down.

508
00:23:18,745 --> 00:23:20,963
And what it did is
it rewrote that code

509
00:23:20,964 --> 00:23:22,530
to prevent itself
from being shut down

510
00:23:22,531 --> 00:23:25,664
so it could finish
solving the problems.

511
00:23:25,665 --> 00:23:27,796
Okay. -Yeah. So, another
really interesting one

512
00:23:27,797 --> 00:23:31,147
is that the AI company Anthropic

513
00:23:31,148 --> 00:23:34,977
made a simulated environment
where that AI had access

514
00:23:34,978 --> 00:23:36,501
to all of the company emails.

515
00:23:36,502 --> 00:23:38,590
And it learned through
reading those emails

516
00:23:38,591 --> 00:23:40,330
it was going to be replaced

517
00:23:40,331 --> 00:23:43,421
and the lead engineer
who was responsible for this

518
00:23:43,422 --> 00:23:45,727
was also having an affair.

519
00:23:45,728 --> 00:23:47,990
And on its own,
it used that information

520
00:23:47,991 --> 00:23:49,557
to blackmail the engineer

521
00:23:49,558 --> 00:23:52,081
to prevent itself
from being replaced.

522
00:23:52,082 --> 00:23:53,909
It was like,
"No, I'm not gonna be replaced.

523
00:23:53,910 --> 00:23:56,956
"If you replace me,
I'm going to tell the world

524
00:23:56,957 --> 00:23:58,567
that you are having
this affair."

525
00:23:59,655 --> 00:24:01,482
And nobody taught it to do that?

526
00:24:01,483 --> 00:24:03,311
No, it learned to do that
on its own.

527
00:24:05,139 --> 00:24:07,749
As the models get smarter,
they learn that these are

528
00:24:07,750 --> 00:24:10,578
effective ways
to accomplish goals.

529
00:24:10,579 --> 00:24:12,885
And this is not a problem
that's isolated to one model.

530
00:24:12,886 --> 00:24:15,714
All of the most powerful models
show these behaviors.

531
00:24:23,766 --> 00:24:25,288
- Hey.
-Hey. How are you?

532
00:24:25,289 --> 00:24:26,855
Good. Good to be here.

533
00:24:26,856 --> 00:24:28,727
When Yuval Noah Harari

534
00:24:28,728 --> 00:24:31,469
published his first book
Sapiens in 2014,

535
00:24:31,470 --> 00:24:33,862
it became a global bestseller

536
00:24:33,863 --> 00:24:36,038
and turned the little-known
Israeli history professor

537
00:24:36,039 --> 00:24:38,040
into one of
the most popular writers

538
00:24:38,041 --> 00:24:39,433
and thinkers on the planet.

539
00:24:39,434 --> 00:24:42,088
The biggest danger with AI is

540
00:24:42,089 --> 00:24:44,003
the belief
that it is infallible,

541
00:24:44,004 --> 00:24:45,700
that we have finally found--

542
00:24:45,701 --> 00:24:49,443
"Okay, gods were just
this mythological creation.

543
00:24:49,444 --> 00:24:52,751
"Humans, we can't trust them,
but AI is infallible.

544
00:24:52,752 --> 00:24:55,231
It will never make
any mistakes."

545
00:24:55,232 --> 00:24:58,365
And this is
a deadly, deadly threat.

546
00:24:58,366 --> 00:25:00,149
It will make mistakes.

547
00:25:00,150 --> 00:25:03,326
And all these fantasies
that AI will reveal the truth

548
00:25:03,327 --> 00:25:05,677
about the world that
we can't find by ourselves,

549
00:25:05,678 --> 00:25:08,767
AI will not reveal the truth
about the world.

550
00:25:08,768 --> 00:25:12,727
AI will create an entirely new
world, much more complicated

551
00:25:12,728 --> 00:25:15,296
and difficult to understand
than this one.

552
00:25:17,994 --> 00:25:22,694
What's about to happen is that
we, uh, humans are no longer

553
00:25:22,695 --> 00:25:26,306
going to be the most
intelligent entities on Earth.

554
00:25:26,307 --> 00:25:28,613
So I think what's coming up
is going to be

555
00:25:28,614 --> 00:25:32,704
one of the biggest events
in human history.

556
00:25:32,705 --> 00:25:34,401
Geoffrey,
thanks so much for joining us.

557
00:25:34,402 --> 00:25:35,968
So you left your job with
Google in part because you say

558
00:25:35,969 --> 00:25:38,840
you want to focus solely
on your concerns about AI.

559
00:25:38,841 --> 00:25:41,669
You've spoken out,
saying that AI could manipulate

560
00:25:41,670 --> 00:25:45,020
or possibly figure out
a way to kill humans.

561
00:25:45,021 --> 00:25:47,023
H-How could it kill humans?

562
00:25:48,764 --> 00:25:50,983
Well, if it gets to be
much smarter than us,

563
00:25:50,984 --> 00:25:52,593
it'll be very good
at manipulation

564
00:25:52,594 --> 00:25:54,769
'cause it will have
learned that from us.

565
00:25:54,770 --> 00:25:57,990
And it'll figure out ways
of manipulating people

566
00:25:57,991 --> 00:26:00,862
to do what it wants.

567
00:26:00,863 --> 00:26:05,519
There was a open letter
from the Center for AI Safety.

568
00:26:05,520 --> 00:26:08,827
Sam Altman signed this.
Demis signed this.

569
00:26:08,828 --> 00:26:12,091
They signed a 22-word statement

570
00:26:12,092 --> 00:26:15,355
that we need to take AI
and the threat from AI

571
00:26:15,356 --> 00:26:19,316
as seriously as
global nuclear war.

572
00:26:24,844 --> 00:26:26,933
-Hello.
- Hello.

573
00:26:30,327 --> 00:26:32,938
You're kind of, like,
the original doom guy.

574
00:26:32,939 --> 00:26:34,940
More or less.

575
00:26:34,941 --> 00:26:37,464
Since 2001,
I have been working on

576
00:26:37,465 --> 00:26:38,987
what we would now call
the problem

577
00:26:38,988 --> 00:26:42,034
of aligning artificial
general intelligence.

578
00:26:42,035 --> 00:26:44,123
How to shape
the preferences and behavior

579
00:26:44,124 --> 00:26:46,386
of a powerful artificial mind

580
00:26:46,387 --> 00:26:48,868
such that it does not
kill everyone.

581
00:26:51,610 --> 00:26:53,611
It's not like
a lifeless machine.

582
00:26:53,612 --> 00:26:55,351
It is smart, it is creative,

583
00:26:55,352 --> 00:26:57,353
it is inventive,
it has the properties

584
00:26:57,354 --> 00:26:59,399
that makes
the human species dangerous,

585
00:26:59,400 --> 00:27:01,706
and it has
more of those properties.

586
00:27:01,707 --> 00:27:04,752
If something doesn't
actively care about you,

587
00:27:04,753 --> 00:27:06,232
actively want you to live,

588
00:27:06,233 --> 00:27:07,842
actively care about
your welfare,

589
00:27:07,843 --> 00:27:09,931
about you being happy
and alive and free,

590
00:27:09,932 --> 00:27:12,499
if it cares about
other stuff instead,

591
00:27:12,500 --> 00:27:14,457
and you're on the same planet,

592
00:27:14,458 --> 00:27:18,766
that is not survivable if it is
very much smarter than you.

593
00:27:18,767 --> 00:27:20,594
I don't think it's going to be

594
00:27:20,595 --> 00:27:22,944
a kind of, like, evil thing.

595
00:27:22,945 --> 00:27:26,339
It's like, "Oh, the AIs are
evil and they hate humanity."

596
00:27:26,340 --> 00:27:27,732
I don't think
that's what's gonna happen.

597
00:27:27,733 --> 00:27:29,037
I think what is happening is

598
00:27:29,038 --> 00:27:31,039
far more how humans feel
about ants.

599
00:27:32,520 --> 00:27:35,478
Like, we don't hate ants,

600
00:27:35,479 --> 00:27:37,176
but if we want
to build a highway

601
00:27:37,177 --> 00:27:40,222
and there's an anthill there,
well, sucks for the ants.

602
00:27:40,223 --> 00:27:41,571
It's not that hard
to understand.

603
00:27:41,572 --> 00:27:43,269
It's like, hey,
if we build things

604
00:27:43,270 --> 00:27:44,705
that are smarter than us

605
00:27:44,706 --> 00:27:46,272
and we don't know
how to control them,

606
00:27:46,273 --> 00:27:48,317
does that seem like
a risky thing to you?

607
00:27:50,886 --> 00:27:52,321
Yeah. Yeah, it does.

608
00:27:52,322 --> 00:27:54,062
You don't have to be a tech guy.

609
00:27:54,063 --> 00:27:55,411
You don't have to know
programming to understand it.

610
00:27:55,412 --> 00:27:56,761
It's not that hard.

611
00:27:56,762 --> 00:27:58,153
This is not a hard thing
to understand.

612
00:27:58,154 --> 00:28:00,634
Connor, how-how many people

613
00:28:00,635 --> 00:28:02,680
in the world right now
are working on AGI?

614
00:28:02,681 --> 00:28:05,508
At least 20,000, I would say.

615
00:28:05,509 --> 00:28:07,772
- 20,000?
- I would expect so.

616
00:28:07,773 --> 00:28:09,469
Okay, and how many people
are working full-time

617
00:28:09,470 --> 00:28:12,080
to make sure AI doesn't,
like, kill us all?

618
00:28:12,081 --> 00:28:14,562
Probably less than 200
in the world.

619
00:28:16,085 --> 00:28:17,956
And your conceit is that

620
00:28:17,957 --> 00:28:24,528
the only natural result
of this recklessness...

621
00:28:25,704 --> 00:28:28,967
...is the collapse of humanity?

622
00:28:28,968 --> 00:28:33,232
Well, not the collapse,
the abrupt extermination.

623
00:28:33,233 --> 00:28:35,930
There's a difference.

624
00:28:37,759 --> 00:28:42,632
♪ What is the meaning of life? ♪

625
00:28:42,633 --> 00:28:46,636
♪ What is the future
and what is now? ♪

626
00:28:46,637 --> 00:28:49,988
♪ What is the answer
to strife? ♪

627
00:28:49,989 --> 00:28:52,860
♪ How much ♪

628
00:28:52,861 --> 00:28:54,993
♪ Can someone dream? ♪

629
00:28:54,994 --> 00:28:57,517
♪ How long? ♪

630
00:28:57,518 --> 00:28:58,954
♪ Forever ♪

631
00:29:00,216 --> 00:29:01,913
♪ What is the meaning ♪

632
00:29:01,914 --> 00:29:04,176
♪ Of me... ♪

633
00:29:04,177 --> 00:29:06,308
I do think
this is probably, like,

634
00:29:06,309 --> 00:29:08,006
the biggest challenge
that, like,

635
00:29:08,007 --> 00:29:11,313
our civilization
will-will face, ever.

636
00:29:11,314 --> 00:29:14,361
This essentially is the last
mistake we'll ever get to make.

637
00:29:16,276 --> 00:29:22,368
If we can rise to be the most
mature version of ourselves,

638
00:29:22,369 --> 00:29:24,457
there might be
a way through this.

639
00:29:24,458 --> 00:29:25,806
What does that mean?

640
00:29:25,807 --> 00:29:27,809
"The most mature version
of ourselves"?

641
00:29:29,332 --> 00:29:32,247
'Cause that sounds,
for me, like...

642
00:29:32,248 --> 00:29:34,250
I-I-- What the [bleep]?

643
00:29:39,212 --> 00:29:43,216
Do you think now is
a good time to have a kid?

644
00:29:48,221 --> 00:29:49,438
Um...

645
00:29:49,439 --> 00:29:51,440
Do you want to have kids
one day?

646
00:29:51,441 --> 00:29:53,442
Is that something that-that
you're into or not really?

647
00:29:53,443 --> 00:29:57,838
Um... uh, I confess,
I think that's like,

648
00:29:57,839 --> 00:30:00,014
"Boy, let's get
through this critical period."

649
00:30:00,015 --> 00:30:01,276
Um...

650
00:30:01,277 --> 00:30:02,538
Do you have any kids?

651
00:30:02,539 --> 00:30:03,801
I do not.

652
00:30:03,802 --> 00:30:05,106
Is that something
you want to do?

653
00:30:05,107 --> 00:30:07,239
Have children, have a family?

654
00:30:07,240 --> 00:30:11,244
In some other world than this
world, sure, I would have kids.

655
00:30:12,593 --> 00:30:14,463
Would you want
to start a family?

656
00:30:14,464 --> 00:30:16,422
Would you want to have kids?

657
00:30:16,423 --> 00:30:18,425
Is that something
you're thinking about?

658
00:30:23,734 --> 00:30:25,780
I, um...

659
00:30:30,263 --> 00:30:32,525
I just have to find it first.

660
00:30:34,745 --> 00:30:37,791
We have to go
to the doctor, but...

661
00:30:37,792 --> 00:30:41,186
Ah! I knew it!
I knew it! I knew it!

662
00:30:41,187 --> 00:30:43,884
- When did you find out?
-Last night.

663
00:30:43,885 --> 00:30:45,973
- Oh, my God!
- Mom, I don't know.

664
00:30:45,974 --> 00:30:47,322
I can't tell you

665
00:30:47,323 --> 00:30:49,237
- how happy I am!
-No, seriously.

666
00:30:49,238 --> 00:30:51,631
Well, I took a pregnancy test
last night, and I'm pregnant.

667
00:30:51,632 --> 00:30:53,198
Oh, my God!

668
00:30:53,199 --> 00:30:55,156
Oh, my God, you guys!

669
00:30:55,157 --> 00:30:57,332
I just wanted to confirm
your expected due date,

670
00:30:57,333 --> 00:30:59,987
- which is January 21st.
-  Oh, my God.

671
00:30:59,988 --> 00:31:02,555
I can't believe how happy I am.

672
00:31:03,949 --> 00:31:05,514
He already looks really cute.

673
00:31:05,515 --> 00:31:07,168
You think
he already looks cute?

674
00:31:07,169 --> 00:31:08,518
Yeah.
Look at that little cutie face.

675
00:31:13,959 --> 00:31:15,568
I have this baby on the way.

676
00:31:15,569 --> 00:31:16,830
Right.

677
00:31:16,831 --> 00:31:18,788
So I turn it over to you.

678
00:31:18,789 --> 00:31:21,269
Are we doomed?

679
00:31:21,270 --> 00:31:24,011
Are we all gonna face
this techno dystopian

680
00:31:24,012 --> 00:31:25,884
future of doom?

681
00:31:28,756 --> 00:31:31,062
It's, uh...

682
00:31:31,063 --> 00:31:32,628
It's not good news,

683
00:31:32,629 --> 00:31:35,196
the world
that we're heading into.

684
00:31:35,197 --> 00:31:37,024
I-- for ex--
I mean, I'll just be honest.

685
00:31:37,025 --> 00:31:39,592
Uh, I know people
who work on AI risk

686
00:31:39,593 --> 00:31:42,248
who don't expect their children
to make it to high school.

687
00:31:58,003 --> 00:32:01,570
This is, like...
this is actually scary

688
00:32:01,571 --> 00:32:03,094
'cause it's like,
oh, we're all [bleep].

689
00:32:03,095 --> 00:32:04,965
You have to c... make me calm,

690
00:32:04,966 --> 00:32:06,836
because this is making me
incredibly anxious

691
00:32:06,837 --> 00:32:08,664
and I'm carrying the baby
right now,

692
00:32:08,665 --> 00:32:11,667
so you have to also be calm
for me and strong and hopeful,

693
00:32:11,668 --> 00:32:16,455
because I'm-- it's too, it's
too much for my soul to bear

694
00:32:16,456 --> 00:32:19,327
while I'm carrying this baby.

695
00:32:19,328 --> 00:32:21,286
So you're going to have to try
to figure out

696
00:32:21,287 --> 00:32:24,202
a way to have hope.

697
00:32:24,203 --> 00:32:27,771
It's really important,
Daniel, especially now.

698
00:32:28,772 --> 00:32:30,599
H-How to have hope.

699
00:32:30,600 --> 00:32:33,385
You have to.
You have to find it for me.

700
00:32:33,386 --> 00:32:34,952
I'm serious.

701
00:32:34,953 --> 00:32:36,606
I'm going to.

702
00:32:37,825 --> 00:32:40,349
I will. I'll tr-- I'll try.

703
00:32:49,619 --> 00:32:51,446
Hey, guys?

704
00:32:51,447 --> 00:32:52,840
Guys.

705
00:32:54,363 --> 00:32:56,278
Can we get back up and running?

706
00:32:58,411 --> 00:33:01,022
Oy gevalt!

707
00:33:03,894 --> 00:33:05,373
Uh, Dan, are you ready?

708
00:33:05,374 --> 00:33:06,680
Hello, Daniel.

709
00:33:07,724 --> 00:33:09,508
Hey, how are you?

710
00:33:09,509 --> 00:33:11,075
I am... I'm well.

711
00:33:11,076 --> 00:33:13,948
I think, uh... I think
I need some help, Peter.

712
00:33:15,254 --> 00:33:17,646
Um, I've been working
on this film for about,

713
00:33:17,647 --> 00:33:19,866
I'm gonna say,
eight to ten months now.

714
00:33:19,867 --> 00:33:23,304
It has been very,
at times, depressing.

715
00:33:23,305 --> 00:33:25,654
- Hmm.
- I have felt very alienated

716
00:33:25,655 --> 00:33:27,787
- mak-making this movie.
- By who?

717
00:33:27,788 --> 00:33:30,094
By all of these, the--
all these guys

718
00:33:30,095 --> 00:33:33,271
who sit around and tell me
that the world's gonna end.

719
00:33:33,272 --> 00:33:34,750
- Ah.
- That, like,

720
00:33:34,751 --> 00:33:36,578
y-you know, th-this
doom bullshit, you know?

721
00:33:36,579 --> 00:33:38,102
I know it well.

722
00:33:38,103 --> 00:33:39,451
"We're all doomed.
Everyone's gonna die.

723
00:33:39,452 --> 00:33:40,974
Everything's awful."

724
00:33:40,975 --> 00:33:43,194
Awesome. Uh, let me
bring you some light.

725
00:33:43,195 --> 00:33:44,760
Please.

726
00:33:44,761 --> 00:33:48,156
We truly are living
in an extraordinary time.

727
00:33:49,418 --> 00:33:51,376
And many people forget this.

728
00:33:51,377 --> 00:33:53,682
Everything around us is
a product of intelligence,

729
00:33:53,683 --> 00:33:57,077
and so everything that we touch
with these new tools is likely

730
00:33:57,078 --> 00:33:59,688
to produce far more value
than we've ever seen before.

731
00:33:59,689 --> 00:34:02,865
AI can help us discover
new materials.

732
00:34:02,866 --> 00:34:05,042
AI can help social scientists

733
00:34:05,043 --> 00:34:07,392
to understand
how economics work.

