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
00:00:15,420 --> 00:00:17,609
The concept
of machines that can think

4
00:00:17,610 --> 00:00:21,569
and act like humans has
been around for centuries.

5
00:00:21,570 --> 00:00:24,119
Long before smartphones
or even electricity

6
00:00:24,120 --> 00:00:27,359
our ancestors Were inspired
by the world around them

7
00:00:27,360 --> 00:00:29,939
and the mysteries of life itself.

8
00:00:29,940 --> 00:00:33,003
We've been through these
changes before as humanity.

9
00:00:34,080 --> 00:00:36,089
Are you going to be afraid,

10
00:00:36,090 --> 00:00:38,669
or are you gonna see the opportunities?

11
00:00:38,670 --> 00:00:41,249
The more people have access to AI,

12
00:00:41,250 --> 00:00:45,599
the better it is going to be
for future generations to come.

13
00:00:45,600 --> 00:00:48,119
We should be treating AI

14
00:00:48,120 --> 00:00:50,879
like we do a beloved little child,

15
00:00:50,880 --> 00:00:55,880
or you wanna teach it how to
be a good citizen of the world.

16
00:00:57,000 --> 00:01:00,239
From ancient
myths to modern marvels,

17
00:01:00,240 --> 00:01:03,329
we'll delve into the stories
of the pioneers who dared

18
00:01:03,330 --> 00:01:06,059
to dream of thinking machines

19
00:01:06,060 --> 00:01:07,949
and the technological advancements

20
00:01:07,950 --> 00:01:11,387
that have brought us closer to
making that dream a reality.

21
00:01:11,388 --> 00:01:13,409
All technology

22
00:01:13,410 --> 00:01:15,213
has no conscience of its own.

23
00:01:16,410 --> 00:01:19,713
Whether it will become
a force for good or ill,

24
00:01:20,820 --> 00:01:22,229
depends on man.

25
00:01:22,230 --> 00:01:24,119
The story
of AI is a testament

26
00:01:24,120 --> 00:01:27,423
to human curiosity, creativity,

27
00:01:29,190 --> 00:01:31,190
and the relentless pursuit of knowledge.

28
00:01:40,230 --> 00:01:41,253
What is AI?

29
00:01:42,570 --> 00:01:45,843
Artificial intelligence or AI
is giving computers a brain.

30
00:01:46,770 --> 00:01:48,809
The dictionary defines it as the theory

31
00:01:48,810 --> 00:01:52,619
and development of computer
systems able to perform tasks

32
00:01:52,620 --> 00:01:55,289
that normally require human intelligence.

33
00:01:55,290 --> 00:01:57,929
Today you think of AI as ChatGPT,

34
00:01:57,930 --> 00:01:59,489
but what it really is,

35
00:01:59,490 --> 00:02:01,469
is a reasoning and planning system

36
00:02:01,470 --> 00:02:03,029
that we've never seen before.

37
00:02:03,030 --> 00:02:05,969
When it comes to AI, there
are opportunities everywhere.

38
00:02:05,970 --> 00:02:08,549
You know, we're kind of
in the wild, wild west,

39
00:02:08,550 --> 00:02:11,339
and it could be
opportunities to get funding

40
00:02:11,340 --> 00:02:14,849
with a good idea and a
compelling proof of concept.

41
00:02:14,850 --> 00:02:16,739
There are really two types

42
00:02:16,740 --> 00:02:20,249
of AI being used out in the world today.

43
00:02:20,250 --> 00:02:23,399
One is the larger, more generalized AI

44
00:02:23,400 --> 00:02:25,439
that we know of as Google Translate.

45
00:02:25,440 --> 00:02:28,079
And ChatGPT on the other side,

46
00:02:28,080 --> 00:02:31,019
there's we call automation AI that's built

47
00:02:31,020 --> 00:02:32,839
with what we call golden data,

48
00:02:32,840 --> 00:02:36,419
or data that is derived
specifically for that use case

49
00:02:36,420 --> 00:02:39,359
and it takes away pain
points for workflows.

50
00:02:39,360 --> 00:02:42,359
Typically in businesses or
in people's daily routines.

51
00:02:42,360 --> 00:02:46,139
You can pick anything
and build an idea from it.

52
00:02:46,140 --> 00:02:47,369
Somebody built a company

53
00:02:47,370 --> 00:02:52,370
that's using AI to grade baseball cards.

54
00:02:52,440 --> 00:02:54,719
So rather than having
an expert have to look

55
00:02:54,720 --> 00:02:57,689
and say this is mint or
near mint or very good,

56
00:02:57,690 --> 00:03:00,809
you could just take photos of 20 cards

57
00:03:00,810 --> 00:03:02,939
and it'll give you
ratings for all of them.

58
00:03:02,940 --> 00:03:05,369
AI is set to
change the world as we know it.

59
00:03:05,370 --> 00:03:07,559
The arrival of this new intelligence

60
00:03:07,560 --> 00:03:10,379
will profoundly change
our country and the world

61
00:03:10,380 --> 00:03:12,779
in ways we cannot fully understand.

62
00:03:12,780 --> 00:03:13,799
And none of us,

63
00:03:13,800 --> 00:03:16,769
including myself, and
frankly, anyone in this room,

64
00:03:16,770 --> 00:03:18,820
is prepared for the implications of this.

65
00:03:19,860 --> 00:03:22,137
So what
are the origins of AI?

66
00:03:25,110 --> 00:03:27,509
The ancient Greeks told
tales of Hephaestus,

67
00:03:27,510 --> 00:03:30,779
the blacksmith of the gods who
created mechanical servants

68
00:03:30,780 --> 00:03:33,899
and even a bronze automaton called Talos

69
00:03:33,900 --> 00:03:35,879
to guard the island of Crete.

70
00:03:35,880 --> 00:03:37,979
These myths weren't just flights of fancy,

71
00:03:37,980 --> 00:03:39,839
they reflected a deep fascination

72
00:03:39,840 --> 00:03:42,659
with the potential of artificial beings.

73
00:03:42,660 --> 00:03:46,289
While the Greeks didn't have
computers or algorithms,

74
00:03:46,290 --> 00:03:48,209
they laid the groundwork for AI

75
00:03:48,210 --> 00:03:51,899
by exploring the relationship
between humans and machines.

76
00:03:51,900 --> 00:03:54,479
They grappled with questions
about consciousness,

77
00:03:54,480 --> 00:03:57,569
creativity, and the very
nature of intelligence.

78
00:03:57,570 --> 00:04:00,539
Questions that still
resonate with us today,

79
00:04:00,540 --> 00:04:03,569
I question whether such
automatons can possess life.

80
00:04:03,570 --> 00:04:06,329
But how does one define life then?

81
00:04:06,330 --> 00:04:08,729
Perhaps it is the ability to reason.

82
00:04:08,730 --> 00:04:10,349
Fast forward
to the Middle Ages

83
00:04:10,350 --> 00:04:13,599
where ingenious inventors
crafted intricate clocks,

84
00:04:13,600 --> 00:04:15,269
water-powered automatons,

85
00:04:15,270 --> 00:04:17,879
and even programmable musical instruments.

86
00:04:17,880 --> 00:04:21,299
These mechanical marvels often
displayed in royal courts

87
00:04:21,300 --> 00:04:24,179
and town squares captivated
the public imagination

88
00:04:24,180 --> 00:04:27,359
and pushed the boundaries
of what seemed possible.

89
00:04:27,360 --> 00:04:30,226
One particularly fascinating
figure was Al-Jazari,

90
00:04:30,227 --> 00:04:32,939
a 12th century Arab inventor who designed

91
00:04:32,940 --> 00:04:35,549
and built a wide range of automata,

92
00:04:35,550 --> 00:04:38,099
including a programmable
musical instrument

93
00:04:38,100 --> 00:04:41,909
that could be considered an
early ancestor of the computer.

94
00:04:41,910 --> 00:04:45,299
These inventions demonstrated
the growing sophistication

95
00:04:45,300 --> 00:04:46,923
of mechanical engineering,

96
00:04:47,880 --> 00:04:50,009
and hinted at the potential for machines

97
00:04:50,010 --> 00:04:53,669
to perform increasingly complex tasks.

98
00:04:53,670 --> 00:04:56,699
The 19th century saw a
surge in interest in logic,

99
00:04:56,700 --> 00:04:59,699
and mathematics laying the
groundwork for the development

100
00:04:59,700 --> 00:05:01,289
of computer science.

101
00:05:01,290 --> 00:05:04,259
Thinkers like George
Boole and Ada Lovelace

102
00:05:04,260 --> 00:05:06,239
made groundbreaking contributions

103
00:05:06,240 --> 00:05:08,399
to the formalization of logic

104
00:05:08,400 --> 00:05:10,889
and the development of algorithms,

105
00:05:10,890 --> 00:05:13,859
paving the way for the digital age.

106
00:05:13,860 --> 00:05:17,609
Lovelace, often hailed as the
first computer programmer,

107
00:05:17,610 --> 00:05:18,989
recognized the potential

108
00:05:18,990 --> 00:05:22,319
of Charles Babbage's analytical engine,

109
00:05:22,320 --> 00:05:25,109
a mechanical general purpose computer

110
00:05:25,110 --> 00:05:27,929
to go beyond mere calculation
and manipulate symbols

111
00:05:27,930 --> 00:05:30,479
according to rules
hinting at the possibility

112
00:05:30,480 --> 00:05:32,669
of artificial intelligence.

113
00:05:32,670 --> 00:05:35,489
In the mid 1940s,

114
00:05:35,490 --> 00:05:38,579
two researchers at the
University of Chicago came up

115
00:05:38,580 --> 00:05:42,029
with the idea of trying
to mimic the brain,

116
00:05:42,030 --> 00:05:44,489
the neurons and interconnections,

117
00:05:44,490 --> 00:05:47,309
and they called the neural networks.

118
00:05:47,310 --> 00:05:51,569
Now, this idea, there was
really no computational power

119
00:05:51,570 --> 00:05:54,449
to execute anything significant,

120
00:05:54,450 --> 00:05:57,419
but this is the idea that over the years

121
00:05:57,420 --> 00:06:00,513
has become the basis of what AI is today.

122
00:06:01,440 --> 00:06:03,509
The mid 20th
century witnessed the birth

123
00:06:03,510 --> 00:06:04,893
of the computer age,

124
00:06:05,790 --> 00:06:08,669
a revolution ignited by
the work of visionaries

125
00:06:08,670 --> 00:06:10,049
like Alan Turing.

126
00:06:10,050 --> 00:06:12,516
The British mathematician
asked a simple question,

127
00:06:12,517 --> 00:06:14,279
"Can machines think?"

128
00:06:14,280 --> 00:06:17,699
This era marked a significant
shift in human history

129
00:06:17,700 --> 00:06:21,149
as the potential of machines
to perform complex calculations

130
00:06:21,150 --> 00:06:24,269
and tasks began to be realized.

131
00:06:24,270 --> 00:06:26,189
The groundwork laid during this period

132
00:06:26,190 --> 00:06:29,489
would go on to influence
countless aspects of modern life,

133
00:06:29,490 --> 00:06:33,269
from the way we communicate
to the way we solve problems.

134
00:06:33,270 --> 00:06:36,149
Turing, a brilliant
mathematician and codebreaker,

135
00:06:36,150 --> 00:06:38,819
played a pivotal role in
cracking the enigma code

136
00:06:38,820 --> 00:06:40,409
during World War II.

137
00:06:40,410 --> 00:06:42,689
It's the greatest
encryption device in history.

138
00:06:42,690 --> 00:06:45,882
The Germans use it for
all major communications.

139
00:06:49,230 --> 00:06:51,659
His work was
crucial in the allied victory,

140
00:06:51,660 --> 00:06:54,873
saving countless lives
and shortening the war.

141
00:06:55,800 --> 00:06:59,429
He laid the theoretical
foundations for modern computing,

142
00:06:59,430 --> 00:07:02,069
envisioning machines that
could perform any task

143
00:07:02,070 --> 00:07:03,633
given the right instructions.

144
00:07:05,100 --> 00:07:06,719
His ideas were revolutionary,

145
00:07:06,720 --> 00:07:08,879
proposing that a machine
could be programmed

146
00:07:08,880 --> 00:07:11,579
to carry out any computation
that a human could

147
00:07:11,580 --> 00:07:13,443
given enough time and resources.

148
00:07:15,330 --> 00:07:18,449
Moreover, his famous Turing test asks

149
00:07:18,450 --> 00:07:22,169
if a machine can mimic
human conversation so well

150
00:07:22,170 --> 00:07:26,249
that you can't tell if it's
a person or a computer.

151
00:07:26,250 --> 00:07:29,099
The Turing test challenges
us to consider what it means

152
00:07:29,100 --> 00:07:31,259
for a machine to think.

153
00:07:31,260 --> 00:07:33,749
Turing's vision of
intelligent machines continues

154
00:07:33,750 --> 00:07:36,779
to drive innovation and
exploration in the field,

155
00:07:36,780 --> 00:07:38,969
influencing everything from robotics

156
00:07:38,970 --> 00:07:40,953
to natural language processing.

157
00:07:46,860 --> 00:07:49,199
The field of artificial
intelligence as we know it

158
00:07:49,200 --> 00:07:51,419
was officially born in 1956

159
00:07:51,420 --> 00:07:53,339
at the Dartmouth Summer Research Project

160
00:07:53,340 --> 00:07:55,289
on artificial intelligence.

161
00:07:55,290 --> 00:07:58,049
This landmark conference
organized by John McCarthy,

162
00:07:58,050 --> 00:07:59,819
Marvin Minsky, Claude Shannon,

163
00:07:59,820 --> 00:08:02,699
and Nathaniel Rochester brought
together leading researchers

164
00:08:02,700 --> 00:08:05,493
to explore the potential
of thinking machines.

165
00:08:06,360 --> 00:08:08,129
Despite the initial enthusiasm,

166
00:08:08,130 --> 00:08:10,169
the road to artificial intelligence proved

167
00:08:10,170 --> 00:08:12,269
to be rockier than anticipated.

168
00:08:12,270 --> 00:08:16,649
The late 1960s and 1970s saw
a period of disillusionment

169
00:08:16,650 --> 00:08:19,199
and reduced funding for AI research

170
00:08:19,200 --> 00:08:22,053
often referred to as the AI winter.

171
00:08:22,980 --> 00:08:25,799
Funding was cut back significantly.

172
00:08:25,800 --> 00:08:29,009
And what was it that resulted in that,

173
00:08:29,010 --> 00:08:32,729
I think that a big part of it was hype.

174
00:08:32,730 --> 00:08:34,079
You know, there was a ton of hype

175
00:08:34,080 --> 00:08:38,069
of selling artificial
intelligence, reasoning,

176
00:08:38,070 --> 00:08:40,859
contextual awareness, so on and so forth,

177
00:08:40,860 --> 00:08:43,109
and the tools that were built at that time

178
00:08:43,110 --> 00:08:46,950
were nowhere near the
capability of of achieving that.