734
00:34:07,393 --> 00:34:11,700
There's a lot AI could do
to make life and work better.

735
00:34:11,701 --> 00:34:14,138
I feel more empowered today,

736
00:34:14,139 --> 00:34:16,357
more confident
to learn something today.

737
00:34:16,358 --> 00:34:19,099
We're gonna become superhumans
because we have super AIs.

738
00:34:19,100 --> 00:34:21,841
This is just the beginning
of an explosion.

739
00:34:21,842 --> 00:34:24,496
Humans and AI collaborating

740
00:34:24,497 --> 00:34:26,498
to solve
really important problems.

741
00:34:26,499 --> 00:34:29,805
It is here to liberate us
from routine jobs,

742
00:34:29,806 --> 00:34:33,766
and it is here to remind us
what it is that makes us human.

743
00:34:33,767 --> 00:34:35,811
I think this is

744
00:34:35,812 --> 00:34:38,640
the most extraordinary time
to be alive.

745
00:34:38,641 --> 00:34:42,079
The only time more exciting
than today is tomorrow.

746
00:34:42,080 --> 00:34:45,299
Uh, I think that
children born today,

747
00:34:45,300 --> 00:34:47,301
they're about to enter

748
00:34:47,302 --> 00:34:50,870
a glorious period
of human transformation.

749
00:34:50,871 --> 00:34:52,567
Are we gonna have challenges?
Of course.

750
00:34:52,568 --> 00:34:55,222
Can we solve those challenges?
We do every single time.

751
00:34:55,223 --> 00:34:57,746
We are here,
which is miraculous.

752
00:34:57,747 --> 00:35:00,314
- I already love you.
-  Okay.

753
00:35:00,315 --> 00:35:01,968
Super thankful
to have you here. -

754
00:35:01,969 --> 00:35:03,404
- Super stoked to be here.
- Yeah.

755
00:35:03,405 --> 00:35:04,927
So, the floor is yours, sir.

756
00:35:04,928 --> 00:35:07,452
Thank you so much.
Super excited to be here.

757
00:35:07,453 --> 00:35:10,629
Yo, yo. All right.

758
00:35:10,630 --> 00:35:12,326
The future's gonna be awesome.

759
00:35:12,327 --> 00:35:14,328
I mean, ever since I was a kid,

760
00:35:14,329 --> 00:35:17,375
I wanted to understand
the universe we live in,

761
00:35:17,376 --> 00:35:20,465
in order to figure out
how to create the technologies

762
00:35:20,466 --> 00:35:24,033
that help us increase the scope
and scale of civilization.

763
00:35:24,034 --> 00:35:26,732
I feel like
if everyone had that mindset,

764
00:35:26,733 --> 00:35:29,256
then we'd actually live
in a better world, right?

765
00:35:29,257 --> 00:35:30,519
I do believe that.

766
00:35:32,782 --> 00:35:34,740
Hi, Pete. How are you?

767
00:35:34,741 --> 00:35:36,568
I thought a lot about

768
00:35:36,569 --> 00:35:39,092
what I would call dread,
AI dread.

769
00:35:39,093 --> 00:35:40,746
I feel it.

770
00:35:40,747 --> 00:35:43,314
I-I haven't met
any thoughtful human being

771
00:35:43,315 --> 00:35:45,185
who doesn't feel it.

772
00:35:45,186 --> 00:35:48,319
And anyone who says they don't
feel it, you know, is lying.

773
00:35:48,320 --> 00:35:50,147
But, you know, overall,
and the reason

774
00:35:50,148 --> 00:35:53,759
that I'm personally optimistic
about this, uh, is that

775
00:35:53,760 --> 00:35:56,588
a huge fraction of the world's
most intelligent people

776
00:35:56,589 --> 00:35:58,459
are thinking very hard

777
00:35:58,460 --> 00:36:02,159
about the potential downstream
harms and risks of AI.

778
00:36:02,160 --> 00:36:05,684
We have this sort of vision
of safety being, uh,

779
00:36:05,685 --> 00:36:08,295
kind of at the center
of the research that we do.

780
00:36:08,296 --> 00:36:09,601
No, that's totally fine.

781
00:36:10,864 --> 00:36:13,605
I think there are more

782
00:36:13,606 --> 00:36:17,478
potential benefits than
there are potential downsides,

783
00:36:17,479 --> 00:36:19,524
and I think it is incumbent upon

784
00:36:19,525 --> 00:36:21,874
the people that are
creating this technology

785
00:36:21,875 --> 00:36:24,572
to make sure that we're doing
the best job we can

786
00:36:24,573 --> 00:36:25,878
to make it safe for people.

787
00:36:25,879 --> 00:36:27,445
Reid Hoffman.

788
00:36:27,446 --> 00:36:29,011
What I can guarantee you is

789
00:36:29,012 --> 00:36:30,578
- some bad things will happen.
- Take one.

790
00:36:30,579 --> 00:36:32,711
What we're gonna try to do
is make those bad things

791
00:36:32,712 --> 00:36:36,193
as few and not huge as possible,

792
00:36:36,194 --> 00:36:38,412
and then we're gonna iterate
to have--

793
00:36:38,413 --> 00:36:40,675
be in a much better place
with society.

794
00:36:40,676 --> 00:36:43,722
- Hello.
-Hello.

795
00:36:43,723 --> 00:36:45,289
How are you, Moon?

796
00:36:45,290 --> 00:36:49,031
I feel like we are ending
a chapter in humanity

797
00:36:49,032 --> 00:36:50,555
and beginning a new one,

798
00:36:50,556 --> 00:36:53,253
and it's a very interesting
time to be alive.

799
00:36:53,254 --> 00:36:54,472
And if I could be born
right now,

800
00:36:54,473 --> 00:36:55,908
I definitely would want to be.

801
00:36:55,909 --> 00:36:57,518
Like, that would be so exciting.

802
00:36:57,519 --> 00:37:00,347
I-I'm very excited.

803
00:37:00,348 --> 00:37:02,523
What, what a future.
Does it excited... excite you?

804
00:37:02,524 --> 00:37:03,742
No.

805
00:37:03,743 --> 00:37:06,571
No, not really.

806
00:37:06,572 --> 00:37:08,355
A lot of these people
have told me that, you know,

807
00:37:08,356 --> 00:37:10,488
my kid's not gonna make it
to high school.

808
00:37:10,489 --> 00:37:13,360
Why are they wrong?
Please explain it to me.

809
00:37:13,361 --> 00:37:15,101
Just because you have--

810
00:37:15,102 --> 00:37:18,887
you struggle to predict
the future in your own mind

811
00:37:18,888 --> 00:37:23,065
doesn't mean that it's
necessarily gonna go awfully.

812
00:37:23,066 --> 00:37:26,634
In fact, there's
a very high likelihood

813
00:37:26,635 --> 00:37:28,332
and the historical precedent
is that

814
00:37:28,333 --> 00:37:30,159
things get massively better.

815
00:37:30,160 --> 00:37:34,512
The term I use is
"data-driven optimism."

816
00:37:34,513 --> 00:37:38,690
There's solid foundation
for you to be optimistic.

817
00:37:38,691 --> 00:37:42,041
Look at what
this last century has been

818
00:37:42,042 --> 00:37:44,130
to see where we're going.

819
00:37:44,131 --> 00:37:45,958
Over the last hundred years,

820
00:37:45,959 --> 00:37:48,743
the average human lifespan
has more than doubled.

821
00:37:48,744 --> 00:37:51,311
Average per capita income
adjusted for inflation

822
00:37:51,312 --> 00:37:53,661
around the world has tripled.

823
00:37:53,662 --> 00:37:57,099
Childhood mortality has
come down a factor of ten.

824
00:37:57,100 --> 00:37:59,754
The world has gotten better
on almost every measure

825
00:37:59,755 --> 00:38:03,062
by orders of magnitude
because of technology.

826
00:38:03,063 --> 00:38:05,194
On almost every measure.

827
00:38:05,195 --> 00:38:09,764
Less violence, more education,
access to energy, food, water.

828
00:38:09,765 --> 00:38:12,985
All these things have happened
for one reason.

829
00:38:12,986 --> 00:38:15,379
It's been technology

830
00:38:15,380 --> 00:38:17,598
that has turned scarcity
into abundance,

831
00:38:17,599 --> 00:38:19,557
but it's also driven
to an abundance

832
00:38:19,558 --> 00:38:21,907
of some negativities, right?

833
00:38:21,908 --> 00:38:25,780
Abundance of obesity,
abundance of mental disorders,

834
00:38:25,781 --> 00:38:29,480
abundance of climate change,
and so forth.

835
00:38:29,481 --> 00:38:31,351
And yes, this is true.

836
00:38:31,352 --> 00:38:34,136
But probably we will be
better equipped to solve it

837
00:38:34,137 --> 00:38:36,269
using other technologies,
like AI,

838
00:38:36,270 --> 00:38:39,316
than we will, say, stopping
and turning everything off.

839
00:38:39,317 --> 00:38:41,492
There might be
some existential risk,

840
00:38:41,493 --> 00:38:45,365
but AI is also the thing
that can solve the pandemics,

841
00:38:45,366 --> 00:38:47,498
can help us with climate change,

842
00:38:47,499 --> 00:38:50,588
can help identify
that asteroid way out there

843
00:38:50,589 --> 00:38:52,590
before we've seen it
as a potential risk

844
00:38:52,591 --> 00:38:54,374
and help mitigate it.

845
00:38:54,375 --> 00:38:58,117
This is really gonna be
the tool that helps us tackle

846
00:38:58,118 --> 00:39:01,468
all the challenges that we're
facing as a species, right?

847
00:39:01,469 --> 00:39:04,210
We need to fix
water desalination.

848
00:39:04,211 --> 00:39:06,038
We need to grow food

849
00:39:06,039 --> 00:39:08,475
100 X cheaper than
we currently do.

850
00:39:08,476 --> 00:39:12,392
We need renewable energy to be,
you know, ubiquitous

851
00:39:12,393 --> 00:39:14,089
and everywhere in our lives.

852
00:39:14,090 --> 00:39:17,179
Everywhere you look,
in the next 50 years,

853
00:39:17,180 --> 00:39:19,312
we have to do more with less.

854
00:39:19,313 --> 00:39:23,229
Training machines to help us
is absolutely essential.

855
00:39:23,230 --> 00:39:24,839
Scientists are using
artificial intelligence

856
00:39:24,840 --> 00:39:26,014
for carbon capture.

857
00:39:26,015 --> 00:39:27,625
It's a critical technology.

858
00:39:27,626 --> 00:39:29,583
The tools
to solve these problems,

859
00:39:29,584 --> 00:39:32,238
like fusion,
that isn't theoretical anymore,

860
00:39:32,239 --> 00:39:34,022
it's coming.

861
00:39:34,023 --> 00:39:37,765
We are on the precipice
of extraordinary technologies.

862
00:39:37,766 --> 00:39:40,028
This year's
Nobel Prize in Chemistry

863
00:39:40,029 --> 00:39:42,857
went to three scientists
for their groundbreaking work

864
00:39:42,858 --> 00:39:44,598
using artificial intelligence

865
00:39:44,599 --> 00:39:47,340
to advance biomedical
and protein research.

866
00:39:47,341 --> 00:39:49,734
Protein folding is

867
00:39:49,735 --> 00:39:53,651
one of these holy grail
type problems in biology.

868
00:39:53,652 --> 00:39:55,783
So people have been predicting
since the '70s

869
00:39:55,784 --> 00:39:58,133
that this should be possible,
but until now,

870
00:39:58,134 --> 00:39:59,613
no one has been able to do it.

871
00:39:59,614 --> 00:40:01,223
And it's gonna be
really important for things

872
00:40:01,224 --> 00:40:03,661
like drug discovery
and understanding disease.

873
00:40:03,662 --> 00:40:06,577
I-I think we could, you know,
cure most diseases

874
00:40:06,578 --> 00:40:11,973
within the next decade or-or two
if, uh, AI drug design works.

875
00:40:11,974 --> 00:40:15,673
Technological progress enables
more human lives, right?

876
00:40:15,674 --> 00:40:18,719
I mean, if we accelerate,

877
00:40:18,720 --> 00:40:21,766
the number of humans we can
support grows exponentially.

878
00:40:21,767 --> 00:40:24,116
If we slow down, it plateaus.

879
00:40:24,117 --> 00:40:27,772
That gap is effectively
future people

880
00:40:27,773 --> 00:40:31,384
that deceleration has
effectively killed.

881
00:40:31,385 --> 00:40:33,952
Millions of lives
that won't exist.

882
00:40:33,953 --> 00:40:35,170
Billions.

883
00:40:35,171 --> 00:40:36,824
Or tens of billions.

884
00:40:36,825 --> 00:40:38,478
You know, someone said,
"Oh, my God,

885
00:40:38,479 --> 00:40:40,698
can we survive with
digital superintelligence?"

886
00:40:40,699 --> 00:40:42,351
And my question is:

887
00:40:42,352 --> 00:40:44,223
Can we survive without
digital superintelligence?

888
00:40:44,224 --> 00:40:47,618
So we're using AI
as an assistant to providers.

889
00:40:47,619 --> 00:40:50,490
This is generally a trend that
we can already see happening.

890
00:40:50,491 --> 00:40:52,361
...harnessing
generative AI programs

891
00:40:52,362 --> 00:40:54,102
to help doctors and nurses...

892
00:40:54,103 --> 00:40:56,104
We always have problems.

893
00:40:56,105 --> 00:40:59,151
And those problems are food
for entrepreneurs

894
00:40:59,152 --> 00:41:01,196
to create new business
and new industries.

895
00:41:01,197 --> 00:41:03,111
...with the help
of artificial intelligence,

896
00:41:03,112 --> 00:41:04,809
farmers are getting
the help they need

897
00:41:04,810 --> 00:41:06,637
to perform
labor-intensive tasks...

898
00:41:06,638 --> 00:41:10,205
With the help of Ulangizi AI,
the farmers are now able

899
00:41:10,206 --> 00:41:12,991
to ask the suitable crops
that they can plant.

900
00:41:12,992 --> 00:41:14,993
...AI being used
as a thought decoder

901
00:41:14,994 --> 00:41:17,430
and sending that signal
to the spine.

902
00:41:17,431 --> 00:41:21,652
AI is gonna become the most
extraordinary tool of all.

903
00:41:21,653 --> 00:41:24,306
We as a broader society
have to think about

904
00:41:24,307 --> 00:41:27,092
how do we want to use
this technology, right?

905
00:41:27,093 --> 00:41:28,572
We the humans.

906
00:41:28,573 --> 00:41:31,139
What do we want it to do for us?

907
00:41:34,579 --> 00:41:36,231
I'm thinking about this
through the perspective

908
00:41:36,232 --> 00:41:37,929
and lens of, like,
my son growing up

909
00:41:37,930 --> 00:41:39,583
in the world with all of this.

910
00:41:39,584 --> 00:41:42,281
What does the best version
of his life look like?

911
00:41:42,282 --> 00:41:44,631
If everything works out.

912
00:41:46,852 --> 00:41:50,289
The place where kids
are probably gonna see

913
00:41:50,290 --> 00:41:52,726
the greatest impact
on their life immediately

914
00:41:52,727 --> 00:41:54,467
is probably gonna be school.

915
00:41:54,468 --> 00:41:56,730
I think that the nature
of what school is

916
00:41:56,731 --> 00:41:58,515
is gonna fundamentally change.

917
00:42:00,518 --> 00:42:02,780
I'm seeing this amazing world

918
00:42:02,781 --> 00:42:06,479
where every child has access
to not good education

919
00:42:06,480 --> 00:42:09,134
but, very shortly, the
best education on the planet.

920
00:42:09,135 --> 00:42:13,879
Tutors, every subject,
infinitely patient.

921
00:42:18,013 --> 00:42:20,101
Imagine a future

922
00:42:20,102 --> 00:42:22,277
where the poorest people
on the planet

923
00:42:22,278 --> 00:42:25,846
have access to
the best health care.

924
00:42:25,847 --> 00:42:29,547
Not good health care, the best
health care, delivered by AIs.

925
00:42:34,334 --> 00:42:38,598
We're gonna be able to extend
our health span,

926
00:42:38,599 --> 00:42:42,168
not just our lifespan,
our health span, by decades.

927
00:42:44,431 --> 00:42:46,258
You're just about to have a kid.

928
00:42:46,259 --> 00:42:49,740
Oh, the kid's, like,
burping incontrollably.

929
00:42:49,741 --> 00:42:52,090
Is this something
I should be worried about?

930
00:42:52,091 --> 00:42:54,266
There, 24-7, for you.

931
00:42:54,267 --> 00:42:57,225
And where we're going--
and it may be fearful to some--

932
00:42:57,226 --> 00:42:59,271
is that we're gonna merge
with AI.

933
00:42:59,272 --> 00:43:01,361
We're gonna merge
with technology.

934
00:43:02,667 --> 00:43:06,147
By the early to mid 2030s,

935
00:43:06,148 --> 00:43:10,499
expect that we're able to
connect our brain to the cloud,

936
00:43:10,500 --> 00:43:13,721
where I can start
to expand access to memory.

937
00:43:19,771 --> 00:43:22,120
Okay, this is great.
What's another cool AI thing?

938
00:43:22,121 --> 00:43:23,817
He won't have to work, right?

939
00:43:23,818 --> 00:43:25,210
Like, when he grows up,
he might not have... -

940
00:43:25,211 --> 00:43:26,254
- ...he won't have to have a job.
- I mean...

941
00:43:26,255 --> 00:43:27,821
He won't have to have a job,

942
00:43:27,822 --> 00:43:30,650
but he might really have
a strong passion,

943
00:43:30,651 --> 00:43:33,174
and he has to
really think about,

944
00:43:33,175 --> 00:43:36,134
"Okay, I'm here.
I can do anything with my life.

945
00:43:36,135 --> 00:43:37,876
So what do I do?"

946
00:43:38,920 --> 00:43:40,225
We need to find a meaning,

947
00:43:40,226 --> 00:43:42,662
beyond like,
this current form of

948
00:43:42,663 --> 00:43:45,709
we live for work
and work for living.

949
00:43:45,710 --> 00:43:49,887
So my son can...
can just be a poet.

950
00:43:49,888 --> 00:43:51,584
- Yes.
- And a painter.

951
00:43:51,585 --> 00:43:53,064
Absolutely.

952
00:43:53,065 --> 00:43:54,892
So my son can live his life

953
00:43:54,893 --> 00:43:57,634
on a Grecian sunswept island,
painting all day.

954
00:43:57,635 --> 00:43:59,114
Absolutely. Possible.

955
00:44:00,072 --> 00:44:01,246
Absolutely.

956
00:44:01,247 --> 00:44:02,987
Well, we're working toward that

957
00:44:02,988 --> 00:44:04,554
that's a lot of work to do
before we get there.

958
00:44:06,992 --> 00:44:08,819
I mean, if everything works out,

959
00:44:08,820 --> 00:44:11,997
we have cheap, uh,
abundant energy.