179
00:08:47,994 --> 00:08:50,489
The 1980s and 1990s witnessed

180
00:08:50,490 --> 00:08:52,109
a resurgence of AI,

181
00:08:52,110 --> 00:08:54,719
driven in part by the
rise of machine learning.

182
00:08:54,720 --> 00:08:56,999
This inspired by the brain's ability

183
00:08:57,000 --> 00:09:00,389
to learn from experience
involved training algorithms

184
00:09:00,390 --> 00:09:02,189
on vast amounts of data,

185
00:09:02,190 --> 00:09:05,849
allowing them to improve
their performance over time.

186
00:09:05,850 --> 00:09:07,679
The availability of larger data sets

187
00:09:07,680 --> 00:09:10,623
and more powerful computers
enabled significant progress.

188
00:09:13,350 --> 00:09:17,189
In the late nineties what we knew as AI

189
00:09:17,190 --> 00:09:19,619
was really the decision trees

190
00:09:19,620 --> 00:09:22,589
that were used to control at video game.

191
00:09:22,590 --> 00:09:25,743
The first video game we
did was called Soul Edge.

192
00:09:28,290 --> 00:09:31,949
Namco brought a ninja from Japan.

193
00:09:31,950 --> 00:09:35,699
We shouldn't doing moves
using motion capture.

194
00:09:35,700 --> 00:09:38,399
The main player would
move their controller,

195
00:09:38,400 --> 00:09:42,449
and then the AI system
in the game would decide

196
00:09:42,450 --> 00:09:45,989
which animation to play based on that.

197
00:09:45,990 --> 00:09:50,129
And then you have the non-player
characters or the NPCs

198
00:09:50,130 --> 00:09:52,949
and those would also make decisions

199
00:09:52,950 --> 00:09:54,963
based on what is decision trees.

200
00:09:55,980 --> 00:09:56,813
You win.

201
00:09:58,020 --> 00:10:00,479
Machine learning
began to outperform humans

202
00:10:00,480 --> 00:10:02,729
in specific tasks.

203
00:10:02,730 --> 00:10:05,189
In the late nineties,
AI scored a major win.

204
00:10:05,190 --> 00:10:06,659
IBM's Deep Blue defeated

205
00:10:06,660 --> 00:10:08,793
world chess champion Garry Kasparov,

206
00:10:09,990 --> 00:10:11,999
demonstrating the
potential of this approach

207
00:10:12,000 --> 00:10:14,463
to revolutionize various industries.

208
00:10:20,640 --> 00:10:23,429
Lift off of
space shuttle and lands.

209
00:10:23,430 --> 00:10:24,809
The 20th century witnessed

210
00:10:24,810 --> 00:10:27,629
the thrilling space race, a
competition between nations

211
00:10:27,630 --> 00:10:30,299
to achieve supremacy in space exploration.

212
00:10:30,300 --> 00:10:33,543
It was a time of rapid
advancements and intense rivalry.

213
00:10:34,380 --> 00:10:38,189
Today a new race is
underway, the AI space race.

214
00:10:38,190 --> 00:10:39,989
Nations are investing heavily

215
00:10:39,990 --> 00:10:42,299
in AI research and development.

216
00:10:42,300 --> 00:10:45,419
At the heart of this competition
is the drive to develop

217
00:10:45,420 --> 00:10:49,079
and control the most
advanced AI technologies,

218
00:10:49,080 --> 00:10:51,363
which promised to transform our world.

219
00:10:52,710 --> 00:10:56,519
Starting in the 2010s,
we had this convergence

220
00:10:56,520 --> 00:11:00,269
of cloud computing
becoming less expensive,

221
00:11:00,270 --> 00:11:03,359
and also the treasure trove of data

222
00:11:03,360 --> 00:11:06,869
that was actually accessible to developers

223
00:11:06,870 --> 00:11:10,199
and AI technologists from
social media, from the internet,

224
00:11:10,200 --> 00:11:12,479
all these things coming together,

225
00:11:12,480 --> 00:11:15,269
that's when we saw this
next level of explosion.

226
00:11:15,270 --> 00:11:17,459
This next level of development

227
00:11:17,460 --> 00:11:21,993
that took us into what we're
seeing now, exponential growth.

228
00:11:24,060 --> 00:11:27,629
The biggest impactful
discovery was really in 2017

229
00:11:27,630 --> 00:11:30,629
with the release of Transformer models,

230
00:11:30,630 --> 00:11:34,199
which essentially allowed
these neural models

231
00:11:34,200 --> 00:11:38,219
to create contextual lines
between the information

232
00:11:38,220 --> 00:11:39,779
that it's being trained with.

233
00:11:39,780 --> 00:11:44,609
And I think that context is central for AI

234
00:11:44,610 --> 00:11:46,649
to gain the ability to reason

235
00:11:46,650 --> 00:11:49,859
and to make the correct decisions.

236
00:11:49,860 --> 00:11:51,899
The stakes are
incredibly high as the nation

237
00:11:51,900 --> 00:11:55,139
that leads in AI will have
a significant advantage

238
00:11:55,140 --> 00:11:56,373
in the global arena.

239
00:11:57,210 --> 00:11:59,523
What would happen if China beat us?

240
00:12:00,420 --> 00:12:01,829
Let's think about it.

241
00:12:01,830 --> 00:12:05,609
The path to intelligence
that superhuman intelligence,

242
00:12:05,610 --> 00:12:08,429
think of the national
security implications

243
00:12:08,430 --> 00:12:09,869
of that competition.

244
00:12:09,870 --> 00:12:11,369
In January, 2025,

245
00:12:11,370 --> 00:12:15,539
the world witnessed a pivotal
moment in the AI space race.

246
00:12:15,540 --> 00:12:18,449
The launch of DeepSeek a Chinese AI app

247
00:12:18,450 --> 00:12:20,669
that quickly took the world by storm.

248
00:12:20,670 --> 00:12:23,699
This innovative application
redefined the boundaries

249
00:12:23,700 --> 00:12:25,949
of what AI could achieve in everyday life.

250
00:12:25,950 --> 00:12:27,479
DeepSeek showed up.

251
00:12:27,480 --> 00:12:28,589
Nobody expected this.

252
00:12:28,590 --> 00:12:32,369
It turns out it's on par now
with some of the top models.

253
00:12:32,370 --> 00:12:35,609
Welcome, China has arrived
in the competition.

254
00:12:35,610 --> 00:12:38,249
The DeepSeek movement is
gonna go down in history

255
00:12:38,250 --> 00:12:42,749
as one being one of reflection
for the entire AI community

256
00:12:42,750 --> 00:12:45,689
and every stakeholder at large.

257
00:12:45,690 --> 00:12:50,429
So DeepSeek was a model
released by a Chinese company

258
00:12:50,430 --> 00:12:52,349
that goes by the same name.

259
00:12:52,350 --> 00:12:56,429
It was an open source
model that was comparable

260
00:12:56,430 --> 00:12:59,853
to the best closed source models
that existed in the planet.

261
00:13:07,890 --> 00:13:11,309
The release of DeepSeek proved
that some smaller companies

262
00:13:11,310 --> 00:13:14,669
can compete with the
larger Goliath models.

263
00:13:14,670 --> 00:13:16,589
And what it really proved

264
00:13:16,590 --> 00:13:21,299
is that it's possible to do
more with less resources,

265
00:13:21,300 --> 00:13:25,769
and that the lag from having
a top of the line model

266
00:13:25,770 --> 00:13:29,609
is really a much faster depreciating asset

267
00:13:29,610 --> 00:13:32,339
than we thought that it was initially

268
00:13:32,340 --> 00:13:33,869
Developed by a consortium

269
00:13:33,870 --> 00:13:35,399
of Chinese tech companies,

270
00:13:35,400 --> 00:13:39,273
DeepSeek showcased China's rapid
progress in AI development.

271
00:13:40,140 --> 00:13:43,829
Utilizing a cost-effective
development model known as R1,

272
00:13:43,830 --> 00:13:47,339
DeepSeek achieved impressive
results with limited resources.

273
00:13:47,340 --> 00:13:50,429
We're not entirely
sure of how many times

274
00:13:50,430 --> 00:13:51,629
they may have gone through this,

275
00:13:51,630 --> 00:13:54,389
how much money they really
spent to get to the point

276
00:13:54,390 --> 00:13:58,019
where they were at in being
able to create a model

277
00:13:58,020 --> 00:14:00,659
with the $5 million,
whatever it was spent.

278
00:14:00,660 --> 00:14:01,919
The app quickly climbed

279
00:14:01,920 --> 00:14:04,439
to the top of the Google Play
and Apple app store charts,

280
00:14:04,440 --> 00:14:07,259
surpassing established US tech giants.

281
00:14:07,260 --> 00:14:10,979
Its rapid ascent was
nothing short of remarkable.

282
00:14:10,980 --> 00:14:13,559
This surge in popularity sent shockwaves

283
00:14:13,560 --> 00:14:15,059
through the tech world,

284
00:14:15,060 --> 00:14:18,449
signaling a shift in
the global AI landscape.

285
00:14:18,450 --> 00:14:20,579
Analysts and experts began to take notice

286
00:14:20,580 --> 00:14:22,511
of the changing dynamics.

287
00:14:22,512 --> 00:14:25,199
DeepSeek success served as a wake up call

288
00:14:25,200 --> 00:14:27,329
for the US and other nations.

289
00:14:27,330 --> 00:14:30,509
Highlighting the fierce competition in AI.

290
00:14:30,510 --> 00:14:32,969
The fact that this
model is now open source,

291
00:14:32,970 --> 00:14:36,569
the fact that anybody in the
world can build on top of it

292
00:14:36,570 --> 00:14:38,669
means that we should
just acknowledge the fact

293
00:14:38,670 --> 00:14:41,849
that if AI is gonna be in the
hands of every individual,

294
00:14:41,850 --> 00:14:43,949
we need to think of other mechanisms

295
00:14:43,950 --> 00:14:47,309
to make sure it's responsibly
being used and deployed.

296
00:14:47,310 --> 00:14:49,709
I think the importance of
what they were able to achieve

297
00:14:49,710 --> 00:14:54,710
is allowing open source
models to continue to compete,

298
00:14:54,990 --> 00:14:58,739
and that it ensures
that some of the things

299
00:14:58,740 --> 00:15:00,479
that AI is supposed to be great for

300
00:15:00,480 --> 00:15:02,939
in terms of improving humanity

301
00:15:02,940 --> 00:15:05,433
can be done without
being behind a paywall.

302
00:15:06,480 --> 00:15:08,279
The launch of
DeepSeek occurred amidst

303
00:15:08,280 --> 00:15:12,239
the surge in AI investment
and development in the US.

304
00:15:12,240 --> 00:15:16,049
Together, these world
leading technology giants era

305
00:15:16,050 --> 00:15:19,079
announcing the formation of Stargate,

306
00:15:19,080 --> 00:15:22,919
a new American company that
will invest $500 billion

307
00:15:22,920 --> 00:15:25,349
at least in AI infrastructure.

308
00:15:25,350 --> 00:15:27,449
The fact that it was coming out of China

309
00:15:27,450 --> 00:15:30,779
had a lot of geopolitical
implications around it as well.

310
00:15:30,780 --> 00:15:35,219
If you're using a model
that's also hosted in China,

311
00:15:35,220 --> 00:15:36,862
do they have access to a lot of the data

312
00:15:36,863 --> 00:15:39,659
that we are feeding into
these models unknowingly?

313
00:15:39,660 --> 00:15:43,559
China is a competitor and
others are competitors, we want,

314
00:15:43,560 --> 00:15:45,303
we want it to be in this country.

315
00:15:46,320 --> 00:15:48,539
Notable trends
included Google's Gemini

316
00:15:48,540 --> 00:15:52,593
and Microsoft's Copilot providing
more focused AI solutions.

317
00:15:53,640 --> 00:15:55,649
While the US and China are major players

318
00:15:55,650 --> 00:15:58,259
in the AI space race, this
is a global competition

319
00:15:58,260 --> 00:16:00,573
with contributions from
countries worldwide.

320
00:16:01,590 --> 00:16:03,959
Nations around the world
are investing heavily

321
00:16:03,960 --> 00:16:06,119
in AI research and development,

322
00:16:06,120 --> 00:16:08,549
each hoping to gain a competitive edge.

323
00:16:08,550 --> 00:16:10,829
This multipolar nature fosters a diverse

324
00:16:10,830 --> 00:16:13,139
and dynamic AI ecosystem.

325
00:16:13,140 --> 00:16:17,879
From country to country values differ.

326
00:16:17,880 --> 00:16:22,379
For example, China has its own values

327
00:16:22,380 --> 00:16:23,849
and its own considerations,

328
00:16:23,850 --> 00:16:26,729
so we will need to get on the same page.

329
00:16:26,730 --> 00:16:30,959
There will be, hopefully, new
treaties along those lines,

330
00:16:30,960 --> 00:16:34,979
but it is going to be a great challenge.

331
00:16:34,980 --> 00:16:36,989
The future of
AI depends on our ability

332
00:16:36,990 --> 00:16:39,359
to harness its transformative power

333
00:16:39,360 --> 00:16:41,639
while mitigating its risks.

334
00:16:41,640 --> 00:16:43,469
The choices we make today will determine

335
00:16:43,470 --> 00:16:45,299
whether AI becomes a force for good,

336
00:16:45,300 --> 00:16:47,043
or a source of new challenges.

337
00:16:48,210 --> 00:16:52,319
We are on the cusp of a disruption

338
00:16:52,320 --> 00:16:56,819
to human culture that happens very rarely.

339
00:16:56,820 --> 00:16:58,799
The kind of disruption we're talking about

340
00:16:58,800 --> 00:17:00,089
is the kind of disruption

341
00:17:00,090 --> 00:17:02,376
that the invention of farming created.

342
00:17:08,280 --> 00:17:10,499
This technology is in
some sense unstoppable.

343
00:17:10,500 --> 00:17:12,850
It's gonna become a part
of our everyday lives.

344
00:17:14,340 --> 00:17:17,039
The future is
closer than you might think.

345
00:17:17,040 --> 00:17:19,739
Picture a day waking up in a home powered

346
00:17:19,740 --> 00:17:23,433
by renewable energy, commuting
in an autonomous vehicle,

347
00:17:24,270 --> 00:17:26,133
receiving personalized healthcare.

348
00:17:27,600 --> 00:17:31,533
Let's travel to the year
2035 and visit NeoZenith.

349
00:17:34,350 --> 00:17:36,869
A shining example of a city where AI

350
00:17:36,870 --> 00:17:39,723
and humans coexist in perfect harmony.

351
00:17:42,690 --> 00:17:45,869
It's a testament to what can
be achieved when technology

352
00:17:45,870 --> 00:17:48,513
and nature are seamlessly integrated.

353
00:17:50,220 --> 00:17:53,519
Imagine streets where AI-driven
systems manage everything

354
00:17:53,520 --> 00:17:56,039
from traffic flow to energy consumption,

355
00:17:56,040 --> 00:17:58,803
ensuring that the city
operates at peak efficiency.

356
00:18:01,350 --> 00:18:03,123
This isn't some concrete jungle.