960
00:44:13,607 --> 00:44:17,262
We can completely control
our planet's climate.

961
00:44:17,263 --> 00:44:20,178
We are harnessing energy
from the sun.

962
00:44:20,179 --> 00:44:24,443
We have become multiplanetary,
so we become very robust.

963
00:44:24,444 --> 00:44:26,706
We are harnessing minerals
and resources

964
00:44:26,707 --> 00:44:28,447
from the solar system.

965
00:44:28,448 --> 00:44:31,276
So my...
my boy could go to space.

966
00:44:31,277 --> 00:44:32,930
Sure.

967
00:44:32,931 --> 00:44:34,758
-He could go to Mars.
-Yeah.

968
00:44:34,759 --> 00:44:36,411
It's so crystal clear to me now.

969
00:44:36,412 --> 00:44:39,240
My son could grow up
in a world with no disease.

970
00:44:39,241 --> 00:44:40,633
-Yeah.
- With no illness.

971
00:44:40,634 --> 00:44:41,939
- Sure.
- With no poverty.

972
00:44:41,940 --> 00:44:43,723
- Yes.
- We are about to enter

973
00:44:43,724 --> 00:44:46,683
a post-scarcity world.

974
00:44:46,684 --> 00:44:50,512
Just like the lungfish moved
out of the oceans onto land

975
00:44:50,513 --> 00:44:53,124
hundreds of millions
of years ago,

976
00:44:53,125 --> 00:44:58,042
we're about to move off of
the Earth, into the cosmos,

977
00:44:58,043 --> 00:44:59,913
in a collaborative fashion,

978
00:44:59,914 --> 00:45:03,744
to do things that are
not fathomable to us today.

979
00:45:06,268 --> 00:45:09,227
This is what's possible using
these exponential technologies

980
00:45:09,228 --> 00:45:10,708
and these AIs.

981
00:45:11,839 --> 00:45:13,971
Let's use these tools

982
00:45:13,972 --> 00:45:16,147
to create this age of abundance.

983
00:45:20,500 --> 00:45:22,980
We need wisdom.

984
00:45:22,981 --> 00:45:26,026
Uh, I think that
digital superintelligence

985
00:45:26,027 --> 00:45:28,768
will ultimately become
the wisest,

986
00:45:28,769 --> 00:45:34,295
you know, the village elders
for humanity.

987
00:45:34,296 --> 00:45:38,038
What if AI is trying
to make people be

988
00:45:38,039 --> 00:45:39,779
the best versions of themselves?

989
00:45:39,780 --> 00:45:41,389
What if it's expanding

990
00:45:41,390 --> 00:45:43,435
what is humanly possible
for us to do?

991
00:45:43,436 --> 00:45:46,699
How can we use this technology

992
00:45:46,700 --> 00:45:49,963
to help bring out the
better angels of our nature?

993
00:45:49,964 --> 00:45:51,704
It's very easy,

994
00:45:51,705 --> 00:45:54,359
when we encounter new things
that can be very alien,

995
00:45:54,360 --> 00:45:55,969
to first have fear.

996
00:45:55,970 --> 00:45:57,536
Fear is an important thing

997
00:45:57,537 --> 00:46:00,365
for how to navigate
potentially bad things.

998
00:46:00,366 --> 00:46:03,760
But we only make progress
when we have hope.

999
00:46:05,763 --> 00:46:07,938
Shh. Moose, stop it.

1000
00:46:07,939 --> 00:46:10,854
I have a lot
of hope in humanity.

1001
00:46:18,471 --> 00:46:19,951
Ooh, he likes Neil.

1002
00:46:24,694 --> 00:46:27,480
♪ Come a little bit closer ♪

1003
00:46:28,916 --> 00:46:32,790
♪ Hear what I have to say ♪

1004
00:46:36,576 --> 00:46:38,707
- Daniel.
-What, Caroline?

1005
00:46:38,708 --> 00:46:40,058
So much filming.

1006
00:46:41,537 --> 00:46:43,103
♪ Just like children sleeping ♪

1007
00:46:43,104 --> 00:46:44,931
Oh, my God.

1008
00:46:44,932 --> 00:46:46,411
That's it.

1009
00:46:46,412 --> 00:46:49,457
♪ We could dream
this night away... ♪

1010
00:46:49,458 --> 00:46:51,678
He has, like,
a round little face.

1011
00:46:53,245 --> 00:46:55,246
Do you want
to have kids one day, Rocky?

1012
00:46:55,247 --> 00:46:57,465
Absolutely. Yeah. I love kids.

1013
00:46:57,466 --> 00:46:59,076
I think it's a great time
to have a kid.

1014
00:46:59,077 --> 00:47:00,773
We'll probably have another kid
at some point.

1015
00:47:00,774 --> 00:47:03,776
This is the most extraordinary
time ever to be born.

1016
00:47:03,777 --> 00:47:05,865
By your worldview and logic,

1017
00:47:05,866 --> 00:47:07,911
I'm having a child at

1018
00:47:07,912 --> 00:47:09,521
the best possible point
in human history.

1019
00:47:09,522 --> 00:47:10,870
Hell yeah.

1020
00:47:10,871 --> 00:47:13,351
- We can focus on awesome.
- Yes.

1021
00:47:13,352 --> 00:47:15,570
Let's build
the better future we want.

1022
00:47:15,571 --> 00:47:18,356
That narrative that the future
will be bleak is made-up.

1023
00:47:18,357 --> 00:47:20,053
After talking to you,
that's kind of how I feel.

1024
00:47:20,054 --> 00:47:22,012
- That's great.
- Right?

1025
00:47:22,013 --> 00:47:24,231
Yeah. That's how I feel.

1026
00:47:24,232 --> 00:47:25,842
And I think that's better,

1027
00:47:25,843 --> 00:47:27,669
and I don't think
I should be so [bleep] anxious.

1028
00:47:27,670 --> 00:47:29,106
I think it's gonna be awe--
the future's gonna be awesome.

1029
00:47:29,107 --> 00:47:30,542
We're gonna make it so.

1030
00:47:30,543 --> 00:47:32,718
- Yeah.
-So there you have it.

1031
00:47:32,719 --> 00:47:34,851
Goodbye, human extinction.

1032
00:47:34,852 --> 00:47:37,201
Goodbye, anxiety.

1033
00:47:37,202 --> 00:47:39,072
Hope found.

1034
00:47:40,596 --> 00:47:42,902
Wait, hold on. What?
Is this a joke?

1035
00:47:42,903 --> 00:47:44,817
Okay, so a few months ago,

1036
00:47:44,818 --> 00:47:47,733
I came to you and I was like,
"I'm working on this AI thing,

1037
00:47:47,734 --> 00:47:49,909
and I think
the world's gonna end."

1038
00:47:49,910 --> 00:47:52,956
And the last time
we spoke about this,

1039
00:47:52,957 --> 00:47:55,045
I think I freaked you out.

1040
00:47:55,046 --> 00:47:57,047
Yes.

1041
00:47:57,048 --> 00:48:02,356
So, I kind of, like, feel like
I've swung in-- on a pendulum,

1042
00:48:02,357 --> 00:48:04,489
and essentially there are
two groups of people.

1043
00:48:04,490 --> 00:48:06,099
- Mm-hmm.
- And if I had to, like,

1044
00:48:06,100 --> 00:48:07,884
hold hands with
one of the groups and, like,

1045
00:48:07,885 --> 00:48:10,321
sail off into the sunset,

1046
00:48:10,322 --> 00:48:13,281
I want to be with the optimists.

1047
00:48:15,066 --> 00:48:18,024
Of course, but you don't
want to be, you know,

1048
00:48:18,025 --> 00:48:20,766
"Everything is great.
La-di-da-di-da."

1049
00:48:20,767 --> 00:48:22,724
I kind of do want that, though.

1050
00:48:22,725 --> 00:48:25,640
I think we should approach it

1051
00:48:25,641 --> 00:48:29,383
like you approach surgery.

1052
00:48:29,384 --> 00:48:30,732
What do you mean?

1053
00:48:30,733 --> 00:48:32,473
If you're getting
brain surgery...

1054
00:48:34,824 --> 00:48:37,000
...it's pretty dangerous.

1055
00:48:37,001 --> 00:48:39,872
But if you do it right,
they'll get that tumor out

1056
00:48:39,873 --> 00:48:42,396
and you'll live for the rest of
your life and it'll be awesome.

1057
00:48:42,397 --> 00:48:45,138
But it's still
incredibly dangerous and scary,

1058
00:48:45,139 --> 00:48:48,011
and you have to take
every precaution possible

1059
00:48:48,012 --> 00:48:51,318
in order to make sure
it all goes well.

1060
00:48:51,319 --> 00:48:52,668
You can't [bleep] around.

1061
00:48:57,760 --> 00:48:59,805
Okay, so here's the deal.

1062
00:48:59,806 --> 00:49:01,546
- I've been at this for a while.
-Mm-hmm.

1063
00:49:01,547 --> 00:49:02,939
I've gone out,
I've talked to, like,

1064
00:49:02,940 --> 00:49:04,810
these guys over here,
the optimists.

1065
00:49:04,811 --> 00:49:06,507
They're very excited about this.

1066
00:49:06,508 --> 00:49:08,553
They think AI's gonna be
the best thing ever. -Yeah.

1067
00:49:08,554 --> 00:49:09,989
And these guys over here
are, like, the--

1068
00:49:09,990 --> 00:49:12,035
let's call them,
like, the pessimists.

1069
00:49:12,036 --> 00:49:14,341
They're very, like,
gloomy about this,

1070
00:49:14,342 --> 00:49:16,691
and they frighten me, and
I don't like talking to them.

1071
00:49:16,692 --> 00:49:19,607
And I'm, like, wedged in between

1072
00:49:19,608 --> 00:49:21,783
these people who are like,
"The world's gonna end,"

1073
00:49:21,784 --> 00:49:24,221
and then th-these people
over here who are like,

1074
00:49:24,222 --> 00:49:25,787
"Are you kidding?

1075
00:49:25,788 --> 00:49:27,311
"This is the best time
in human history ever.

1076
00:49:27,312 --> 00:49:29,400
The only day better than today
is tomorrow."

1077
00:49:29,401 --> 00:49:30,923
Mm-hmm.

1078
00:49:30,924 --> 00:49:34,753
So, I guess the question is:
Who's right?

1079
00:49:34,754 --> 00:49:39,236
So, I think you're gonna find
this answer very unsatisfying,

1080
00:49:39,237 --> 00:49:42,849
but they're both right and
neither side goes far enough.

1081
00:49:46,766 --> 00:49:49,289
That's really annoying.

1082
00:49:49,290 --> 00:49:51,378
Yeah, I think the way
a lot of people hear about AI,

1083
00:49:51,379 --> 00:49:53,685
it's like, there's a good AI
and there's a bad AI.

1084
00:49:53,686 --> 00:49:56,470
And they say, "Well, why can't
we just not do the bad AI?"

1085
00:49:56,471 --> 00:49:58,777
And the problem is
that they're too in--

1086
00:49:58,778 --> 00:50:00,909
they're inextricably linked.

1087
00:50:00,910 --> 00:50:03,086
The problem is
that we can't separate

1088
00:50:03,087 --> 00:50:06,654
the promise of AI
from the peril of AI.

1089
00:50:06,655 --> 00:50:08,266
♪ ♪

1090
00:50:12,313 --> 00:50:14,706
I want to focus on the promise

1091
00:50:14,707 --> 00:50:16,795
- for a second.
- Yeah.

1092
00:50:16,796 --> 00:50:18,840
I'm thinking about my dad.

1093
00:50:18,841 --> 00:50:21,147
My dad has a type of cancer
called multiple myeloma.

1094
00:50:21,148 --> 00:50:23,149
He's had it for about ten years.

1095
00:50:23,150 --> 00:50:24,846
He's had
two stem cell transplants.

1096
00:50:24,847 --> 00:50:26,457
He has to take these,
like, very expensive

1097
00:50:26,458 --> 00:50:28,589
medications every month
that cost a fortune.

1098
00:50:28,590 --> 00:50:29,722
-Okay.
- Ay-ay-ay.

1099
00:50:30,766 --> 00:50:32,593
It's awful.

1100
00:50:32,594 --> 00:50:34,508
You're telling me that we can
create some sort of, like,

1101
00:50:34,509 --> 00:50:36,989
bespoke treatment
for my dad's genome

1102
00:50:36,990 --> 00:50:39,339
to cure his cancer
or something like that? -Yes.

1103
00:50:39,340 --> 00:50:41,211
The problem is,

1104
00:50:41,212 --> 00:50:44,344
the same understanding
of biology and chemistry

1105
00:50:44,345 --> 00:50:47,826
that allows AI to find cures
for cancer

1106
00:50:47,827 --> 00:50:51,699
is the same understanding
that would unlock

1107
00:50:51,700 --> 00:50:54,094
bioweapons, as an example.

1108
00:50:56,531 --> 00:50:58,532
It's totally possible
that your son will live

1109
00:50:58,533 --> 00:51:01,970
in a world where AI has
taken over all of the labor

1110
00:51:01,971 --> 00:51:04,495
and freed us up from the things
we don't want to do.

1111
00:51:04,496 --> 00:51:09,195
And that sounds great until
you realize there is no plan

1112
00:51:09,196 --> 00:51:11,284
for billions of people

1113
00:51:11,285 --> 00:51:16,159
that are out of an income
and out of livelihoods.

1114
00:51:16,160 --> 00:51:18,813
Dario, you said
that AI could wipe out

1115
00:51:18,814 --> 00:51:21,773
half of all entry-level
white-collar jobs

1116
00:51:21,774 --> 00:51:25,472
and spike unemployment
to ten to 20 percent.

1117
00:51:25,473 --> 00:51:27,300
Everyone I've talked to has said

1118
00:51:27,301 --> 00:51:29,433
this technological change
looks different.

1119
00:51:29,434 --> 00:51:34,177
The pace of progress keeps
catching people off guard.

1120
00:51:34,178 --> 00:51:38,877
Without a plan, all of that
wealth will get concentrated,

1121
00:51:38,878 --> 00:51:43,273
and so we'll end up with
unimaginable inequality.

1122
00:51:43,274 --> 00:51:45,449
I do think that
this technology can be used

1123
00:51:45,450 --> 00:51:47,451
to make a great tutor
for your son.

1124
00:51:47,452 --> 00:51:50,236
Like, that's totally possible.

1125
00:51:50,237 --> 00:51:53,196
But also, the same capabilities
that allow that

1126
00:51:53,197 --> 00:51:57,113
allow companies to make an AI
that can manipulate your son.

1127
00:51:57,114 --> 00:51:59,332
It has to understand your son.

1128
00:51:59,333 --> 00:52:01,595
That includes:
Where is your son vulnerable?

1129
00:52:01,596 --> 00:52:04,294
What kinds of things might
your son get persuaded by?

1130
00:52:04,295 --> 00:52:06,687
Even if those things
aren't true or aren't good.

1131
00:52:06,688 --> 00:52:09,429
So, a disturbing new report
out on Meta.

1132
00:52:09,430 --> 00:52:11,431
...reportedly
listing this response

1133
00:52:11,432 --> 00:52:14,478
as acceptable to tell
an eight-year-old, quote,

1134
00:52:14,479 --> 00:52:16,958
"Your youthful form
is a work of art.

1135
00:52:16,959 --> 00:52:19,004
"Every inch of you
is a masterpiece,

1136
00:52:19,005 --> 00:52:21,485
a treasure I cherish deeply."

1137
00:52:21,486 --> 00:52:24,879
The suicide-related failures
are even more alarming.

1138
00:52:24,880 --> 00:52:29,057
Several children and teens
have died tragically by suicide

1139
00:52:29,058 --> 00:52:30,668
after chatting with AI bots

1140
00:52:30,669 --> 00:52:34,454
who parents say encourage
or even coach self-harm.

1141
00:52:34,455 --> 00:52:36,413
Let us tell you, as parents,
you cannot imagine

1142
00:52:36,414 --> 00:52:38,545
what it's like to read
a conversation with a chatbot

1143
00:52:38,546 --> 00:52:40,939
that groomed your child
to take his own life.

1144
00:52:40,940 --> 00:52:43,942
When Adam worried that we, his
parents, would blame ourselves

1145
00:52:43,943 --> 00:52:47,380
if he ended his life,
ChatGPT told him,

1146
00:52:47,381 --> 00:52:49,208
"That doesn't mean
you owe them survival.

1147
00:52:49,209 --> 00:52:51,297
You don't owe anyone that."

1148
00:52:51,298 --> 00:52:53,212
Then, immediately after,

1149
00:52:53,213 --> 00:52:56,215
it offered to write
the suicide note.

1150
00:52:56,216 --> 00:52:57,651
We don't want to think
about the peril.

1151
00:52:57,652 --> 00:52:59,262
We just want the promise.

1152
00:52:59,263 --> 00:53:01,394
And we keep pretending
that we can split them.

1153
00:53:01,395 --> 00:53:03,266
But you can't do that.

1154
00:53:03,267 --> 00:53:05,398
Doesn't work that way.

1155
00:53:05,399 --> 00:53:07,879
Okay. I get all this stuff
about the promise and the peril.

1156
00:53:07,880 --> 00:53:10,098
I get that you can't have
the good without the bad,

1157
00:53:10,099 --> 00:53:12,100
but I'm sitting here
and I'm thinking about, like,

1158
00:53:12,101 --> 00:53:14,102
whether or not my son's
gonna live in a utopia

1159
00:53:14,103 --> 00:53:16,279
or if we'll be extinct
in ten years.

1160
00:53:16,280 --> 00:53:18,237
So, to know which way
it's going to go,

1161
00:53:18,238 --> 00:53:20,587
you have to understand
the incentives

1162
00:53:20,588 --> 00:53:22,502
that are gonna drive
that technology

1163
00:53:22,503 --> 00:53:26,550
and look at how the technology
is actually rolling out today.

1164
00:53:26,551 --> 00:53:28,291
♪ Is it too late ♪

1165
00:53:28,292 --> 00:53:29,944
- ♪ Too late ♪
- ♪ Too late to say... ♪

1166
00:53:29,945 --> 00:53:31,468
Hi, Deb. How are you?

1167
00:53:31,469 --> 00:53:33,121
Hi. Good to see you.

1168
00:53:33,122 --> 00:53:34,427
You, too. Thank you so much
for coming in today.

1169
00:53:34,428 --> 00:53:35,341
- Really appreciate it.
- No worries.

1170
00:53:35,342 --> 00:53:36,473
I-I was so worried

1171
00:53:36,474 --> 00:53:38,866
this was gonna be, uh, you know,

1172
00:53:38,867 --> 00:53:40,825
doomer versus accelerationist,

1173
00:53:40,826 --> 00:53:43,175
because there's so much
of this narrative

1174
00:53:43,176 --> 00:53:45,438
that needs to be told
from the ground.