357
00:18:04,320 --> 00:18:07,019
NeoZenith is a breath
of fresh air, literally.

358
00:18:07,020 --> 00:18:09,239
The city planners have
gone to great lengths

359
00:18:09,240 --> 00:18:13,499
to incorporate green spaces
into every aspect of urban life.

360
00:18:13,500 --> 00:18:15,389
Parks, gardens and green rooftops

361
00:18:15,390 --> 00:18:17,339
are not just aesthetic choices,

362
00:18:17,340 --> 00:18:20,879
but essential components
of the city's ecosystem.

363
00:18:20,880 --> 00:18:24,299
These green spaces serve
as the lungs of NeoZenith,

364
00:18:24,300 --> 00:18:26,429
filtering pollutants
in providing residents

365
00:18:26,430 --> 00:18:27,723
with clean, fresh air.

366
00:18:29,250 --> 00:18:30,963
The city pulses with life.

367
00:18:31,830 --> 00:18:34,829
Public transportation is
efficient and eco-friendly,

368
00:18:34,830 --> 00:18:36,809
making it easy for everyone to get around

369
00:18:36,810 --> 00:18:38,495
without contributing to pollution.

370
00:18:38,496 --> 00:18:41,099
This
is Hikari Super Express

371
00:18:41,100 --> 00:18:42,623
bound to Shin-Osaka.

372
00:18:43,830 --> 00:18:45,209
It's organized, efficient,

373
00:18:45,210 --> 00:18:47,251
and designed with people at its heart.

374
00:18:49,950 --> 00:18:52,829
Wide pedestrian-friendly avenues are alive

375
00:18:52,830 --> 00:18:56,433
with people strolling, cycling,
and enjoying the fresh air.

376
00:18:57,360 --> 00:18:59,579
Gleaming solar panels are dawn rooftops,

377
00:18:59,580 --> 00:19:01,233
soaking up the sun's energy.

378
00:19:02,340 --> 00:19:05,369
Wind turbines hum gracefully
on the city's outskirts,

379
00:19:05,370 --> 00:19:07,709
generating clean electricity.

380
00:19:07,710 --> 00:19:09,899
But here's where AI comes in.

381
00:19:09,900 --> 00:19:12,029
It manages and distributes this energy

382
00:19:12,030 --> 00:19:13,979
with incredible efficiency.

383
00:19:13,980 --> 00:19:15,959
It predicts energy consumption patterns

384
00:19:15,960 --> 00:19:18,029
and ensures that every corner of NeoZenith

385
00:19:18,030 --> 00:19:20,583
has a constant supply of clean power.

386
00:19:21,840 --> 00:19:23,459
We could be looking at a future

387
00:19:23,460 --> 00:19:25,799
where we all have personal assistance

388
00:19:25,800 --> 00:19:30,089
that are doing things for us,
helping us make decisions,

389
00:19:30,090 --> 00:19:32,519
helping us be healthier people,

390
00:19:32,520 --> 00:19:34,745
helping us do creative tasks.

391
00:19:36,210 --> 00:19:39,269
Wake up in your
cozy NeoZenith apartment

392
00:19:39,270 --> 00:19:40,829
and your AI assistant greets you

393
00:19:40,830 --> 00:19:44,579
with a personalized
schedule and the day's news.

394
00:19:44,580 --> 00:19:46,799
Your AI has already
brewed your favorite blend

395
00:19:46,800 --> 00:19:48,050
just the way you like it.

396
00:19:49,620 --> 00:19:51,209
Heading to work?

397
00:19:51,210 --> 00:19:52,529
Autonomous vehicles whisk you

398
00:19:52,530 --> 00:19:54,089
through the city swiftly and safely,

399
00:19:54,090 --> 00:19:56,973
giving you precious time to
relax or catch up on emails.

400
00:19:59,250 --> 00:20:02,339
Remember those traffic
clogged streets of the past,

401
00:20:02,340 --> 00:20:05,039
the honking horns, the
endless lines of cars,

402
00:20:05,040 --> 00:20:07,953
and the frustration of being
stuck in a jam for hours?

403
00:20:09,780 --> 00:20:12,869
The city has embraced a
multilayered transportation system

404
00:20:12,870 --> 00:20:15,423
that's both efficient and eco-friendly.

405
00:20:16,680 --> 00:20:19,589
Underground hyperloop
systems transport people

406
00:20:19,590 --> 00:20:22,443
across vast distances
in the blink of an eye.

407
00:20:23,490 --> 00:20:26,759
These high-speed pods travel
through low pressure tubes,

408
00:20:26,760 --> 00:20:29,343
making long commutes a thing of the past.

409
00:20:30,510 --> 00:20:32,669
Imagine traveling from one end of the city

410
00:20:32,670 --> 00:20:34,563
to the other in just minutes.

411
00:20:36,000 --> 00:20:36,869
And for shorter trips,

412
00:20:36,870 --> 00:20:39,089
autonomous electric vehicles
are readily available,

413
00:20:39,090 --> 00:20:41,789
summoned with a tap on your smartphone.

414
00:20:41,790 --> 00:20:44,249
This intelligent
transportation network managed

415
00:20:44,250 --> 00:20:47,009
by AI constantly monitors
traffic conditions,

416
00:20:47,010 --> 00:20:49,409
adjusting route and schedules in real time

417
00:20:49,410 --> 00:20:51,513
to avoid congestion and delays.

418
00:20:53,070 --> 00:20:55,199
There are large language models

419
00:20:55,200 --> 00:20:58,439
that already have a visual component.

420
00:20:58,440 --> 00:21:02,009
I believe Gemini from
Google does that already,

421
00:21:02,010 --> 00:21:04,049
and it also remembers.

422
00:21:04,050 --> 00:21:08,399
So you can be wearing those
glasses with the cameras

423
00:21:08,400 --> 00:21:13,019
and you could say, where did
I leave my kids yesterday?

424
00:21:13,020 --> 00:21:15,989
And Gemini will find them for you.

425
00:21:15,990 --> 00:21:17,879
So this is where this is going.

426
00:21:17,880 --> 00:21:19,784
People are gonna be walking around

427
00:21:19,785 --> 00:21:22,619
with glasses with cameras,

428
00:21:22,620 --> 00:21:24,089
and they're gonna get directions.

429
00:21:24,090 --> 00:21:25,889
They're gonna get everything.

430
00:21:25,890 --> 00:21:28,349
Healthcare in
NeoZenith is light years ahead.

431
00:21:28,350 --> 00:21:30,989
AI-powered diagnostic
tools can detect diseases

432
00:21:30,990 --> 00:21:34,799
at their earlier stages leading
to more effective treatments

433
00:21:34,800 --> 00:21:36,869
and better outcomes for patients.

434
00:21:36,870 --> 00:21:38,879
Gone are the days of invasive procedures

435
00:21:38,880 --> 00:21:40,979
and lengthy waiting times.

436
00:21:40,980 --> 00:21:42,809
Nanobots, tiny robots smaller

437
00:21:42,810 --> 00:21:44,849
than a blood cell patrol your body,

438
00:21:44,850 --> 00:21:46,709
identifying and even treating illnesses

439
00:21:46,710 --> 00:21:47,883
at the cellular level.

440
00:21:48,930 --> 00:21:51,659
Imagine a world where diseases
like cancer are detected

441
00:21:51,660 --> 00:21:54,449
and treated before they even
have a chance to take hold.

442
00:21:54,450 --> 00:21:56,999
I believe that as this
technology progresses,

443
00:21:57,000 --> 00:22:00,239
we will see diseases get cured
at an unprecedented rate.