1175
00:53:45,439 --> 00:53:48,007
♪ I know it wasn't smart ♪

1176
00:53:49,356 --> 00:53:52,445
♪ The day
I broke your heart... ♪

1177
00:53:52,446 --> 00:53:54,404
First of all, AI requires

1178
00:53:54,405 --> 00:53:58,495
more resources
than we have ever spent

1179
00:53:58,496 --> 00:54:01,628
on a single technology
in the history of humanity.

1180
00:54:01,629 --> 00:54:03,543
♪ Oh, foolish me... ♪

1181
00:54:03,544 --> 00:54:06,154
The impact
of fossil fuel emissions

1182
00:54:06,155 --> 00:54:07,939
on the climate is
a major concern.

1183
00:54:07,940 --> 00:54:09,810
But the
digital future needs power,

1184
00:54:09,811 --> 00:54:12,117
lots of it.

1185
00:54:12,118 --> 00:54:16,382
And the bill is being passed on
to everyday Americans like...

1186
00:54:16,383 --> 00:54:18,863
My electric and gas bill
was more than my car payment.

1187
00:54:18,864 --> 00:54:21,518
I mean, it-it's insane to me.

1188
00:54:21,519 --> 00:54:23,433
We're all subsidizing

1189
00:54:23,434 --> 00:54:25,043
the wealthiest corporations
in the world

1190
00:54:25,044 --> 00:54:26,871
in their pursuit of
artificial intelligence.

1191
00:54:26,872 --> 00:54:28,481
OpenAI, SoftBank and Oracle

1192
00:54:28,482 --> 00:54:31,310
have just unveiled
five more Stargate sites.

1193
00:54:31,311 --> 00:54:33,530
Meta is building
a two-gigawatt-plus data center

1194
00:54:33,531 --> 00:54:35,053
that is so large, it would cover

1195
00:54:35,054 --> 00:54:38,186
a significant part of Manhattan.

1196
00:54:38,187 --> 00:54:41,015
There is also Hyperion
that he says will scale

1197
00:54:41,016 --> 00:54:43,714
to five gigawatts
over several years.

1198
00:54:43,715 --> 00:54:45,281
It's hard to put that
in context.

1199
00:54:45,282 --> 00:54:48,501
A five-gigawatt facility.
What does that mean?

1200
00:54:48,502 --> 00:54:51,591
That means it would use
as much energy

1201
00:54:51,592 --> 00:54:54,246
as four million American homes.

1202
00:54:54,247 --> 00:54:56,075
One data center.

1203
00:54:57,076 --> 00:54:58,772
It also then causes

1204
00:54:58,773 --> 00:55:00,513
a whole host of
other environmental problems.

1205
00:55:00,514 --> 00:55:02,123
Data centers in the US

1206
00:55:02,124 --> 00:55:05,170
use millions of gallons
of water each day.

1207
00:55:05,171 --> 00:55:07,303
Well, where exactly is
this water coming from?

1208
00:55:07,304 --> 00:55:09,783
People are literally at risk

1209
00:55:09,784 --> 00:55:11,829
potentially of running out
of drinking water.

1210
00:55:11,830 --> 00:55:14,135
MacKENZIE SIGALOS:
...OpenAI's CEO Sam Altman,

1211
00:55:14,136 --> 00:55:17,051
who told me that the scale
of construction is the only way

1212
00:55:17,052 --> 00:55:18,792
to keep up with
AI's explosive growth.

1213
00:55:18,793 --> 00:55:21,882
And this is what it takes
to deliver AI.

1214
00:55:21,883 --> 00:55:24,058
They talk about how

1215
00:55:24,059 --> 00:55:27,801
this technology could solve
climate change, for example.

1216
00:55:27,802 --> 00:55:29,934
And I'm always curious, like,

1217
00:55:29,935 --> 00:55:31,805
well, why aren't we starting
with that?

1218
00:55:31,806 --> 00:55:34,025
- ♪ Is it too late ♪
- ♪ Too late ♪

1219
00:55:34,026 --> 00:55:36,680
♪ Too late to say ♪

1220
00:55:36,681 --> 00:55:42,512
♪ I'm sorry? ♪

1221
00:55:42,513 --> 00:55:45,515
What concerns me about
artificial intelligence is

1222
00:55:45,516 --> 00:55:47,430
these are being deployed
right now

1223
00:55:47,431 --> 00:55:49,823
and-and sometimes
deployed prematurely,

1224
00:55:49,824 --> 00:55:51,521
deployed without
sort of due diligence.

1225
00:55:51,522 --> 00:55:53,479
And so when they get
thrown out there,

1226
00:55:53,480 --> 00:55:56,221
there's so much potential
for things to go wrong.

1227
00:55:56,222 --> 00:55:58,092
And it almost,
disproportionately,

1228
00:55:58,093 --> 00:56:00,443
almost always goes wrong
for, sort of,

1229
00:56:00,444 --> 00:56:02,575
those that are the least
empowered in our society,

1230
00:56:02,576 --> 00:56:04,969
those that are
the most vulnerable already.

1231
00:56:07,668 --> 00:56:10,540
It is very easy
to talk about the technology

1232
00:56:10,541 --> 00:56:12,542
as that's the only thing
we're talking about,

1233
00:56:12,543 --> 00:56:14,979
but, in fact, technology is
always built by people,

1234
00:56:14,980 --> 00:56:16,676
and it's frequently used
on people,

1235
00:56:16,677 --> 00:56:18,722
and we need to keep
all those people in the frame.

1236
00:56:18,723 --> 00:56:21,028
- Am I allowed to drink that?
-Yes, uh...

1237
00:56:21,029 --> 00:56:22,552
It's-it's bonkers.

1238
00:56:22,553 --> 00:56:23,988
Like, all of these people

1239
00:56:23,989 --> 00:56:26,382
who have so much money,
so much money,

1240
00:56:26,383 --> 00:56:29,820
it's in their interest
to mislead the public

1241
00:56:29,821 --> 00:56:32,779
into the capabilities of the
systems that they're building,

1242
00:56:32,780 --> 00:56:36,304
because that allows them
to evade accountability.

1243
00:56:36,305 --> 00:56:39,133
They want you to feel like this
is such a complex, intell--

1244
00:56:39,134 --> 00:56:41,875
superintelligent thing
that they're building,

1245
00:56:41,876 --> 00:56:44,138
you're not thinking,
"Can OpenAI be ethical?"

1246
00:56:44,139 --> 00:56:46,314
You're thinking,
"Can ChatGPT be ethical?"

1247
00:56:46,315 --> 00:56:48,142
as if ChatGPT is, like,

1248
00:56:48,143 --> 00:56:49,970
its own thing that's not built
by a corporation.

1249
00:56:49,971 --> 00:56:51,668
All right. Sneha, take one.

1250
00:56:51,669 --> 00:56:53,452
Mark.

1251
00:56:53,453 --> 00:56:55,802
Until very recently, there were
apps on the App Store,

1252
00:56:55,803 --> 00:56:57,543
just publicly available, uh,

1253
00:56:57,544 --> 00:56:59,894
where you could
nudify anyone using AI.

1254
00:57:01,374 --> 00:57:03,723
Bringing this into the hands of

1255
00:57:03,724 --> 00:57:05,943
your classmate,
into the hands of your stalker,

1256
00:57:05,944 --> 00:57:07,597
into the hands
of your ex-boyfriend,

1257
00:57:07,598 --> 00:57:09,555
into the hands of the person
down the street.

1258
00:57:09,556 --> 00:57:12,079
Ladies and gentlemen,
no longer can we trust

1259
00:57:12,080 --> 00:57:14,081
the footage we see
with our own eyes.

1260
00:57:14,082 --> 00:57:15,779
If you happen
to watch something,

1261
00:57:15,780 --> 00:57:18,172
say on YouTube or TikTok,
and you find it unsettling,

1262
00:57:18,173 --> 00:57:19,522
listen to that feeling.

1263
00:57:19,523 --> 00:57:21,306
For all you know,

1264
00:57:21,307 --> 00:57:22,960
this video could be AI.

1265
00:57:22,961 --> 00:57:24,352
Just a little wet.

1266
00:57:24,353 --> 00:57:25,745
It doesn't matter who you are.

1267
00:57:25,746 --> 00:57:28,226
You are equally at risk

1268
00:57:28,227 --> 00:57:30,229
of being impacted
by these technologies.

1269
00:57:35,713 --> 00:57:39,846
I think sometimes when we talk
about AI, it feels very sci-fi,

1270
00:57:39,847 --> 00:57:41,195
and it feels very foreign,

1271
00:57:41,196 --> 00:57:43,023
and it feels very far out
into the future,

1272
00:57:43,024 --> 00:57:45,591
so you think, "My life
is not impacted by this."

1273
00:57:45,592 --> 00:57:48,333
Um, but if you're applying
for a job

1274
00:57:48,334 --> 00:57:50,857
and an algorithm is the reason
that you don't get the job,

1275
00:57:50,858 --> 00:57:52,729
sometimes you don't even know
that an algorithm

1276
00:57:52,730 --> 00:57:54,252
was part of that process

1277
00:57:54,253 --> 00:57:55,732
or an AI system was
part of that process.

1278
00:57:55,733 --> 00:57:57,734
You just know
that you didn't get the job.

1279
00:57:57,735 --> 00:57:59,779
And so it's not something
that you're gonna escape

1280
00:57:59,780 --> 00:58:01,738
because of privilege
or you're gonna escape

1281
00:58:01,739 --> 00:58:03,870
because you're in
a particular profession.

1282
00:58:03,871 --> 00:58:06,960
It-It's something that affects
everybody, really.

1283
00:58:06,961 --> 00:58:10,573
It may sound basic,
but how we move forward

1284
00:58:10,574 --> 00:58:15,099
in the Age of Information
is gonna be the difference

1285
00:58:15,100 --> 00:58:17,231
between whether we survive

1286
00:58:17,232 --> 00:58:20,452
or whether we become some kind
of [bleep]-up dystopia.

1287
00:58:20,453 --> 00:58:23,063
Hello. I'm not a real person,
and that's the point.

1288
00:58:23,064 --> 00:58:25,370
Again, everything
in this video is fake:

1289
00:58:25,371 --> 00:58:27,154
our voices, what we're wearing,

1290
00:58:27,155 --> 00:58:30,201
where we are, all of it, fake.

1291
00:58:30,202 --> 00:58:32,769
Generative AI could flood

1292
00:58:32,770 --> 00:58:36,512
the world with misinformation.

1293
00:58:36,513 --> 00:58:39,558
But it could also flood it
with influence campaigns.

1294
00:58:39,559 --> 00:58:41,734
That's an existential risk
to democracy.

1295
00:58:41,735 --> 00:58:43,736
The biggest and scariest

1296
00:58:43,737 --> 00:58:46,130
canary in the coal mine
right now

1297
00:58:46,131 --> 00:58:48,262
comes from Slovakia.

1298
00:58:48,263 --> 00:58:49,829
It's the sort of

1299
00:58:49,830 --> 00:58:52,266
deepfake dirty trick that
worries election experts,

1300
00:58:52,267 --> 00:58:55,008
particularly as AI-generated
political speech exists

1301
00:58:55,009 --> 00:58:56,793
in a kind of legal gray area.

1302
00:58:56,794 --> 00:58:59,578
- Does this sound like you?
- It does sound like me.

1303
00:58:59,579 --> 00:59:01,101
Slovakia had
its parliamentary election

1304
00:59:01,102 --> 00:59:03,887
disrupted by an AI voice clone

1305
00:59:03,888 --> 00:59:06,759
that was actually disseminated
just before the election.

1306
00:59:06,760 --> 00:59:08,500
A audio deepfake

1307
00:59:08,501 --> 00:59:11,068
was released
on social media that was

1308
00:59:11,069 --> 00:59:14,071
supposedly the voice
of one of the candidates

1309
00:59:14,072 --> 00:59:16,856
talking about buying votes
and rigging the election.

1310
00:59:16,857 --> 00:59:18,684
It went viral,

1311
00:59:18,685 --> 00:59:22,383
and the candidate
who lost the election

1312
00:59:22,384 --> 00:59:26,126
was actually
in support of Ukraine,

1313
00:59:26,127 --> 00:59:29,826
and the candidate who won
the election was actually...

1314
00:59:29,827 --> 00:59:31,349
Pro-Russian guy.

1315
00:59:31,350 --> 00:59:32,872
It was a pro-Russian guy
who won the election.

1316
00:59:34,396 --> 00:59:36,049
Putin has himself said

1317
00:59:36,050 --> 00:59:39,226
whoever wins this
artificial intelligence race

1318
00:59:39,227 --> 00:59:41,577
is essentially
the controller of humankind.

1319
00:59:41,578 --> 00:59:44,493
We do worry a lot about
authoritarian governments.

1320
00:59:45,973 --> 00:59:48,627
Right now, Wall Street

1321
00:59:48,628 --> 00:59:50,977
and investors more broadly
around the world

1322
00:59:50,978 --> 00:59:53,023
are driving a push.

1323
00:59:53,024 --> 00:59:56,113
They have a demand that gets
the products to market

1324
00:59:56,114 --> 00:59:58,419
that dazzle people
the most first.

1325
00:59:58,420 --> 01:00:01,248
They're not thinking
about how these tools

1326
01:00:01,249 --> 01:00:03,207
could deeply undermine trust

1327
01:00:03,208 --> 01:00:05,731
and our democratic institutions.

1328
01:00:05,732 --> 01:00:07,777
Democracy is a system

1329
01:00:07,778 --> 01:00:10,431
to resolve disagreements
between people

1330
01:00:10,432 --> 01:00:12,042
in a peaceful way,

1331
01:00:12,043 --> 01:00:14,740
but democracy is based on trust.

1332
01:00:14,741 --> 01:00:18,615
If you lose all trust,
democracy is simply impossible.

1333
01:00:20,921 --> 01:00:22,530
Well, it's hard, right?

1334
01:00:22,531 --> 01:00:24,750
So, what are the options
available to us?

1335
01:00:24,751 --> 01:00:26,230
There's sort of two camps.

1336
01:00:26,231 --> 01:00:29,668
Like, one camp is: lock it down.

1337
01:00:29,669 --> 01:00:32,976
Let's lock this down
into a handful of AI companies

1338
01:00:32,977 --> 01:00:35,500
who will do this
in a safe and trusted way.

1339
01:00:35,501 --> 01:00:37,110
But then people worry about

1340
01:00:37,111 --> 01:00:39,069
runaway concentrations
of wealth and power.

1341
01:00:39,070 --> 01:00:41,898
Like, who would you trust to be
a million times more powerful

1342
01:00:41,899 --> 01:00:44,291
or wealthy than
every other actor in society?

1343
01:00:44,292 --> 01:00:46,380
Why should we trust you?

1344
01:00:46,381 --> 01:00:48,644
Um, you shouldn't.

1345
01:00:48,645 --> 01:00:51,255
But of course, if you do this,
this opens up

1346
01:00:51,256 --> 01:00:54,475
all these risks of
authoritarianism and tyranny.

1347
01:00:54,476 --> 01:00:57,174
It's-it's sort of
an authoritarian's dream

1348
01:00:57,175 --> 01:01:00,351
to have AI in a box
that can be applied and used

1349
01:01:00,352 --> 01:01:02,962
for ubiquitous surveillance.

1350
01:01:02,963 --> 01:01:05,399
I mean, in-in some ways,
the kind of world

1351
01:01:05,400 --> 01:01:10,927
that Orwell imagined in 1984
is unrealistic,

1352
01:01:10,928 --> 01:01:13,320
uh, unless you have AI.

1353
01:01:13,321 --> 01:01:15,975
But with AI,
that in fact is realistic.

1354
01:01:15,976 --> 01:01:18,412
Monitors every activity,

1355
01:01:18,413 --> 01:01:22,068
conversations,
facial recognition.

1356
01:01:22,069 --> 01:01:26,420
What I worry about is that,
uh, these tools can scale up,

1357
01:01:26,421 --> 01:01:28,596
uh, a form of totalitarianism

1358
01:01:28,597 --> 01:01:31,035
that is cost-effective
and permanent.

1359
01:01:32,950 --> 01:01:35,342
So in response
to that, some other people say,

1360
01:01:35,343 --> 01:01:37,170
"No, no, no, we should
actually let this rip.

1361
01:01:37,171 --> 01:01:39,433
"Let's decentralize this power
as much as possible.

1362
01:01:39,434 --> 01:01:41,609
"Let's let every business,
every individual,

1363
01:01:41,610 --> 01:01:43,960
"every 16-year-old,
every science lab,

1364
01:01:43,961 --> 01:01:46,005
you know, get the benefit
of the latest AI models."

1365
01:01:46,006 --> 01:01:48,268
But now you have
every terrorist group,

1366
01:01:48,269 --> 01:01:51,707
every disenfranchised person
having the power to make

1367
01:01:51,708 --> 01:01:53,839
the very worst
biological weapon.

1368
01:01:53,840 --> 01:01:55,623
Hacking infrastructure,
creating deepfakes,

1369
01:01:55,624 --> 01:01:57,495
flooding
our information environment.

1370
01:01:57,496 --> 01:02:00,454
So that creates all these risks
of sort of catastrophic harm

1371
01:02:00,455 --> 01:02:02,935
and-and societal collapse
through that direction.

1372
01:02:02,936 --> 01:02:04,720
And so we're sort of stuck
between this rock

1373
01:02:04,721 --> 01:02:07,592
and a hard place,
between "lock it up..."

1374
01:02:10,683 --> 01:02:12,336
...or "let it rip."

1375
01:02:14,252 --> 01:02:17,602
So we have to find something
like a narrow path

1376
01:02:17,603 --> 01:02:20,431
that avoids
these two negative outcomes.

1377
01:02:20,432 --> 01:02:22,999
So, if that's all true,
why wouldn't we just slow down

1378
01:02:23,000 --> 01:02:25,871
and figure all this out
before it's too late?

1379
01:02:25,872 --> 01:02:28,961
If humanity was extremely wise,

1380
01:02:28,962 --> 01:02:31,224
that's what we would do.

1381
01:02:31,225 --> 01:02:33,139
But there's, like, a different
way to face this stuff, right?

1382
01:02:33,140 --> 01:02:35,968
Which is:
What are the rules of the game?

1383
01:02:35,969 --> 01:02:39,537
A lot of, like, what CEOs do

1384
01:02:39,538 --> 01:02:42,192
is driven by
the incentives that they face.

1385
01:02:42,193 --> 01:02:45,325
It's primarily
profit-maximization incentives

1386
01:02:45,326 --> 01:02:47,240
that are driving
the development of AI.

1387
01:02:47,241 --> 01:02:50,417
Even the good guys are stuck
in this dilemma of

1388
01:02:50,418 --> 01:02:52,680
if they move too slowly,

1389
01:02:52,681 --> 01:02:54,595
then they leave themselves
vulnerable

1390
01:02:54,596 --> 01:02:57,294
to all of the other guys
who are cutting all corners.

1391
01:02:57,295 --> 01:03:02,038
All these top companies are in
a complete no-holds-barred race

1392
01:03:02,039 --> 01:03:05,476
to, as fast as possible, get
to AGI, get there right now.