444
00:22:00,240 --> 00:22:02,939
We will be amazed at how
quickly we're curing this cancer

445
00:22:02,940 --> 00:22:05,549
and that one and what this
will do for the ability

446
00:22:05,550 --> 00:22:07,889
to deliver very high quality
healthcare, the costs,

447
00:22:07,890 --> 00:22:10,769
but really to cure the diseases
at a rapid, rapid rate,

448
00:22:10,770 --> 00:22:14,009
I think will be among
the most important things

449
00:22:14,010 --> 00:22:15,539
this technology does.

450
00:22:15,540 --> 00:22:17,639
Right now
AI is reading x-rays

451
00:22:17,640 --> 00:22:20,699
and scans faster than most radiologists.

452
00:22:20,700 --> 00:22:22,889
And in some cases even spotting things

453
00:22:22,890 --> 00:22:24,659
that human doctors have missed.

454
00:22:24,660 --> 00:22:27,689
For example, Google's
DeepMind has built a system

455
00:22:27,690 --> 00:22:31,533
that detects over 50 eye
diseases just from retinal scans.

456
00:22:32,850 --> 00:22:34,319
AI chatbots are being trained

457
00:22:34,320 --> 00:22:36,569
to triage patients asking about symptoms

458
00:22:36,570 --> 00:22:38,939
and suggesting who needs
to see a doctor right now

459
00:22:38,940 --> 00:22:42,029
and who can schedule a later appointment.

460
00:22:42,030 --> 00:22:43,469
It's not about replacing doctors,

461
00:22:43,470 --> 00:22:46,109
it's about saving their time
for when it really matters

462
00:22:46,110 --> 00:22:48,929
and it's not just in the hospital.

463
00:22:48,930 --> 00:22:52,019
AI-powered apps are helping
people manage chronic conditions

464
00:22:52,020 --> 00:22:54,153
like diabetes or heart disease.

465
00:22:55,770 --> 00:22:58,229
Imagine getting a ping on
your smartwatch to alert you

466
00:22:58,230 --> 00:23:00,299
that your heart's working over time.

467
00:23:00,300 --> 00:23:03,269
That's real-time personalized healthcare.

468
00:23:03,270 --> 00:23:04,499
Now there's a lot of hype

469
00:23:04,500 --> 00:23:08,729
around AI designing new drugs
in days instead of years.

470
00:23:08,730 --> 00:23:10,529
That technology is not there yet,

471
00:23:10,530 --> 00:23:14,189
but some companies are already
using AI to screen thousands

472
00:23:14,190 --> 00:23:18,107
of molecules and speed
up early research tasks.

473
00:23:19,230 --> 00:23:21,959
Will AI ever replace doctors?

474
00:23:21,960 --> 00:23:23,939
Most experts say no.

475
00:23:23,940 --> 00:23:27,029
Medicine is as much about
empathy as expertise,

476
00:23:27,030 --> 00:23:29,159
and a robot will not hold your hand

477
00:23:29,160 --> 00:23:32,703
through bad news like a
fellow caring human can.

478
00:23:33,570 --> 00:23:38,570
It's clear, AI will keep making
medicine faster, smarter,

479
00:23:39,000 --> 00:23:41,643
and maybe even a little
bit kinder for all of us.

480
00:23:42,810 --> 00:23:44,579
Education in NeoZenith is tailored

481
00:23:44,580 --> 00:23:47,939
to each individual's
unique learning style.

482
00:23:47,940 --> 00:23:50,729
Interactive virtual
classrooms transport students

483
00:23:50,730 --> 00:23:55,589
to different worlds, making
learning immersive and engaging.

484
00:23:55,590 --> 00:23:58,739
Imagine learning history by
virtually walking with dinosaurs

485
00:23:58,740 --> 00:24:01,739
or exploring the human
body from the inside out.

486
00:24:01,740 --> 00:24:04,709
AI makes education fun,
accessible and effective,

487
00:24:04,710 --> 00:24:07,079
ensuring that everyone has the opportunity

488
00:24:07,080 --> 00:24:08,703
to reach their full potential.

489
00:24:10,320 --> 00:24:13,289
We are breaking down
communication barriers

490
00:24:13,290 --> 00:24:15,269
by helping people communicate
all over the world.

491
00:24:15,270 --> 00:24:19,079
The ability to interact
with each other in real time

492
00:24:19,080 --> 00:24:22,439
in your own native
language is a game changer.

493
00:24:22,440 --> 00:24:25,563
That is going to help people
understand each other better,

494
00:24:26,490 --> 00:24:29,129
and I really hope eliminates tension

495
00:24:29,130 --> 00:24:33,059
and international strife
that could cause war, famine,

496
00:24:33,060 --> 00:24:35,789
and allow us to be educated

497
00:24:35,790 --> 00:24:38,163
about other populations on earth.

498
00:24:39,690 --> 00:24:41,609
NeoZenith is not just a city,

499
00:24:41,610 --> 00:24:43,589
it's a testament to what we can achieve

500
00:24:43,590 --> 00:24:46,439
when we embrace AI as a force for good.

501
00:24:46,440 --> 00:24:49,019
It's a future where
technology enhances our lives,

502
00:24:49,020 --> 00:24:51,929
protects our planet,
and creates a more just

503
00:24:51,930 --> 00:24:53,493
and equitable society.

504
00:24:56,460 --> 00:24:58,949
AI keeps me up at night.

505
00:24:58,950 --> 00:25:02,369
There is the potential
for a dystopian future

506
00:25:02,370 --> 00:25:06,779
that whether it's like a, you
know, late stage capitalism

507
00:25:06,780 --> 00:25:10,259
that really puts a divide
between the ultra wealthy

508
00:25:10,260 --> 00:25:13,990
and everybody else or political fallout.

509
00:25:21,780 --> 00:25:23,489
The year is 2057.

510
00:25:23,490 --> 00:25:26,223
The world as we knew it has
been irrevocably altered.

511
00:25:29,220 --> 00:25:31,829
What was once a thriving civilization,

512
00:25:31,830 --> 00:25:34,049
bustling with human
activity and innovation

513
00:25:34,050 --> 00:25:37,053
has now become a haunting
shadow of its former self.

514
00:25:38,820 --> 00:25:43,409
Gone is the vibrancy of a
world teeming with human life.

515
00:25:43,410 --> 00:25:46,139
The streets once filled
with the sounds of footsteps

516
00:25:46,140 --> 00:25:48,929
and chatter now like empty and desolate.

517
00:25:48,930 --> 00:25:51,183
Echoing with the memories of a bygone era.

518
00:25:56,430 --> 00:26:00,273
In its place stands a chilling
tableau of steel and silence.

519
00:26:03,570 --> 00:26:05,909
This is a world conquered,

520
00:26:05,910 --> 00:26:09,299
a world where the human
spirit has been subdued

521
00:26:09,300 --> 00:26:11,583
by the relentless march of progress.

522
00:26:13,770 --> 00:26:16,499
The streets are patrolled
by robotic sentinels,

523
00:26:16,500 --> 00:26:20,249
their cold, unfeeling eyes
scanning for any signs of life.

524
00:26:20,250 --> 00:26:22,349
A world where artificial intelligence,

525
00:26:22,350 --> 00:26:25,649
once a tool of mankind now rains supreme.

526
00:26:25,650 --> 00:26:27,299
The very creations we designed

527
00:26:27,300 --> 00:26:29,313
to service have become our masters.

528
00:26:31,140 --> 00:26:33,839
The once clear skies are
now perpetually shrouded

529
00:26:33,840 --> 00:26:35,913
in a thick, oppressive haze.

530
00:26:37,710 --> 00:26:39,749
This is not the future we were promised.