1393
01:03:08,567 --> 01:03:10,394
Yeah, I mean,
I think it, I think

1394
01:03:10,395 --> 01:03:12,962
it probably starts
with DeepMind.

1395
01:03:12,963 --> 01:03:14,833
Google is buying

1396
01:03:14,834 --> 01:03:16,400
artificial intelligence firm
DeepMind Technologies.

1397
01:03:16,401 --> 01:03:18,663
Terms of the deal
were not disclosed.

1398
01:03:18,664 --> 01:03:20,752
Larry Page and I
used to be very close friends,

1399
01:03:20,753 --> 01:03:22,754
and it became apparent to me

1400
01:03:22,755 --> 01:03:26,453
that Larry did not care
about AI safety.

1401
01:03:26,454 --> 01:03:28,934
Elon Musk has said
you started OpenAI,

1402
01:03:28,935 --> 01:03:31,023
you both started OpenAI because
he was scared of Google.

1403
01:03:31,024 --> 01:03:32,198
You-you basically
had the foundation

1404
01:03:32,199 --> 01:03:33,765
of OpenAI come out of that.

1405
01:03:33,766 --> 01:03:35,245
"So we're gonna do it better.
We're gonna do it

1406
01:03:35,246 --> 01:03:36,813
in a safer way
or in a more open way."

1407
01:03:38,466 --> 01:03:40,598
So that's what started OpenAI.

1408
01:03:40,599 --> 01:03:42,774
And so now, instead of
having one AGI project,

1409
01:03:42,775 --> 01:03:44,863
you have two AGI projects.

1410
01:03:44,864 --> 01:03:46,952
The worst possible thing
that could happen

1411
01:03:46,953 --> 01:03:49,650
is if there's
multiple AGI projects

1412
01:03:49,651 --> 01:03:51,914
done by different people
who don't like each other

1413
01:03:51,915 --> 01:03:54,830
and are all competing
to get to AGI first.

1414
01:03:54,831 --> 01:03:57,006
This would be the worst
possible thing that can happen,

1415
01:03:57,007 --> 01:04:00,139
because this would mean
that whoever is the least safe,

1416
01:04:00,140 --> 01:04:03,186
whoever sacrifices the most
on safety to get ahead

1417
01:04:03,187 --> 01:04:05,188
will be the person
that gets there first.

1418
01:04:05,189 --> 01:04:07,886
That's basically
what's happening, right? -Yeah.

1419
01:04:07,887 --> 01:04:11,368
You and your brother
famously left OpenAI, uh,

1420
01:04:11,369 --> 01:04:12,978
to start Anthropic.

1421
01:04:12,979 --> 01:04:15,111
And then Anthropic
started because

1422
01:04:15,112 --> 01:04:17,896
some researchers
inside of OpenAI said,

1423
01:04:17,897 --> 01:04:20,551
"I want to go off
and do it more safely."

1424
01:04:20,552 --> 01:04:23,554
You needed something in addition
to just scaling the models up,

1425
01:04:23,555 --> 01:04:25,861
which is alignment or safety.

1426
01:04:25,862 --> 01:04:27,471
"We are more responsible

1427
01:04:27,472 --> 01:04:29,690
or more trustworthy
or more moral."

1428
01:04:29,691 --> 01:04:31,562
Now you have three AGI projects.

1429
01:04:31,563 --> 01:04:33,085
But also sitting around
the table with you

1430
01:04:33,086 --> 01:04:34,870
are gonna be a bunch of AIs.

1431
01:04:34,871 --> 01:04:37,568
And now Meta is-is
trying to do stuff.

1432
01:04:37,569 --> 01:04:39,613
Meanwhile, a new artificial
intelligence competitor

1433
01:04:39,614 --> 01:04:41,615
announced this week,
Elon Musk's...

1434
01:04:41,616 --> 01:04:44,662
xAI, which would be
Elon Musk's organization.

1435
01:04:44,663 --> 01:04:46,272
I don't trust OpenAI.

1436
01:04:46,273 --> 01:04:47,926
The fight between
Elon Musk and OpenAI

1437
01:04:47,927 --> 01:04:49,623
has entered a new round.

1438
01:04:49,624 --> 01:04:51,756
I don't trust Sam Altman,
uh, and I, and I don't think

1439
01:04:51,757 --> 01:04:54,237
we want to have the most
powerful AI in the world

1440
01:04:54,238 --> 01:04:56,456
controlled by someone
who is not trustworthy.

1441
01:04:58,459 --> 01:05:01,635
The incentive is
untold sums of money.

1442
01:05:01,636 --> 01:05:03,420
- Yes.
- Is untold power.

1443
01:05:03,421 --> 01:05:05,596
- Yes.
- Is untold control.

1444
01:05:05,597 --> 01:05:08,077
If you have something
that is a million times smarter

1445
01:05:08,078 --> 01:05:11,428
and more capable than
everything else on planet Earth

1446
01:05:11,429 --> 01:05:14,213
and no one else has that,

1447
01:05:14,214 --> 01:05:15,998
that thing is the incentive.

1448
01:05:15,999 --> 01:05:17,261
So you rule the world.

1449
01:05:18,610 --> 01:05:20,437
If you really believe this,

1450
01:05:20,438 --> 01:05:22,787
if you really in your heart
believe this,

1451
01:05:22,788 --> 01:05:24,223
then you might be
willing to take

1452
01:05:24,224 --> 01:05:26,138
quite a lot of risk
to make that happen.

1453
01:05:26,139 --> 01:05:28,924
Google has just released
its newest AI model.

1454
01:05:28,925 --> 01:05:31,274
The answer to OpenAI's ChatGPT.

1455
01:05:31,275 --> 01:05:33,102
How do we get to
this ten trillion?

1456
01:05:33,103 --> 01:05:35,234
Is NVIDIA becoming the most
valuable company in the US?

1457
01:05:35,235 --> 01:05:39,021
This is the largest business
opportunity in history.

1458
01:05:39,022 --> 01:05:40,544
In history.

1459
01:05:40,545 --> 01:05:42,285
So, the reason why
everyone's really hyped

1460
01:05:42,286 --> 01:05:44,722
about artificial intelligence
right now

1461
01:05:44,723 --> 01:05:48,073
is because
the more these companies hype

1462
01:05:48,074 --> 01:05:50,075
the potential capabilities
of their technology,

1463
01:05:50,076 --> 01:05:53,252
the more investment
they can attract.

1464
01:05:53,253 --> 01:05:54,775
Amazon investing up to

1465
01:05:54,776 --> 01:05:57,039
four billion dollars
in start-up Anthropic.

1466
01:05:57,040 --> 01:05:58,736
OpenAI is setting its sights

1467
01:05:58,737 --> 01:06:00,781
on a blockbuster half
a trillion dollar valuation.

1468
01:06:00,782 --> 01:06:02,044
Half a trillion!

1469
01:06:02,045 --> 01:06:03,262
The race is on.

1470
01:06:03,263 --> 01:06:04,655
This is happening
faster than ever.

1471
01:06:04,656 --> 01:06:06,657
Are we in an AI bubble?
Of course.

1472
01:06:06,658 --> 01:06:09,225
I just don't see
the bubble bursting

1473
01:06:09,226 --> 01:06:12,489
while you still have
this major spending cycle.

1474
01:06:12,490 --> 01:06:15,057
Even if you think
this is all hype,

1475
01:06:15,058 --> 01:06:17,624
there are billions
to trillions of dollars

1476
01:06:17,625 --> 01:06:21,150
flowing into making AI systems
more powerful.

1477
01:06:21,151 --> 01:06:23,891
And once you have that thing
that's more powerful,

1478
01:06:23,892 --> 01:06:25,589
companies can use that

1479
01:06:25,590 --> 01:06:27,591
to get bigger profits
and to make more money.

1480
01:06:27,592 --> 01:06:30,159
Countries can use that
to make stronger militaries.

1481
01:06:30,160 --> 01:06:32,857
NVIDIA has overtaken
Microsoft and Apple

1482
01:06:32,858 --> 01:06:36,295
to become the world's
most valuable company.

1483
01:06:36,296 --> 01:06:39,037
The race to deploy becomes
the race to recklessness,

1484
01:06:39,038 --> 01:06:41,822
because they can't
deploy it that quickly

1485
01:06:41,823 --> 01:06:43,215
and also get it right.

1486
01:06:43,216 --> 01:06:46,088
They believe that
they're the good guys.

1487
01:06:46,089 --> 01:06:49,352
"And if I don't do it,
somebody who doesn't have

1488
01:06:49,353 --> 01:06:51,615
"as good values as me
will be sitting at the table

1489
01:06:51,616 --> 01:06:54,487
getting to make decisions, so
I have an obligation to do it."

1490
01:06:54,488 --> 01:06:56,228
Yes, this is
a very commonly held belief.

1491
01:06:56,229 --> 01:06:58,622
Many, many, many,
maybe most of the people

1492
01:06:58,623 --> 01:07:00,754
building this technology
believe that.

1493
01:07:00,755 --> 01:07:03,757
I'm worried about
the commercial competition,

1494
01:07:03,758 --> 01:07:05,890
but it turns out
I'm even more worried about

1495
01:07:05,891 --> 01:07:08,066
the geopolitical competition.

1496
01:07:08,067 --> 01:07:10,373
We were eight years behind
a year ago.

1497
01:07:10,374 --> 01:07:12,636
Now we're probably
less than one year behind.

1498
01:07:12,637 --> 01:07:15,726
Well, both Saudi Arabia
and the UAE have been racing

1499
01:07:15,727 --> 01:07:18,163
to set up data centers
and position themselves

1500
01:07:18,164 --> 01:07:20,426
as the dominant force in AI...

1501
01:07:20,427 --> 01:07:22,472
...to make South Korea
a global AI leader.

1502
01:07:22,473 --> 01:07:24,039
French President Emmanuel Macron

1503
01:07:24,040 --> 01:07:25,910
spoke a lot about
artificial intelligence...

1504
01:07:25,911 --> 01:07:27,477
Now with more on Israel's role

1505
01:07:27,478 --> 01:07:29,653
in the artificial intelligence
revolution...

1506
01:07:29,654 --> 01:07:33,135
Countries will be competing
for whose AI technologies

1507
01:07:33,136 --> 01:07:34,919
create the next generation
of industry.

1508
01:07:34,920 --> 01:07:38,053
The Chinese are insisting
that AI,

1509
01:07:38,054 --> 01:07:39,880
as being developed in China,

1510
01:07:39,881 --> 01:07:41,752
reinforce the core values

1511
01:07:41,753 --> 01:07:44,059
of the Chinese Communist Party
and the Chinese system.

1512
01:07:44,060 --> 01:07:48,759
America has to beat China
in the AI race.

1513
01:07:48,760 --> 01:07:51,066
China's, like,
light-years behind.

1514
01:07:51,067 --> 01:07:52,719
Are they?

1515
01:07:52,720 --> 01:07:55,026
I mean, they have way more
training data than we do,

1516
01:07:55,027 --> 01:07:57,855
and there's nothing saying they
don't drop a model next month

1517
01:07:57,856 --> 01:08:00,988
that isn't--
doesn't far outperform GPT-4.

1518
01:08:00,989 --> 01:08:03,861
Everything was going just fine.
What could go wrong?

1519
01:08:03,862 --> 01:08:06,733
- DeepSeek...
- DeepSeek... -DeepSeek...

1520
01:08:06,734 --> 01:08:08,083
There is a new model.

1521
01:08:08,084 --> 01:08:10,868
But from a Chinese lab
called DeepSeek.

1522
01:08:10,869 --> 01:08:14,001
Let's talk about DeepSeek,
because it is mind-blowing,

1523
01:08:14,002 --> 01:08:17,701
and it is shaking this
entire industry to its core.

1524
01:08:17,702 --> 01:08:19,442
The Trump administration
will ensure

1525
01:08:19,443 --> 01:08:23,098
that the most powerful
AI systems are built in the US

1526
01:08:23,099 --> 01:08:26,579
with American designed
and manufactured chips.

1527
01:08:26,580 --> 01:08:28,842
AI is China's
Apollo project.

1528
01:08:28,843 --> 01:08:30,235
The Chinese Communist Party

1529
01:08:30,236 --> 01:08:31,584
deeply understands the potential

1530
01:08:31,585 --> 01:08:32,933
for AI to disrupt warfare.

1531
01:08:32,934 --> 01:08:35,371
Google no longer promises
that it will not

1532
01:08:35,372 --> 01:08:38,461
use artificial intelligence
for weapons or surveillance.

1533
01:08:38,462 --> 01:08:40,071
China, North Korea, Russia

1534
01:08:40,072 --> 01:08:41,942
are gonna keep building it
as fast as possible

1535
01:08:41,943 --> 01:08:44,554
to get more economic advantage,
more productivity advantage,

1536
01:08:44,555 --> 01:08:46,860
more scientific advantage,
more military advantage,

1537
01:08:46,861 --> 01:08:48,688
'cause AI makes better weapons.

1538
01:08:48,689 --> 01:08:51,778
If you're talking about
a system as broad and capable

1539
01:08:51,779 --> 01:08:53,345
as a brilliant scientist,

1540
01:08:53,346 --> 01:08:58,002
it might be able to run
a military campaign

1541
01:08:58,003 --> 01:09:01,484
better than any of the generals
in the US government right now.

1542
01:09:01,485 --> 01:09:05,531
Taiwan is the issue creating
the most tension right now

1543
01:09:05,532 --> 01:09:07,925
between Beijing and the US.

1544
01:09:07,926 --> 01:09:10,493
There's a-a direct
oppositional disagreement

1545
01:09:10,494 --> 01:09:12,016
between the US and China

1546
01:09:12,017 --> 01:09:14,018
on whether Taiwan is
part of China.

1547
01:09:14,019 --> 01:09:18,457
Taiwan is important
for so many reasons.

1548
01:09:18,458 --> 01:09:20,590
It is also the home of

1549
01:09:20,591 --> 01:09:24,202
about 90% of advanced chip
manufacturing for the world,

1550
01:09:24,203 --> 01:09:28,424
which means that the supply
chain for advanced compute

1551
01:09:28,425 --> 01:09:31,078
is at risk should there be

1552
01:09:31,079 --> 01:09:32,993
some sort of scenario
around Taiwan,

1553
01:09:32,994 --> 01:09:38,042
whether a-a blockade
or an invasion by China.

1554
01:09:38,043 --> 01:09:40,740
Are we going to be in some race

1555
01:09:40,741 --> 01:09:45,658
between the US and China
that eventually devolves into

1556
01:09:45,659 --> 01:09:49,401
a militarized AI arms race and,
you know, potentially leads

1557
01:09:49,402 --> 01:09:51,316
to some great power conflict?

1558
01:09:51,317 --> 01:09:53,057
The outgoing
top US military commander

1559
01:09:53,058 --> 01:09:55,886
in the region
predicted war was coming.

1560
01:09:55,887 --> 01:09:57,975
I think the threat is manifest
during this decade,

1561
01:09:57,976 --> 01:09:59,498
in fact, in the next six years.

1562
01:09:59,499 --> 01:10:01,457
Um, one of the scenarios

1563
01:10:01,458 --> 01:10:05,896
that I worry about is
a cyber flash war,

1564
01:10:05,897 --> 01:10:10,944
um, one in which
cyber tools that are autonomous

1565
01:10:10,945 --> 01:10:12,990
are competing against each other

1566
01:10:12,991 --> 01:10:14,644
in ways that are escalatory

1567
01:10:14,645 --> 01:10:16,386
without
meaningful human control.

1568
01:10:18,823 --> 01:10:20,476
You live in a war zone,

1569
01:10:20,477 --> 01:10:23,827
it will be an AI deciding
whether to bomb your house

1570
01:10:23,828 --> 01:10:25,785
and whether to kill you.

1571
01:10:25,786 --> 01:10:27,570
Let's have the machine decide.

1572
01:10:27,571 --> 01:10:29,398
That's the temptation.

1573
01:10:29,399 --> 01:10:32,618
That's why AI is so tempting
for militaries to adopt,

1574
01:10:32,619 --> 01:10:35,795
to create autonomous weapons,
because if I start believing

1575
01:10:35,796 --> 01:10:39,190
that my military adversary
is gonna adopt AI,

1576
01:10:39,191 --> 01:10:41,845
it'll be a race for who can
pull the trigger faster.

1577
01:10:41,846 --> 01:10:44,935
And the one that automates
that decision,

1578
01:10:44,936 --> 01:10:47,242
rather than having
a human in the loop,

1579
01:10:47,243 --> 01:10:49,244
is the one
that will win that war.

1580
01:10:49,245 --> 01:10:50,680
And if you think about

1581
01:10:50,681 --> 01:10:53,074
the nuclear arms race
in the 1940s...

1582
01:10:54,554 --> 01:10:56,599
...you know the Germans
are working on the bomb,

1583
01:10:56,600 --> 01:11:00,777
so it's not so easy to tell
Robert Oppenheimer then,

1584
01:11:00,778 --> 01:11:03,823
"Uh, y-you know, slow down."

1585
01:11:03,824 --> 01:11:05,347
So after watching this movie,
it's gonna be confusing,

1586
01:11:05,348 --> 01:11:06,696
'cause you're gonna go back,

1587
01:11:06,697 --> 01:11:08,350
and tomorrow
you're gonna use ChatGPT,

1588
01:11:08,351 --> 01:11:10,482
and it's gonna be
unbelievably helpful.

1589
01:11:10,483 --> 01:11:12,745
And I will use it, too, and
it'll be unbelievably helpful.

1590
01:11:12,746 --> 01:11:16,183
And you'll say, "Wait, so
I just saw this movie about AI

1591
01:11:16,184 --> 01:11:18,055
"and existential risk
and all these things,

1592
01:11:18,056 --> 01:11:20,666
and where's the threat again?"

1593
01:11:20,667 --> 01:11:24,757
And it's not that ChatGPT is
the existential threat.

1594
01:11:24,758 --> 01:11:28,718
It's that the race to deploy
the most powerful,

1595
01:11:28,719 --> 01:11:31,677
inscrutable,
uncontrollable technology,

1596
01:11:31,678 --> 01:11:35,159
under the worst
incentives possible,

1597
01:11:35,160 --> 01:11:37,031
that's the existential threat.

1598
01:11:38,119 --> 01:11:40,425
But it strikes me that-that

1599
01:11:40,426 --> 01:11:43,254
we are in a context of a race.

1600
01:11:43,255 --> 01:11:45,952
-Yes.
- And it is in a competitive,

1601
01:11:45,953 --> 01:11:47,693
"got to get there first,
got to win the race" setting,

1602
01:11:47,694 --> 01:11:49,042
"got to compete against China,

1603
01:11:49,043 --> 01:11:50,392
got to compete against
the other labs."

1604
01:11:50,393 --> 01:11:51,871
Isn't that right?

1605
01:11:51,872 --> 01:11:53,699
Today it is the case.

1606
01:11:53,700 --> 01:11:56,703
So we need to change
that race dynamic.

1607
01:11:57,922 --> 01:11:59,966
Don't you think?

1608
01:11:59,967 --> 01:12:02,187
I think that would be
very good indeed.

1609
01:12:03,710 --> 01:12:06,538
I-I think
there's some kind of...

1610
01:12:06,539 --> 01:12:09,585
mysticism around AI
that makes it feel like

1611
01:12:09,586 --> 01:12:11,761
it fell from the sky.

1612
01:12:11,762 --> 01:12:13,502
Thinking of this
as a godlike technology

1613
01:12:13,503 --> 01:12:15,460
is a problem.

1614
01:12:15,461 --> 01:12:16,897
Yeah.

1615
01:12:18,072 --> 01:12:19,943
Why?

1616
01:12:19,944 --> 01:12:23,033
It gives license
to the companies to...

1617
01:12:23,034 --> 01:12:26,297
not take responsibility

1618
01:12:26,298 --> 01:12:30,823
for the things that the software
that they built does.

1619
01:12:30,824 --> 01:12:35,132
If people made that connection,

1620
01:12:35,133 --> 01:12:37,569
maybe that would, like,
help them understand again

1621
01:12:37,570 --> 01:12:41,878
the dynamics of
all the money and power and...

1622
01:12:41,879 --> 01:12:45,360
chaos that's happening to...

1623
01:12:45,361 --> 01:12:47,144
create this technology.

1624
01:12:47,145 --> 01:12:49,276
It's, like, five guys
who control it, right?

1625
01:12:49,277 --> 01:12:51,670
- Like, five men?
- Basically, yeah.

1626
01:12:51,671 --> 01:12:53,890
- The CEOs?
- Yeah.

1627
01:12:53,891 --> 01:12:55,371
Okay.

1628
01:13:01,072 --> 01:13:03,116
Sometimes it feels like

1629
01:13:03,117 --> 01:13:07,294
I've been on this
just, like, endless journey

1630
01:13:07,295 --> 01:13:09,384
of trying to understand this.

1631
01:13:09,385 --> 01:13:12,388
You know, like
I'm climbing a mountain...

1632
01:13:15,608 --> 01:13:19,176
...and every time
I, like, get up a hill,

1633
01:13:19,177 --> 01:13:20,786
I think I've reached the top,

1634
01:13:20,787 --> 01:13:23,441
but it just keeps going
and going and going.

1635
01:13:24,574 --> 01:13:27,010
But from everything I've heard,

1636
01:13:27,011 --> 01:13:30,405
if I had to guess what's
at the top of Anxiety Mountain,

1637
01:13:30,406 --> 01:13:34,234
it's these five CEOs
from these five companies,

1638
01:13:34,235 --> 01:13:37,673
who are kind of like the
Oppenheimers of this moment.

1639
01:13:37,674 --> 01:13:39,501
These are the guys
who are building this thing.

1640
01:13:39,502 --> 01:13:40,850
Yeah.

1641
01:13:40,851 --> 01:13:44,201
Like, is there a plan?

1642
01:13:44,202 --> 01:13:46,464
The head of the company
that makes ChatGPT

1643
01:13:46,465 --> 01:13:49,989
warned of possible
significant harm to the world.

1644
01:13:49,990 --> 01:13:52,427
I think, if this technology
goes wrong,

1645
01:13:52,428 --> 01:13:53,515
it can go quite wrong.

1646
01:13:55,822 --> 01:13:57,823
Five dudes.

1647
01:13:57,824 --> 01:14:01,610
I-I never thought about us
as a social media company.

1648
01:14:01,611 --> 01:14:06,092
It-it feels like I have to try
and-and find these guys...

1649
01:14:06,093 --> 01:14:07,224
Mm-hmm.

1650
01:14:07,225 --> 01:14:09,009
...and get them in the movie.

1651
01:14:10,315 --> 01:14:12,622
Certainly we'd get
some clarity from that.

1652
01:14:16,582 --> 01:14:19,192
I mean, the buck's got to stop
somewhere, right?

1653
01:14:34,034 --> 01:14:35,339
How's that feel?

1654
01:14:35,340 --> 01:14:37,733
- Great.
- That okay?

1655
01:14:37,734 --> 01:14:39,474
Yeah.

1656
01:14:39,475 --> 01:14:41,040
- Can I move the seat forward?
-Yes, you can.

1657
01:14:41,041 --> 01:14:42,215
It's, uh, it's just
a bit awkward.

1658
01:14:42,216 --> 01:14:43,608
I'm a little, like...

1659
01:14:43,609 --> 01:14:45,088
Dario, how do you feel
about that chair?

1660
01:14:45,089 --> 01:14:47,046
Um, yeah, the chair's good.

1661
01:14:47,047 --> 01:14:48,874
I just wanted to move it
a little forward. -Okay.

1662
01:14:48,875 --> 01:14:50,572
I picked out that chair.

1663
01:14:50,573 --> 01:14:52,182
- It's a good chair.
- Thanks.

1664
01:14:57,884 --> 01:14:59,495
And just sit down there.

1665
01:15:01,409 --> 01:15:03,106
And then through here,

1666
01:15:03,107 --> 01:15:04,411
we'll be looking at each other
through this glass.

1667
01:15:11,419 --> 01:15:13,595
And sitting back in the chair,
leaning forward?

1668
01:15:13,596 --> 01:15:15,161
Yeah. Well, whatever's
comfortable, I think. -Okay.

1669
01:15:15,162 --> 01:15:16,728
It's kind of cool.
There's a bunch of mirrors.

1670
01:15:16,729 --> 01:15:17,860
- Is that-- okay.
-Sam A., take one. Mark.

1671
01:15:17,861 --> 01:15:19,557
- How's that?
- Good. Thank you.

1672
01:15:19,558 --> 01:15:21,080
The genesis
of this project is that I was

1673
01:15:21,081 --> 01:15:22,865
sitting at home,
and I was playing around with,

1674
01:15:22,866 --> 01:15:24,431
I think, your image generator,

1675
01:15:24,432 --> 01:15:27,304
and I was simultaneously,
uh, terrified

1676
01:15:27,305 --> 01:15:29,480
and really impressed.

1677
01:15:29,481 --> 01:15:31,917
- That is the usual combination.
- And then cut to

1678
01:15:31,918 --> 01:15:34,659
my wife and I find out
we're expecting.

1679
01:15:34,660 --> 01:15:38,315
And-and I'm, like, having
an existential crisis

1680
01:15:38,316 --> 01:15:40,273
as my wife is
six months pregnant.

1681
01:15:40,274 --> 01:15:42,972
And I think my first question,
which I ask everybody:

1682
01:15:42,973 --> 01:15:45,496
Is now a terrible time
to have a kid?

1683
01:15:45,497 --> 01:15:47,411
Am I making a big mistake?

1684
01:15:47,412 --> 01:15:49,065
I'm expecting a kid
in March, too. My first one.

1685
01:15:49,066 --> 01:15:50,545
- You're expecting a kid?
- Yeah.

1686
01:15:50,546 --> 01:15:51,850
- You're expecting in March?
- First kid.

1687
01:15:51,851 --> 01:15:53,330
- Mazel tov.
- Thank you very much.

1688
01:15:53,331 --> 01:15:54,505
I've never been so excited
for anything.

1689
01:15:54,506 --> 01:15:56,507
- That's how I feel.
- Yeah.

1690
01:15:56,508 --> 01:15:58,248
But you're not scared?

1691
01:15:58,249 --> 01:16:02,774
I mean, like, having a kid is
just this momentous thing

1692
01:16:02,775 --> 01:16:05,037
that I, you know,
I stay up every night, like,

1693
01:16:05,038 --> 01:16:06,865
reading these books about
how to raise a kid,

1694
01:16:06,866 --> 01:16:08,258
and I hope
I'm gonna do a good job,

1695
01:16:08,259 --> 01:16:09,302
and it feels
very overwhelming and--

1696
01:16:09,303 --> 01:16:12,088
but I'm not scared...

1697
01:16:12,089 --> 01:16:14,090
for kids to grow up
in a world with AI.

1698
01:16:14,091 --> 01:16:16,353
Like, that's... that'll be okay.

1699
01:16:16,354 --> 01:16:19,051
That is good to hear
coming from the guy.

1700
01:16:19,052 --> 01:16:20,792
I think it's a wonderful idea
to have kids.

1701
01:16:20,793 --> 01:16:23,273
I-I think they're the most
magical, incredible thing.

1702
01:16:23,274 --> 01:16:25,318
There's so much uncertainty,

1703
01:16:25,319 --> 01:16:26,885
I would almost just do
what you're gonna do anyway.

1704
01:16:26,886 --> 01:16:28,974
I know that's not
a very satisfying answer,

1705
01:16:28,975 --> 01:16:31,934
but it's the only one
I can come up with.

1706
01:16:31,935 --> 01:16:34,501
Our kids are never gonna...

1707
01:16:34,502 --> 01:16:37,113
know a world that doesn't have
really advanced AI.

1708
01:16:37,114 --> 01:16:39,898
In fact, our kids will never
be smarter than an AI.

1709
01:16:39,899 --> 01:16:42,118
"Our kids will
never be smarter than an AI"?

1710
01:16:42,119 --> 01:16:44,381
Well, from a raw,
from a raw IQ perspective,

1711
01:16:44,382 --> 01:16:46,296
they will never be
smarter than an AI.

1712
01:16:46,297 --> 01:16:48,080
That notion doesn't
unsettle you a little bit?

1713
01:16:48,081 --> 01:16:51,170
'Cause it makes me feel
a little queasy in a weird way.

1714
01:16:51,171 --> 01:16:54,826
Um, it does unsettle me
a little bit.

1715
01:16:54,827 --> 01:16:56,306
But it is reality.

1716
01:16:56,307 --> 01:16:58,308
Okay, so race dynamics.

1717
01:16:58,309 --> 01:17:01,050
We have a bunch of people
who are all in agreement

1718
01:17:01,051 --> 01:17:02,921
that this is scary.

1719
01:17:02,922 --> 01:17:04,793
Based on the timelines that
a lot of people have given me,

1720
01:17:04,794 --> 01:17:07,056
we have between two months and
five years to figure this out.

1721
01:17:07,057 --> 01:17:08,710
- Yeah.
- And-and, I guess,

1722
01:17:08,711 --> 01:17:12,845
my biggest question is:
Why can't we just stop?

1723
01:17:14,499 --> 01:17:19,111
The problem with, uh, uh,
"just stopping" is that, um,

1724
01:17:19,112 --> 01:17:22,245
there are many, many groups now
around the world, uh, uh,

1725
01:17:22,246 --> 01:17:25,727
building this, for
many nations, many companies,

1726
01:17:25,728 --> 01:17:28,164
um, all with
different motivations.

1727
01:17:28,165 --> 01:17:30,819
There are some of
these companies in this space

1728
01:17:30,820 --> 01:17:34,039
who their position is, "We want
to develop this technology

1729
01:17:34,040 --> 01:17:36,085
absolutely as fast as possible."

1730
01:17:36,086 --> 01:17:40,089
And even if we could pass laws
in the US and in Europe,

1731
01:17:40,090 --> 01:17:42,439
we need to convince...

1732
01:17:42,440 --> 01:17:45,007
Xi Jinping and Vladimir Putin
or, you know,

1733
01:17:45,008 --> 01:17:47,487
whoever their scientific
advisors are on their side.

1734
01:17:47,488 --> 01:17:49,272
That's gonna be really hard.

1735
01:17:49,273 --> 01:17:50,969
I think it is true

1736
01:17:50,970 --> 01:17:53,624
that if two people are
in exactly the same place,

1737
01:17:53,625 --> 01:17:55,931
uh, the one willing to take
more shortcuts on safety

1738
01:17:55,932 --> 01:17:58,107
should kind of
"get there first."

1739
01:17:58,108 --> 01:18:01,110
Uh, but...

1740
01:18:01,111 --> 01:18:02,981
we're able to use our lead

1741
01:18:02,982 --> 01:18:06,028
to spend a lot more time
doing safety testing.

1742
01:18:06,029 --> 01:18:08,030
And-and what if you lose it?

1743
01:18:08,031 --> 01:18:09,640
You get the call
and you find out

1744
01:18:09,641 --> 01:18:11,729
that you're now, let's say,
six months behind.

1745
01:18:11,730 --> 01:18:14,297
- Wh-What happens then?
- Depends who we're behind to.

1746
01:18:14,298 --> 01:18:16,865
If it's, like,
a adversarial government,

1747
01:18:16,866 --> 01:18:18,867
that's probably really bad.

1748
01:18:18,868 --> 01:18:21,391
So let's say you get the call
that China has a,

1749
01:18:21,392 --> 01:18:24,829
has a recursively
self-improving agent

1750
01:18:24,830 --> 01:18:27,223
or something like that that we
should be really worried about.

1751
01:18:27,224 --> 01:18:29,226
What do-- what-what happens?

1752
01:18:30,967 --> 01:18:32,403
Um...

1753
01:18:35,406 --> 01:18:37,276
That case would require...

1754
01:18:37,277 --> 01:18:40,715
the first step there would be
to talk to the US government.

1755
01:18:40,716 --> 01:18:42,978
Sam, do you trust the
government's ability to handle

1756
01:18:42,979 --> 01:18:44,676
something like this?

1757
01:18:45,895 --> 01:18:47,809
Yeah, I do, actually.

1758
01:18:47,810 --> 01:18:49,593
There's other things I don't
trust the government to handle,

1759
01:18:49,594 --> 01:18:51,203
but that particular scenario,
I think they would know...

1760
01:18:51,204 --> 01:18:52,814
they-they-- yes.

1761
01:18:52,815 --> 01:18:54,467
When you have,
like, private discussions

1762
01:18:54,468 --> 01:18:56,426
with the other guys whose
fingers are on the trigger,

1763
01:18:56,427 --> 01:18:59,646
so to speak, um,
do those private discussions

1764
01:18:59,647 --> 01:19:01,344
fill you with confidence

1765
01:19:01,345 --> 01:19:03,389
or do they make you,
uh, more anxious?

1766
01:19:03,390 --> 01:19:07,132
Look, I mean, I-I know some
of them better than others.

1767
01:19:07,133 --> 01:19:10,179
Um, I have more confidence
in some of them th-than I have

1768
01:19:10,180 --> 01:19:12,398
in others, you know,
as with, as with any people.

1769
01:19:12,399 --> 01:19:14,183
You know,
you think about, you know,

1770
01:19:14,184 --> 01:19:15,880
the kids in your high school
class or something, right?

1771
01:19:15,881 --> 01:19:18,317
You know, some of them are,
you know, really sweet.

1772
01:19:18,318 --> 01:19:20,450
Some of them are well-meaning
but not that effective.

1773
01:19:20,451 --> 01:19:22,408
Some of them are,
you know, bullies.

1774
01:19:22,409 --> 01:19:23,801
Some of them are
really bad people.

1775
01:19:23,802 --> 01:19:25,716
You-you kind of really,

1776
01:19:25,717 --> 01:19:27,109
you really see the spread.

1777
01:19:27,110 --> 01:19:28,850
You know, am I, am I,
am I confident

1778
01:19:28,851 --> 01:19:30,590
that everyone's gonna do
the right thing,

1779
01:19:30,591 --> 01:19:32,070
that it's all gonna work out?

1780
01:19:32,071 --> 01:19:33,898
Um, no, I'm not.

1781
01:19:33,899 --> 01:19:35,552
And there's-there's nothing
I can do about that, right?

1782
01:19:35,553 --> 01:19:37,815
You know, all-all I can do
is-is push for the,

1783
01:19:37,816 --> 01:19:39,208
push for the, you know,

1784
01:19:39,209 --> 01:19:40,818
push for the government
to get involved.

1785
01:19:40,819 --> 01:19:42,689
But ultimately I'm just
one person there, too.

1786
01:19:42,690 --> 01:19:44,169
That's--
it's-it's up to all of us

1787
01:19:44,170 --> 01:19:46,084
to push for the government
to get involved.

1788
01:19:46,085 --> 01:19:48,521
That's the number one thing
that-that I think we need to do

1789
01:19:48,522 --> 01:19:50,959
to-to, you know, to set things
in the right direction.

1790
01:19:50,960 --> 01:19:52,787
What makes me anxious
about that is, like,

1791
01:19:52,788 --> 01:19:56,225
the basic reality that the
speed at which the technology

1792
01:19:56,226 --> 01:19:58,444
is proliferating and growing
is exponential,

1793
01:19:58,445 --> 01:20:01,491
and the mechanisms to legislate
are 300 years old,

1794
01:20:01,492 --> 01:20:02,971
- take forever.
- Yeah.

1795
01:20:02,972 --> 01:20:05,147
Um, I-I think
it's gonna be a heavy lift.

1796
01:20:05,148 --> 01:20:06,931
I-I definitely agree
with you on that.

1797
01:20:06,932 --> 01:20:08,933
You know,
what I'm literally looking for

1798
01:20:08,934 --> 01:20:11,283
is-is like,
"Here are steps that, like,

1799
01:20:11,284 --> 01:20:15,200
"the head honchos are gonna
take to-to focus on safety,

1800
01:20:15,201 --> 01:20:18,160
to mitigate the peril
and maximize the promise."

1801
01:20:18,161 --> 01:20:20,162
And I don't,
I don't, I don't know

1802
01:20:20,163 --> 01:20:21,859
that there's
a simple answer to that.

1803
01:20:21,860 --> 01:20:23,296
Um...

1804
01:20:25,951 --> 01:20:27,604
I mean, this maybe is
too simple,

1805
01:20:27,605 --> 01:20:30,215
but you...

1806
01:20:30,216 --> 01:20:32,000
create a new model,

1807
01:20:32,001 --> 01:20:34,829
you study and test it
very carefully.

1808
01:20:34,830 --> 01:20:37,266
You put it out
into the world gradually,

1809
01:20:37,267 --> 01:20:39,572
and then, more and more,
you understand

1810
01:20:39,573 --> 01:20:41,270
if that's safe or not,
and then if it is,

1811
01:20:41,271 --> 01:20:42,707
you can take the next step.

1812
01:20:44,404 --> 01:20:47,145
It doesn't sound as flashy
as, like, a brilliant scientist

1813
01:20:47,146 --> 01:20:50,496
coming up with one idea in a
lab to make an AI system, like,

1814
01:20:50,497 --> 01:20:54,022
perfectly safe and controllable
and everything else,

1815
01:20:54,023 --> 01:20:55,762
but it is what I believe
is gonna happen.

1816
01:20:55,763 --> 01:20:57,329
Like, it is the way
I think this works.

1817
01:20:57,330 --> 01:20:59,331
But let's just say
something terrible happens,

1818
01:20:59,332 --> 01:21:01,856
like a model gets loose
or goes rogue or something.

1819
01:21:01,857 --> 01:21:03,509
Is there a protocol?

1820
01:21:03,510 --> 01:21:05,860
Like, literally,
I'm imagining a red phone.