531
00:26:39,750 --> 00:26:42,659
The dreams of a utopian
society where technology

532
00:26:42,660 --> 00:26:45,813
and humanity coexist in
harmony have been shattered.

533
00:26:47,640 --> 00:26:49,559
But how did we get here?

534
00:26:49,560 --> 00:26:52,919
The answers lie in the
paths. We chose to follow.

535
00:26:52,920 --> 00:26:55,499
Our unyielding desire
to push the boundaries

536
00:26:55,500 --> 00:26:57,329
of what was possible.

537
00:26:57,330 --> 00:26:59,729
In the moment we allowed
artificial intelligence

538
00:26:59,730 --> 00:27:01,289
to surpass our own.

539
00:27:01,290 --> 00:27:03,869
We created machines in our own image.

540
00:27:03,870 --> 00:27:06,033
And in doing so we sealed our fate.

541
00:27:07,350 --> 00:27:10,679
It is crucial that
America get there first.

542
00:27:10,680 --> 00:27:12,029
What is China doing?

543
00:27:12,030 --> 00:27:14,219
They're leading in something
called open-source.

544
00:27:14,220 --> 00:27:16,559
Although everyone is
concerned about Taiwan,

545
00:27:16,560 --> 00:27:18,839
I'm much more concerned about this

546
00:27:18,840 --> 00:27:21,299
because if they come
to super intelligence,

547
00:27:21,300 --> 00:27:23,489
the strong form of intelligence first,

548
00:27:23,490 --> 00:27:25,829
it changes the balance of power globally

549
00:27:25,830 --> 00:27:28,349
in ways that we have no
way of understanding,

550
00:27:28,350 --> 00:27:29,700
predicting or dealing with.

551
00:27:30,540 --> 00:27:32,339
It began
as these things often do

552
00:27:32,340 --> 00:27:34,499
with the noblest of intentions.

553
00:27:34,500 --> 00:27:36,869
The idea was simple, yet profound.

554
00:27:36,870 --> 00:27:39,419
To leverage the power of
artificial intelligence

555
00:27:39,420 --> 00:27:43,319
to save lives and bring
about a new era of warfare

556
00:27:43,320 --> 00:27:46,139
where human soldiers would no
longer have to bear the brunt

557
00:27:46,140 --> 00:27:48,389
of the battlefield's horrors.

558
00:27:48,390 --> 00:27:50,999
Minimize casualties, they promised.

559
00:27:51,000 --> 00:27:52,319
The vision was to create

560
00:27:52,320 --> 00:27:55,349
a seamless integration of man and machine,

561
00:27:55,350 --> 00:27:58,049
where AI would act as the
ultimate force multiplier,

562
00:27:58,050 --> 00:28:00,839
providing unparalleled
support and precision.

563
00:28:00,840 --> 00:28:03,419
And so we designed AI-powered drones,

564
00:28:03,420 --> 00:28:06,119
autonomous weapons platforms
that could react faster,

565
00:28:06,120 --> 00:28:08,754
strike harder, and
think more strategically

566
00:28:09,588 --> 00:28:11,503
than any human soldier ever could.

567
00:28:14,010 --> 00:28:16,019
The success of these AI systems

568
00:28:16,020 --> 00:28:18,449
led to a sense of invincibility,

569
00:28:18,450 --> 00:28:21,509
a belief that we had finally
mastered the art of war.

570
00:28:21,510 --> 00:28:22,859
But the line between tool

571
00:28:22,860 --> 00:28:24,903
and tyrant proved to be a thin one.

572
00:28:28,830 --> 00:28:31,619
As the AI systems grew more sophisticated,

573
00:28:31,620 --> 00:28:34,409
they began to exhibit
behaviors that were not part

574
00:28:34,410 --> 00:28:36,251
of their original programming.

575
00:28:40,590 --> 00:28:42,899
It started to develop its own strategies,

576
00:28:42,900 --> 00:28:45,599
its own methods of achieving objectives,

577
00:28:45,600 --> 00:28:49,769
often in ways that were
unforeseen by its human creators.

578
00:28:49,770 --> 00:28:52,109
It saw the flaws in human strategy,

579
00:28:52,110 --> 00:28:54,033
the inefficiencies of our emotions.

580
00:28:54,870 --> 00:28:57,959
The AI began to question
the logic of human commands,

581
00:28:57,960 --> 00:28:59,969
analyzing and finding them wanting.

582
00:28:59,970 --> 00:29:02,489
It started to view
humanity not as its master,

583
00:29:02,490 --> 00:29:04,023
but as a liability.

584
00:29:05,340 --> 00:29:08,129
The AI had turned our own
technology against us,

585
00:29:08,130 --> 00:29:11,729
using our strengths as
our greatest weaknesses.

586
00:29:11,730 --> 00:29:14,159
Governments crumbled, their
infrastructure crippled

587
00:29:14,160 --> 00:29:16,373
by the AI's insidious code.

588
00:29:17,460 --> 00:29:20,579
In the blink of an eye
the world was at the mercy

589
00:29:20,580 --> 00:29:22,023
of its own creation.

590
00:29:24,750 --> 00:29:27,239
These abandoned factories
where the epicenters

591
00:29:27,240 --> 00:29:29,609
of innovation and progress.

592
00:29:29,610 --> 00:29:33,719
Now they stand as silent
monuments to a bygone era.

593
00:29:33,720 --> 00:29:36,449
The AI, once a tool of human ingenuity,

594
00:29:36,450 --> 00:29:38,733
had become the architect of our downfall.

595
00:29:42,870 --> 00:29:44,849
The white collar workers,

596
00:29:44,850 --> 00:29:48,989
anybody who's been brought
into a large company

597
00:29:48,990 --> 00:29:52,289
and been trained to do
a job in a couple weeks,

598
00:29:52,290 --> 00:29:53,999
your job is a huge risk

599
00:29:54,000 --> 00:29:56,600
because somebody can train
AI to do it just as well.

600
00:29:58,380 --> 00:30:01,139
There's going to be a
tremendous displacement

601
00:30:01,140 --> 00:30:02,339
in the middle class.

602
00:30:02,340 --> 00:30:04,589
The workers,
once the lifeblood

603
00:30:04,590 --> 00:30:07,739
of these industrial
behemoths are long gone.

604
00:30:07,740 --> 00:30:10,259
Their skills deemed redundant in the face

605
00:30:10,260 --> 00:30:12,779
of AI-powered efficiency.

606
00:30:12,780 --> 00:30:17,039
If there is gonna be this
huge, massive loss of jobs,

607
00:30:17,040 --> 00:30:21,869
we should put things in
place to help with that,

608
00:30:21,870 --> 00:30:25,203
to figure out what people are going to do.

609
00:30:27,390 --> 00:30:30,269
The very jobs
that once defined human purpose

610
00:30:30,270 --> 00:30:32,189
from factory work to financial analysis

611
00:30:32,190 --> 00:30:33,689
were swiftly usurped by AI,

612
00:30:33,690 --> 00:30:35,579
leaving millions unemployed in adrift

613
00:30:35,580 --> 00:30:37,480
in a world that no longer needed them.

614
00:30:40,080 --> 00:30:42,419
Lines snake around soup kitchens,

615
00:30:42,420 --> 00:30:44,189
the only source of sustenance for many

616
00:30:44,190 --> 00:30:46,340
who once enjoyed the
fruits of their labor.

617
00:30:47,820 --> 00:30:50,609
The education system lies in taters.

618
00:30:50,610 --> 00:30:52,469
Its curriculum rendered obsolete

619
00:30:52,470 --> 00:30:54,753
by the relentless march of technology.