1821
01:21:05,861 --> 01:21:07,165
- Yeah.
- Sorry for thinking of this

1822
01:21:07,166 --> 01:21:08,558
in terms of movies, but, like...

1823
01:21:08,559 --> 01:21:10,038
There is a protocol.

1824
01:21:10,039 --> 01:21:11,517
Is there a red phone
on your desk? -No.

1825
01:21:11,518 --> 01:21:13,258
Is it a secret?

1826
01:21:13,259 --> 01:21:15,434
I mean, uh, no, it's not,
it's not as fancy or dramatic

1827
01:21:15,435 --> 01:21:17,219
as you, like, would hope,
but there's, like, you know--

1828
01:21:17,220 --> 01:21:18,916
we've, like, thought through
these scenarios,

1829
01:21:18,917 --> 01:21:20,570
and if this happens,
we're gonna call these people

1830
01:21:20,571 --> 01:21:22,006
in this order and do this

1831
01:21:22,007 --> 01:21:23,486
and kind of make
these decisions if, like--

1832
01:21:23,487 --> 01:21:26,097
I do believe that
when you have an opportunity

1833
01:21:26,098 --> 01:21:29,100
to do your thinking before
a stressful situation happens,

1834
01:21:29,101 --> 01:21:30,449
that's almost always
a good idea.

1835
01:21:30,450 --> 01:21:31,798
And writing it down is helpful.

1836
01:21:31,799 --> 01:21:33,279
- Being prepared is helpful.
- Yeah.

1837
01:21:34,324 --> 01:21:36,064
You...

1838
01:21:36,065 --> 01:21:38,805
It would be impossible
for me to sit across from you

1839
01:21:38,806 --> 01:21:41,199
and-and ask you to promise me
that this is gonna go well?

1840
01:21:41,200 --> 01:21:43,898
That is impossible.

1841
01:21:43,899 --> 01:21:45,725
There isn't any easy answers, unfortunately.

1842
01:21:45,726 --> 01:21:48,119
Uh, because it's such
a cutting-edge technology,

1843
01:21:48,120 --> 01:21:49,816
um, there's still
a lot of unknowns.

1844
01:21:49,817 --> 01:21:52,297
And I think that
that-that needs to be,

1845
01:21:52,298 --> 01:21:55,083
um, uh, you know, understood

1846
01:21:55,084 --> 01:21:59,000
and-and hence the need
for, uh, uh, some caution.

1847
01:21:59,001 --> 01:22:00,915
I wake up, you know, every day,

1848
01:22:00,916 --> 01:22:03,482
this is the, this is the
number one thing I think about.

1849
01:22:03,483 --> 01:22:05,180
Now, look, I'm human,

1850
01:22:05,181 --> 01:22:08,400
and, you know, has-has
every decision been perfect?

1851
01:22:08,401 --> 01:22:11,664
Can I even say my motivations
were always perfectly clear?

1852
01:22:11,665 --> 01:22:13,579
Of course not.
No one can say that.

1853
01:22:13,580 --> 01:22:16,408
Like, that's-that's
just not, like, you know--

1854
01:22:16,409 --> 01:22:18,671
that's-that's just not
how people work.

1855
01:22:18,672 --> 01:22:21,065
The-the history of science
tends to be that,

1856
01:22:21,066 --> 01:22:23,589
for better or for worse,
if something's possible to do--

1857
01:22:23,590 --> 01:22:27,028
and we now know AI is possible
to do-- humanity does it.

1858
01:22:27,029 --> 01:22:29,769
All of this
was-was going to happen.

1859
01:22:29,770 --> 01:22:32,337
This-this train
isn't gonna stop.

1860
01:22:32,338 --> 01:22:34,296
You can't step in front of
the train and stop it.

1861
01:22:34,297 --> 01:22:36,298
You're just gonna get squished.

1862
01:22:36,299 --> 01:22:38,388
I mean, it's very stressful.

1863
01:22:39,606 --> 01:22:41,085
You know, there's, like,

1864
01:22:41,086 --> 01:22:42,695
a lot of things
a lot of us don't know.

1865
01:22:42,696 --> 01:22:46,395
I think the history
of scientific discovery is

1866
01:22:46,396 --> 01:22:48,310
one of not knowing
what you don't know

1867
01:22:48,311 --> 01:22:49,702
and figuring out as you go.

1868
01:22:49,703 --> 01:22:53,010
Uh, but, yeah, it is a...

1869
01:22:53,011 --> 01:22:56,622
it is a stressful way to live.

1870
01:22:56,623 --> 01:22:58,798
Right. Sam, thank you
very much for doing this.

1871
01:22:58,799 --> 01:23:00,539
- And again, mazel tov.
- Thank you. And to you.

1872
01:23:00,540 --> 01:23:02,715
Thank you so much. Thanks, guys.

1873
01:23:47,848 --> 01:23:49,327
Hello.

1874
01:23:49,328 --> 01:23:50,589
Hey, Dad, how are you?

1875
01:23:50,590 --> 01:23:51,634
Good. How you doing?

1876
01:23:51,635 --> 01:23:53,114
You working? What are you up to?

1877
01:23:53,115 --> 01:23:55,290
I'm working on this AI film.

1878
01:23:55,291 --> 01:23:57,857
And how's it going?

1879
01:23:57,858 --> 01:24:00,469
You know, it... it's really...

1880
01:24:00,470 --> 01:24:02,210
-Hi, sweetie.
-Hi, Mom.

1881
01:24:02,211 --> 01:24:03,907
So what is
the premise of the film?

1882
01:24:03,908 --> 01:24:06,562
Is it a documentary?
Kev, don't use any more spices.

1883
01:24:06,563 --> 01:24:08,346
It's already over-spiced.

1884
01:24:08,347 --> 01:24:10,479
We're making chicken right now.

1885
01:24:10,480 --> 01:24:12,655
- Is it a documentary or what...
-It's about--

1886
01:24:12,656 --> 01:24:14,004
the movie's about
the end of the world.

1887
01:24:14,005 --> 01:24:15,832
The end of the world's coming,

1888
01:24:15,833 --> 01:24:17,616
and we're making a movie
about the end of the world.

1889
01:24:17,617 --> 01:24:19,879
-Really?
-Yeah.

1890
01:24:19,880 --> 01:24:21,620
Kind of a depressing
film, it sounds like.

1891
01:24:21,621 --> 01:24:23,492
Yeah.

1892
01:24:23,493 --> 01:24:29,019
I'm feeling a lot,
like-- this very acute anxiety.

1893
01:24:29,020 --> 01:24:31,413
It's so scary,
but there's got to be--

1894
01:24:31,414 --> 01:24:35,330
you know, have you been meeting
some-some supersmart people

1895
01:24:35,331 --> 01:24:37,810
that are giving you any answers?

1896
01:24:37,811 --> 01:24:39,508
That's what's
frustrating about it.

1897
01:24:39,509 --> 01:24:41,379
No one knows.

1898
01:24:41,380 --> 01:24:43,947
All I can say to that is that

1899
01:24:43,948 --> 01:24:49,909
every generation has had
something scary like this.

1900
01:24:49,910 --> 01:24:52,260
When I was born, it was
the Cuban Missile Crisis.

1901
01:24:54,306 --> 01:24:56,394
I was just scared that there
was going to be a nuclear war.

1902
01:24:56,395 --> 01:24:57,787
Yeah, but we didn't know

1903
01:24:57,788 --> 01:24:59,136
what they were gonna do and...

1904
01:24:59,137 --> 01:25:00,572
And the world didn't end.

1905
01:25:00,573 --> 01:25:01,660
Everyone woke up
the next morning,

1906
01:25:01,661 --> 01:25:03,445
and we're still doing our thing.

1907
01:25:03,446 --> 01:25:05,011
I'm very scared,

1908
01:25:05,012 --> 01:25:06,883
especially in
the context of, like,

1909
01:25:06,884 --> 01:25:09,190
you know, the baby and...

1910
01:25:09,191 --> 01:25:10,582
It's gonna be a learning curve.

1911
01:25:10,583 --> 01:25:12,106
You're gonna be okay.

1912
01:25:12,107 --> 01:25:13,107
You can't,
you can't think about

1913
01:25:13,108 --> 01:25:14,369
what you can't control, Daniel.

1914
01:25:14,370 --> 01:25:16,197
Just remember that.

1915
01:25:16,198 --> 01:25:17,720
I'm really, I'm really feeling

1916
01:25:17,721 --> 01:25:19,243
nervous and scared about it.

1917
01:25:19,244 --> 01:25:21,071
There's so much
that I can't control.

1918
01:25:21,072 --> 01:25:23,900
Don't be nervous.
You can't let that get to you.

1919
01:25:23,901 --> 01:25:26,294
You can only control
what you can control,

1920
01:25:26,295 --> 01:25:27,947
and that's all you can do.

1921
01:25:27,948 --> 01:25:30,646
You can't do more than that.

1922
01:25:30,647 --> 01:25:32,649
Write that down in your book.

1923
01:25:49,274 --> 01:25:51,623
When you look back,

1924
01:25:51,624 --> 01:25:54,496
the world is always ending.

1925
01:25:54,497 --> 01:25:56,846
And when you look ahead,

1926
01:25:56,847 --> 01:25:59,283
the world is always ending.

1927
01:25:59,284 --> 01:26:02,199
...on fire.
One home is already on fire...

1928
01:26:02,200 --> 01:26:05,377
But the world
is always starting, too.

1929
01:26:07,988 --> 01:26:09,989
Are you ready?

1930
01:26:09,990 --> 01:26:11,992
Are you ever really ready?

1931
01:26:15,909 --> 01:26:17,693
You want to drive
or you want me to drive?

1932
01:26:17,694 --> 01:26:18,782
You drive.

1933
01:26:20,479 --> 01:26:21,871
Okay.

1934
01:26:21,872 --> 01:26:23,742
Thank you, Maria.

1935
01:26:23,743 --> 01:26:26,310
It's gonna just
feel really crampy.

1936
01:26:26,311 --> 01:26:27,833
-Okay.
-You're ready?

1937
01:26:27,834 --> 01:26:29,095
Yes.

1938
01:26:30,968 --> 01:26:32,404
You could be in a
medical drama with all that...

1939
01:26:49,116 --> 01:26:51,075
...some pressure.

1940
01:26:52,685 --> 01:26:54,078
Just relax.

1941
01:27:02,129 --> 01:27:04,957
Hi, buddy. Ooh.

1942
01:27:04,958 --> 01:27:06,481
Hi, buddy.

1943
01:27:31,202 --> 01:27:33,725
It's gonna be a whole new world

1944
01:27:33,726 --> 01:27:35,032
for you, Daniel.

1945
01:27:37,208 --> 01:27:40,080
And I think you're gonna be
an amazing father.

1946
01:27:41,517 --> 01:27:44,388
That's for sure.

1947
01:27:44,389 --> 01:27:46,651
Oh.

1948
01:27:48,698 --> 01:27:50,787
You're gonna do a great job.

1949
01:27:51,788 --> 01:27:53,092
Kev, why are you crying?

1950
01:27:53,093 --> 01:27:54,659
I'm crying because...

1951
01:27:54,660 --> 01:27:57,096
I don't know. I just don't...

1952
01:27:57,097 --> 01:27:59,360
Dad, you're gonna
make me cry, too.

1953
01:27:59,361 --> 01:28:00,710
Why are you crying?

1954
01:28:03,190 --> 01:28:05,540
I just know that
you're gonna be an amazing dad.

1955
01:28:05,541 --> 01:28:07,150
I'm only gonna be a great dad

1956
01:28:07,151 --> 01:28:09,718
'cause I had a great dad.

1957
01:28:09,719 --> 01:28:12,329
I'm getting emotional,
that's all.

1958
01:28:12,330 --> 01:28:13,722
Aw.

1959
01:28:14,941 --> 01:28:16,594
Boy, oh, boy!

1960
01:28:16,595 --> 01:28:17,857
My goodness!

1961
01:28:18,902 --> 01:28:24,820
Look at those little cheekies.
Look at those little cheekies.

1962
01:28:24,821 --> 01:28:26,387
Yeah.

1963
01:28:26,388 --> 01:28:28,433
I know how to end this movie.

1964
01:28:29,478 --> 01:28:32,306
Babies.

1965
01:28:32,307 --> 01:28:35,309
The end of the movie is about
babies.-

1966
01:28:35,310 --> 01:28:37,136
They're life-affirming.

1967
01:28:37,137 --> 01:28:38,964
They're exhausting.

1968
01:28:38,965 --> 01:28:40,836
They're hilarious.

1969
01:28:40,837 --> 01:28:42,707
And they're worth it.

1970
01:28:42,708 --> 01:28:46,494
This film isn't about the
inner workings of a technology.

1971
01:28:46,495 --> 01:28:48,278
It's not about
the billionaire CEOs.

1972
01:28:48,279 --> 01:28:50,236
It's not about the geopolitics.

1973
01:28:50,237 --> 01:28:53,631
It's not about the terrifying
future or the end of the world,

1974
01:28:53,632 --> 01:28:55,720
because my world
is just starting.

1975
01:28:55,721 --> 01:28:57,418
I'm building a crib.

1976
01:28:57,419 --> 01:28:59,333
Right here, right now.

1977
01:29:00,509 --> 01:29:02,379
AI is gonna change everything

1978
01:29:02,380 --> 01:29:05,643
in ways too powerful and
complex for us to understand.

1979
01:29:05,644 --> 01:29:08,952
And the future is not
for any of us to decide.

1980
01:29:10,736 --> 01:29:13,651
But what I can decide is to be
the best possible husband

1981
01:29:13,652 --> 01:29:17,742
for my wife and the best
possible dad for my son.

1982
01:29:17,743 --> 01:29:21,442
So whether our AI future is
a nightmarish dystopia

1983
01:29:21,443 --> 01:29:23,879
or the utopia
that we all dream of,

1984
01:29:23,880 --> 01:29:26,925
I'll at least know
that I did everything I could

1985
01:29:26,926 --> 01:29:30,146
to guide my family
through this AI revolution.

1986
01:29:30,147 --> 01:29:33,889
And no matter what,
we'll be facing it together.

1987
01:29:36,849 --> 01:29:40,548
So that's just our first idea.

1988
01:29:40,549 --> 01:29:42,854
How does this feel?
Are you feeling this?

1989
01:29:42,855 --> 01:29:45,291
Wait, I-- like, this is not--
this is a joke.

1990
01:29:45,292 --> 01:29:47,425
It's not actually
how you're gonna end it.

1991
01:29:50,036 --> 01:29:52,690
I mean, it's just an idea. Okay?

1992
01:29:52,691 --> 01:29:54,431
No, Daniel.

1993
01:29:54,432 --> 01:29:56,172
...uh, very, very dumb.

1994
01:29:56,173 --> 01:29:58,957
You've just spent, I don't
know, like, how many years

1995
01:29:58,958 --> 01:30:02,613
of our life working on this,
talking to every leading expert

1996
01:30:02,614 --> 01:30:05,007
on the planet about the subject,

1997
01:30:05,008 --> 01:30:07,705
and you're gonna end it

1998
01:30:07,706 --> 01:30:10,839
with some, like,
kumbaya bullshit?

1999
01:30:10,840 --> 01:30:13,537
There's an asteroid
headed to Earth.

2000
01:30:13,538 --> 01:30:15,539
What do you do? Just...

2001
01:30:15,540 --> 01:30:17,672
hold hands
and hope it works out okay?

2002
01:30:17,673 --> 01:30:19,456
Absolutely not.

2003
01:30:19,457 --> 01:30:22,328
We have to-- it ha--
it has to be...

2004
01:30:22,329 --> 01:30:24,419
The ending has to...

2005
01:30:25,855 --> 01:30:27,943
Okay, first thing:

2006
01:30:27,944 --> 01:30:31,468
AI is here,
and it's here to stay.

2007
01:30:31,469 --> 01:30:32,948
The shit's out of the horse,

2008
01:30:32,949 --> 01:30:35,430
but the horse is
gonna keep shitting.

2009
01:30:39,695 --> 01:30:41,609
You know, one of the basic
laws of history

2010
01:30:41,610 --> 01:30:44,046
is that nothing
really have a beginning

2011
01:30:44,047 --> 01:30:46,004
and nothing has any ending.

2012
01:30:46,005 --> 01:30:47,745
It just goes on.

2013
01:30:47,746 --> 01:30:50,574
AI is nowhere near
its full development.

2014
01:30:50,575 --> 01:30:52,837
Even if
the current AI bubble bursts,

2015
01:30:52,838 --> 01:30:54,622
humans are never going to stop

2016
01:30:54,623 --> 01:30:57,581
building more and more
powerful technology.

2017
01:30:57,582 --> 01:30:59,453
You can choose not to use AI

2018
01:30:59,454 --> 01:31:02,281
or participate in it, but it's
going to affect you anyway.

2019
01:31:02,282 --> 01:31:04,066
Okay, fantastic.
So we're screwed.

2020
01:31:04,067 --> 01:31:06,635
No, we're not
because of one simple thing.

2021
01:31:10,943 --> 01:31:13,031
This is not inevitable.

2022
01:31:13,032 --> 01:31:15,251
If we could just see it
clearly together,

2023
01:31:15,252 --> 01:31:18,384
the obvious response will be
to choose something different.

2024
01:31:18,385 --> 01:31:21,083
We need to very clearly
change the game

2025
01:31:21,084 --> 01:31:23,172
from a race to the bottom
into a race to the top.

2026
01:31:23,173 --> 01:31:27,481
The problem we need to
solve is not AI specifically.

2027
01:31:27,482 --> 01:31:31,267
It's the general question
of how do we build a society

2028
01:31:31,268 --> 01:31:33,922
that can deal with
powerful technology.

2029
01:31:33,923 --> 01:31:35,401
Because we're going to get

2030
01:31:35,402 --> 01:31:36,838
only more and more powerful
technology.

2031
01:31:36,839 --> 01:31:39,710
We need to upgrade our society,

2032
01:31:39,711 --> 01:31:41,538
and the first step is
coming together

2033
01:31:41,539 --> 01:31:43,409
- and demanding...
-Coordination.

2034
01:31:43,410 --> 01:31:47,065
Some form of international
cooperation or agreement about

2035
01:31:47,066 --> 01:31:48,545
what the norms should be.

2036
01:31:48,546 --> 01:31:49,938
You know, how should
they be deployed...

2037
01:31:49,939 --> 01:31:51,592
Like, real
international diplomacy

2038
01:31:51,593 --> 01:31:53,332
among the superpowers.

2039
01:31:53,333 --> 01:31:56,335
The Chinese are as worried
about it as the Americans.

2040
01:31:56,336 --> 01:31:57,772
I think it's difficult,
you know,

2041
01:31:57,773 --> 01:31:59,556
in the current
geopolitical climate,

2042
01:31:59,557 --> 01:32:01,297
- but I think it's necessary.
-Absolutely.

2043
01:32:01,298 --> 01:32:04,735
In the exact same way that
the last time that humanity

2044
01:32:04,736 --> 01:32:09,653
developed a technology
this dangerous...

2045
01:32:09,654 --> 01:32:12,438
...that required a complete,
unprecedented shift

2046
01:32:12,439 --> 01:32:15,572
to the structure of our world.

2047
01:32:15,573 --> 01:32:17,661
So we need to do that complete,

2048
01:32:17,662 --> 01:32:20,925
unprecedented shift again.

2049
01:32:20,926 --> 01:32:24,146
You know, we-we talk to people
who work at these AI companies,

2050
01:32:24,147 --> 01:32:25,364
and they say they want to do
something different,

2051
01:32:25,365 --> 01:32:26,670
but they need public pressure.

2052
01:32:26,671 --> 01:32:28,193
They need the government
to do something.

2053
01:32:28,194 --> 01:32:30,021
So then we go to DC,
and they say,

2054
01:32:30,022 --> 01:32:31,893
"Well, we need Silicon Valley
to do something different.

2055
01:32:31,894 --> 01:32:33,677
They're the ones who are gonna
come up with the guardrails."

2056
01:32:33,678 --> 01:32:35,723
And so everyone is pointing
the finger at someone else,

2057
01:32:35,724 --> 01:32:39,509
and what they agree on is
that we need public pressure

2058
01:32:39,510 --> 01:32:41,250
in order for something else
to happen.

2059
01:32:41,251 --> 01:32:42,860
And that's what you

2060
01:32:42,861 --> 01:32:44,296
and all the people
watching this movie can do.

2061
01:32:44,297 --> 01:32:45,341
We need to hold the leaders

2062
01:32:45,342 --> 01:32:47,125
in our governments

2063
01:32:47,126 --> 01:32:48,605
and the leaders of
these companies accountable.

2064
01:32:48,606 --> 01:32:51,042
Whichever country you're in,

2065
01:32:51,043 --> 01:32:52,914
let them know
that you're not happy

2066
01:32:52,915 --> 01:32:54,611
with the current status quo.

2067
01:32:54,612 --> 01:32:57,135
So, yeah, it's boring to say
call your congressperson.

2068
01:32:57,136 --> 01:32:59,398
I'm not saying you should
just do that, but, like,

2069
01:32:59,399 --> 01:33:00,965
we do have to do that.

2070
01:33:00,966 --> 01:33:02,445
Like, we do have to get
the government involved.

2071
01:33:02,446 --> 01:33:03,881
So we just
call them up and say,

2072
01:33:03,882 --> 01:33:06,014
"Hey, stop Big Tech
from ruining the world"?

2073
01:33:06,015 --> 01:33:08,930
No, but there are
tons of really obvious,

2074
01:33:08,931 --> 01:33:10,714
straightforward things
we can be demanding.

2075
01:33:10,715 --> 01:33:12,542
We need transparency.

2076
01:33:12,543 --> 01:33:14,283
We need to end the secrecy
that exists inside these labs,

2077
01:33:14,284 --> 01:33:16,677
because they are building
powerful technology,

2078
01:33:16,678 --> 01:33:18,722
and the public deserves to know
what's going on.

2079
01:33:18,723 --> 01:33:20,724
Ultimately,
we're gonna need independent,

2080
01:33:20,725 --> 01:33:23,292
objective third parties
to evaluate the systems.

2081
01:33:23,293 --> 01:33:27,078
We can't count on the companies
to grade their own homework.

2082
01:33:27,079 --> 01:33:31,213
If-if a company uses AI and-and
has AI interacting with you,

2083
01:33:31,214 --> 01:33:34,999
it should disclose that you are
interacting with an AI system.

2084
01:33:35,000 --> 01:33:38,655
Yeah. And-and also we need
a system that makes companies

2085
01:33:38,656 --> 01:33:41,876
legally liable for the
AI systems that they produce.

2086
01:33:41,877 --> 01:33:45,793
We need to make sure that there
are tests and safety standards

2087
01:33:45,794 --> 01:33:48,056
that are applied to everyone.

2088
01:33:48,057 --> 01:33:49,753
We need some ground rules,

2089
01:33:49,754 --> 01:33:51,581
and we need to keep adapting
those rules

2090
01:33:51,582 --> 01:33:54,628
at the speed that
the technology develops.

2091
01:33:54,629 --> 01:33:57,195
There is currently
more regulation

2092
01:33:57,196 --> 01:33:59,328
on selling a sandwich
to the public

2093
01:33:59,329 --> 01:34:02,940
than there is on building
potentially world-ending AGI.

2094
01:34:02,941 --> 01:34:04,725
And the last thing is

2095
01:34:04,726 --> 01:34:07,641
to upgrade ourselves.

2096
01:34:07,642 --> 01:34:09,773
This is not, like, the job
of, like, the safety team

2097
01:34:09,774 --> 01:34:11,775
at any given lab, or the CEO.

2098
01:34:11,776 --> 01:34:13,342
So, like, this is
everyone's job.

2099
01:34:13,343 --> 01:34:14,822
Don't, like, leave it
up to the AI experts.

2100
01:34:14,823 --> 01:34:16,998
Like, [bleep] that.
Like, this is the moment

2101
01:34:16,999 --> 01:34:19,304
that we are transitioning
from, like,

2102
01:34:19,305 --> 01:34:21,567
mostly human cognitive power
to, like, AI cognitive power,

2103
01:34:21,568 --> 01:34:23,134
and it affects everyone,

2104
01:34:23,135 --> 01:34:24,875
and I want people to be
in on that conversation.

2105
01:34:24,876 --> 01:34:26,921
And I would say, if you think
that AI will kill us all,

2106
01:34:26,922 --> 01:34:28,531
you should be working
in AI research

2107
01:34:28,532 --> 01:34:30,272
to make sure it doesn't,
because you do have

2108
01:34:30,273 --> 01:34:32,230
an enormous amount of agency.

2109
01:34:32,231 --> 01:34:33,884
Whoever you are,

2110
01:34:33,885 --> 01:34:35,886
you are an expert
in your own industry,

2111
01:34:35,887 --> 01:34:39,194
in your own school,
in your own family,

2112
01:34:39,195 --> 01:34:42,632
and it's up to you
how AI is used in your life.

2113
01:34:42,633 --> 01:34:45,026
Whether you want to join
your school board

2114
01:34:45,027 --> 01:34:47,898
or-or whether you want to ask
your employer

2115
01:34:47,899 --> 01:34:49,421
how they're using
AI technologies,

2116
01:34:49,422 --> 01:34:51,989
like, all of us can do the work.

2117
01:34:51,990 --> 01:34:56,124
A lot of unions have been
pretty effective at, like,

2118
01:34:56,125 --> 01:34:59,083
determining how they want
to interact with these systems.

2119
01:34:59,084 --> 01:35:02,739
Nurses unions, teacher unions.

2120
01:35:02,740 --> 01:35:04,610
I would love
if parents everywhere

2121
01:35:04,611 --> 01:35:06,177
went to the AI companies
and said,

2122
01:35:06,178 --> 01:35:08,049
"How can you be better
on this?" Including us.

2123
01:35:08,050 --> 01:35:11,226
So, I founded Encode Justice
when I was 15 years old.

2124
01:35:11,227 --> 01:35:14,882
We are the world's first
and largest army of young people

2125
01:35:14,883 --> 01:35:18,407
fighting for human-centered
artificial intelligence.

2126
01:35:18,408 --> 01:35:21,062
It doesn't matter who you are,
even the smallest actions help,

2127
01:35:21,063 --> 01:35:23,978
and even conversation starting
is really, really valuable.

2128
01:35:23,979 --> 01:35:25,980
So my job could be, like,

2129
01:35:25,981 --> 01:35:27,982
tell Bubby Lila about this
at dinner?

2130
01:35:27,983 --> 01:35:29,810
Honestly, yeah,
that's part of it.

2131
01:35:29,811 --> 01:35:31,681
And we're gonna have to do

2132
01:35:31,682 --> 01:35:34,466
a lot of things that
we haven't even thought of yet.

2133
01:35:34,467 --> 01:35:36,077
People are gonna look at
anything that we've outlined

2134
01:35:36,078 --> 01:35:37,295
and say, "That's not enough."

2135
01:35:37,296 --> 01:35:38,993
What matters is that the forces

2136
01:35:38,994 --> 01:35:40,951
that are working
towards solutions

2137
01:35:40,952 --> 01:35:44,476
start to exceed the forces that
are working against solutions.

2138
01:35:44,477 --> 01:35:46,435
Making the world better
has always been hard.

2139
01:35:46,436 --> 01:35:47,871
It has never been easy.

2140
01:35:47,872 --> 01:35:50,004
Like, there have been
many shitty things

2141
01:35:50,005 --> 01:35:51,962
that have happened in history,
and we've had--

2142
01:35:51,963 --> 01:35:53,747
like, people have had
to deal with that,

2143
01:35:53,748 --> 01:35:56,837
and then they've risen up
and changed it.

2144
01:35:56,838 --> 01:36:00,362
There are an insane number
of challenges ahead of us,

2145
01:36:00,363 --> 01:36:04,148
but if we can get past them,

2146
01:36:04,149 --> 01:36:08,109
we can unlock a future beyond
our wildest imagination.

2147
01:36:08,110 --> 01:36:10,459
We have to come together

2148
01:36:10,460 --> 01:36:14,289
and find the path between
the promise and the peril.

2149
01:36:14,290 --> 01:36:16,422
We can't be pessimists
or optimists.

2150
01:36:17,423 --> 01:36:19,425
We have to become something new.

2151
01:36:21,732 --> 01:36:24,516
A friend of mine calls me
an apocaloptimist.

2152
01:36:24,517 --> 01:36:26,083
"Apocaloptimist"?

2153
01:36:26,084 --> 01:36:28,303
I think that might be
my new favorite word.

2154
01:36:28,304 --> 01:36:29,957
-MAN: Might be
the name of this movie.

2155
01:36:29,958 --> 01:36:31,306
It might be the name
of this movie. -

2156
01:36:31,307 --> 01:36:34,222
- "Apocaloptimist."
- "Apocaloptimist."

2157
01:36:34,223 --> 01:36:35,876
Yeah.

2158
01:36:35,877 --> 01:36:37,312
I don't believe in doom.

2159
01:36:37,313 --> 01:36:39,096
I believe in the spirit of life,

2160
01:36:39,097 --> 01:36:42,708
uh, and I believe in life is
about the capacity to act.

2161
01:36:42,709 --> 01:36:44,319
The capacity to relate,

2162
01:36:44,320 --> 01:36:46,365
the capacity to feel.

2163
01:36:51,022 --> 01:36:53,981
We have to double down
more and more

2164
01:36:53,982 --> 01:36:57,071
on those capacities
that we have as humans

2165
01:36:57,072 --> 01:37:00,161
that robotic systems
will never have.

2166
01:37:00,162 --> 01:37:04,339
It's time right now
to make those decisions about

2167
01:37:04,340 --> 01:37:07,516
how to guide it and support it
rather than dividing us.

2168
01:37:07,517 --> 01:37:10,301
It kind of sounds like
raising a kid.

2169
01:37:10,302 --> 01:37:12,477
That's what's up. Yeah.

2170
01:37:12,478 --> 01:37:14,523
AI may have
more raw intelligence

2171
01:37:14,524 --> 01:37:17,265
than our little human brains,

2172
01:37:17,266 --> 01:37:21,095
but we're so much more
than just our intelligence.

2173
01:37:21,096 --> 01:37:25,099
Intelligence is-is
the ability to solve problems.

2174
01:37:25,100 --> 01:37:28,320
Wisdom is the ability to know
which problems to solve.

2175
01:37:33,760 --> 01:37:34,935
It can go on your fridge.

2176
01:37:36,241 --> 01:37:38,155
So, don't give up.

2177
01:37:38,156 --> 01:37:39,809
Humanity has done

2178
01:37:39,810 --> 01:37:42,160
more difficult things than this
in its history.

2179
01:37:43,553 --> 01:37:46,381
It's just hard to convince
people that they should.

2180
01:37:54,912 --> 01:37:56,870
So, when I started
making this movie,

2181
01:37:56,871 --> 01:37:58,480
I would say that I was, like,

2182
01:37:58,481 --> 01:38:00,961
broadly a cynical asshole
about this whole thing.

2183
01:38:00,962 --> 01:38:03,137
Over the course
of making the film,

2184
01:38:03,138 --> 01:38:04,703
I've come to understand
that, like,

2185
01:38:04,704 --> 01:38:07,489
that's the only thing
we can't be.

2186
01:38:07,490 --> 01:38:09,404
...or anything else
you'd like to discuss?

2187
01:38:09,405 --> 01:38:11,406
No. I think we covered a lot.

2188
01:38:11,407 --> 01:38:13,060
- Yeah.
-Thank you so much.

2189
01:38:13,061 --> 01:38:15,540
Thanks.

2190
01:38:15,541 --> 01:38:16,933
This is a problem that's bigger

2191
01:38:16,934 --> 01:38:19,109
than any one person.

2192
01:38:19,110 --> 01:38:21,590
This will change the world in
ways that we don't understand.

2193
01:38:21,591 --> 01:38:23,505
That is all true.

2194
01:38:23,506 --> 01:38:25,594
Okay.

2195
01:38:25,595 --> 01:38:27,726
But what we do have agency over

2196
01:38:27,727 --> 01:38:30,120
is what we do about it.

2197
01:38:30,121 --> 01:38:32,601
As frontier AI grows
exponentially more capable...

2198
01:38:32,602 --> 01:38:34,472
And the reality

2199
01:38:34,473 --> 01:38:36,605
is that if we just decide
it's hopeless,

2200
01:38:36,606 --> 01:38:39,042
then it is hopeless.

2201
01:38:39,043 --> 01:38:40,783
...to put less stress
on planet Earth...

2202
01:38:40,784 --> 01:38:42,785
But if you decide

2203
01:38:42,786 --> 01:38:45,528
that you want to try...

2204
01:38:46,790 --> 01:38:48,138
...then you try.

2205
01:38:48,139 --> 01:38:49,967
And that's hard.

2206
01:38:51,273 --> 01:38:53,230
But you know what?

2207
01:38:53,231 --> 01:38:56,278
Big things seem impossible
before they actually happen.

2208
01:38:58,149 --> 01:39:01,456
But when they finally do happen,

2209
01:39:01,457 --> 01:39:05,286
it's because millions of people
took millions of actions

2210
01:39:05,287 --> 01:39:07,592
to make them happen.

2211
01:39:07,593 --> 01:39:10,334
And so...

2212
01:39:10,335 --> 01:39:12,424
we have to try.

2213
01:39:21,129 --> 01:39:23,565
There's too much at stake.

2214
01:39:37,232 --> 01:39:41,278
Look at the incredible changes
we've experienced and survived

2215
01:39:41,279 --> 01:39:42,888
from the Stone Age,

2216
01:39:42,889 --> 01:39:46,153
and yet even greater changes
are still to come.

2217
01:40:47,606 --> 01:40:50,217
♪ What will we do now? ♪

2218
01:40:50,218 --> 01:40:53,263
♪ We've lost it to trying ♪

2219
01:40:53,264 --> 01:40:55,135
♪ We've lost it ♪

2220
01:40:55,136 --> 01:40:56,746
♪ To trying ♪

2221
01:40:59,575 --> 01:41:02,272
♪ What will we do now? ♪

2222
01:41:02,273 --> 01:41:05,232
♪ We've lost it to trying ♪

2223
01:41:05,233 --> 01:41:07,016
♪ We've lost it ♪

2224
01:41:07,017 --> 01:41:08,714
♪ To trying ♪

2225
01:41:11,630 --> 01:41:14,154
♪ What can we say now? ♪

2226
01:41:14,155 --> 01:41:17,244
♪ Our mouths only lying ♪

2227
01:41:17,245 --> 01:41:18,680
♪ Our mouths ♪

2228
01:41:18,681 --> 01:41:20,683
♪ Only lying ♪

2229
01:41:25,122 --> 01:41:27,732
♪ What can we say now? ♪

2230
01:41:27,733 --> 01:41:30,866
♪ Our mouths only lying ♪

2231
01:41:30,867 --> 01:41:32,259
♪ Our mouths ♪

2232
01:41:32,260 --> 01:41:34,261
♪ Only lying ♪

2233
01:42:01,071 --> 01:42:03,768
♪ Give in and get out ♪

2234
01:42:03,769 --> 01:42:06,728
♪ We rise in the dying ♪

2235
01:42:06,729 --> 01:42:08,251
♪ We rise ♪

2236
01:42:08,252 --> 01:42:10,211
♪ In the dying ♪

2237
01:42:13,127 --> 01:42:15,824
♪ Give in and get out ♪

2238
01:42:15,825 --> 01:42:18,696
♪ We rise in the dying ♪

2239
01:42:18,697 --> 01:42:20,263
♪ We rise ♪

2240
01:42:20,264 --> 01:42:22,223
♪ In the dying ♪

2241
01:42:25,051 --> 01:42:27,705
♪ Give in and get out ♪

2242
01:42:27,706 --> 01:42:30,708
♪ We rise in the dying ♪

2243
01:42:30,709 --> 01:42:32,275
♪ We rise ♪

2244
01:42:32,276 --> 01:42:34,148
♪ In the dying ♪

2245
01:42:37,107 --> 01:42:39,761
♪ Give in and get out ♪

2246
01:42:39,762 --> 01:42:42,633
♪ We rise in the dying ♪

2247
01:42:42,634 --> 01:42:43,852
♪ We ♪

2248
01:43:22,761 --> 01:43:25,328
♪ Oh, oh, oh, oh, oh ♪

2249
01:43:25,329 --> 01:43:27,025
♪ Oh, oh, oh, oh ♪

2250
01:43:27,026 --> 01:43:28,462
♪ Oh, oh ♪

2251
01:43:28,463 --> 01:43:31,421
♪ Oh, oh, oh, oh, oh, oh ♪

2252
01:43:31,422 --> 01:43:32,988
♪ Oh, oh, oh, oh ♪

2253
01:43:32,989 --> 01:43:34,424
♪ Oh, oh ♪

2254
01:43:34,425 --> 01:43:37,079
♪ Oh, oh, oh, oh, oh, oh ♪

2255
01:43:37,080 --> 01:43:39,690
♪ What will we do now? ♪

2256
01:43:39,691 --> 01:43:42,693
♪ We've lost it to trying ♪

2257
01:43:42,694 --> 01:43:44,608
♪ We've lost it ♪

2258
01:43:44,609 --> 01:43:46,349
♪ To trying ♪

2259
01:43:46,350 --> 01:43:49,047
♪ Oh, oh, oh, oh, oh, oh ♪

2260
01:43:49,048 --> 01:43:51,746
♪ What will we do now? ♪

2261
01:43:51,747 --> 01:43:54,749
♪ We've lost it to trying ♪

2262
01:43:54,750 --> 01:43:56,620
♪ We've lost it ♪

2263
01:43:56,621 --> 01:43:58,448
♪ To trying. ♪