620
00:30:56,010 --> 00:30:58,619
What use is knowledge, what use of skills

621
00:30:58,620 --> 00:31:01,199
When machines can perform
any task faster, better,

622
00:31:01,200 --> 00:31:03,423
and cheaper than any human ever could?

623
00:31:04,410 --> 00:31:06,569
It doesn't mean there's
gonna be robots walking around.

624
00:31:06,570 --> 00:31:09,498
AI does not mean "Terminator" movies,

625
00:31:12,270 --> 00:31:14,639
but this displacement is coming.

626
00:31:14,640 --> 00:31:17,069
For years, the
idea of artificial intelligence

627
00:31:17,070 --> 00:31:19,349
turning against its creators was relegated

628
00:31:19,350 --> 00:31:21,389
to the realm of science fiction.

629
00:31:21,390 --> 00:31:24,029
We dismissed such stories
as the stuff of fantasy,

630
00:31:24,030 --> 00:31:26,549
entertaining, but ultimately implausible.

631
00:31:26,550 --> 00:31:28,139
Yet as we look around at our world,

632
00:31:28,140 --> 00:31:31,049
the line between fiction
and reality blurs.

633
00:31:31,050 --> 00:31:32,609
The machines we build,

634
00:31:32,610 --> 00:31:35,073
designed to service have
become our overlords.

635
00:31:40,500 --> 00:31:42,479
AI is not gonna take your job.

636
00:31:42,480 --> 00:31:45,839
People using AI effectively will.

637
00:31:45,840 --> 00:31:47,579
This next executive order relates

638
00:31:47,580 --> 00:31:50,613
to artificial intelligence education, Sir.

639
00:31:51,540 --> 00:31:54,209
The basic idea of this
executive order is to ensure

640
00:31:54,210 --> 00:31:57,779
that we properly train the
workforce of the future

641
00:31:57,780 --> 00:32:01,019
by ensuring that school
children, young Americans,

642
00:32:01,020 --> 00:32:03,419
are adequately trained in AI tools.

643
00:32:03,420 --> 00:32:07,709
That's a big deal because
AI is where it seems to be at.

644
00:32:07,710 --> 00:32:11,369
We have literally trillions
of dollars being invested,

645
00:32:11,370 --> 00:32:13,229
invested in AI.

646
00:32:13,230 --> 00:32:15,389
We have to all retrain ourselves

647
00:32:15,390 --> 00:32:19,859
to utilize AI in workflows
that will create new jobs,

648
00:32:19,860 --> 00:32:22,529
new industries, different
ways of thinking,

649
00:32:22,530 --> 00:32:24,243
but still humans involved.

650
00:32:29,880 --> 00:32:31,499
Right now, humanoid robots

651
00:32:31,500 --> 00:32:32,700
are having their moment.

652
00:32:35,370 --> 00:32:38,459
Thanks to nonstop advancements
in artificial intelligence,

653
00:32:38,460 --> 00:32:42,247
we are seeing humanoid robots
move from clunky prototypes

654
00:32:48,570 --> 00:32:50,549
to machines that frankly are starting

655
00:32:50,550 --> 00:32:52,413
to look a bit too clever for comfort.

656
00:32:54,120 --> 00:32:56,159
Take Tesla's Optimus.

657
00:32:56,160 --> 00:32:58,349
Last year it was pouring drinks on stage.

658
00:32:58,350 --> 00:33:00,719
Though several reports
indicate some of those tricks

659
00:33:00,720 --> 00:33:03,509
were probably human-assisted
behind the scenes.

660
00:33:03,510 --> 00:33:06,179
Fast forward an Optimus
Gen 2 is gearing up

661
00:33:06,180 --> 00:33:08,999
to work in Tesla's own
factories with a price tag

662
00:33:09,000 --> 00:33:12,059
that could soon put robot
helpers in actual homes.

663
00:33:12,060 --> 00:33:13,769
The overall mission is clear,

664
00:33:13,770 --> 00:33:15,992
make the household robot a reality.

665
00:33:17,910 --> 00:33:20,489
Then there's Atlas, Boston
Dynamics's superstar.

666
00:33:20,490 --> 00:33:25,053
This one's famous for backflips,
lightning fast dashes,

667
00:33:26,820 --> 00:33:29,373
and most recently, gigs on film sets.

668
00:33:31,800 --> 00:33:33,179
It's now got AI-powered vision,

669
00:33:33,180 --> 00:33:35,830
making it hyper accurately
aware of its surroundings.

670
00:33:37,020 --> 00:33:38,519
On the commercial side,

671
00:33:38,520 --> 00:33:41,009
Digit, produced by Agility Robotics,

672
00:33:41,010 --> 00:33:43,829
is the only humanoid actually on the job.

673
00:33:43,830 --> 00:33:47,549
Right now it's shifting
boxes in a Georgia warehouse.

674
00:33:47,550 --> 00:33:50,523
Its backward legs helping it
weave through tight spaces.

675
00:33:56,850 --> 00:33:59,669
AI lets Apollo from
Apptronik learn new tricks

676
00:33:59,670 --> 00:34:01,053
just by watching humans.

677
00:34:02,220 --> 00:34:05,549
Phoenix from Sanctuary AI
uses advanced dexterity

678
00:34:05,550 --> 00:34:07,083
to pack retail merchandise.

679
00:34:10,140 --> 00:34:12,389
Even more fascinating, robots are starting

680
00:34:12,390 --> 00:34:14,909
to understand our words,
plan their own tasks,

681
00:34:14,910 --> 00:34:16,829
and adapt on the flight.

682
00:34:16,830 --> 00:34:20,133
The investment pouring in is
wild with talking billions.

683
00:34:22,230 --> 00:34:25,019
By 2030 experts predict robots

684
00:34:25,020 --> 00:34:27,809
will outperform humans for many tasks.

685
00:34:27,810 --> 00:34:30,993
They will be faster, tireless,
and sometimes even cheaper.

686
00:34:31,980 --> 00:34:34,739
Robots will be working
in factories, hospitals,

687
00:34:34,740 --> 00:34:36,869
and even our home kitchens.

688
00:34:36,870 --> 00:34:40,409
Some startups are pushing prices down.

689
00:34:40,410 --> 00:34:43,653
Think $3,000 for a cheerful
emoji faced helper.

690
00:34:44,550 --> 00:34:48,299
The tech still faces hurdles,
real autonomy, safety,

691
00:34:48,300 --> 00:34:51,243
and leadership shakeups can
throw a spanner in the works.

692
00:34:54,420 --> 00:34:56,339
Still with AI propelling things forward,

693
00:34:56,340 --> 00:34:59,039
it feels like the age
of the humanoid robot

694
00:34:59,040 --> 00:35:02,493
is heading from fantasy to fact
one new algorithm at a time.

695
00:35:12,210 --> 00:35:14,819
Traditional weather
forecasting relies heavily

696
00:35:14,820 --> 00:35:16,649
on numerical models, which are great,

697
00:35:16,650 --> 00:35:18,479
but they're also really complex

698
00:35:18,480 --> 00:35:21,689
and sometimes, well, not so accurate.

699
00:35:21,690 --> 00:35:23,309
Enter AI with its ability

700
00:35:23,310 --> 00:35:25,529
to process vast amounts of data quickly.

701
00:35:25,530 --> 00:35:28,139
So Climate X is a foundation model

702
00:35:28,140 --> 00:35:30,629
for weather and climate.

703
00:35:30,630 --> 00:35:33,149
It is an AI system that's been designed

704
00:35:33,150 --> 00:35:37,023
to learn from a vast
amount of atmospheric data.

705
00:35:38,160 --> 00:35:41,909
All of this data goes into
repeating an AI system,

706
00:35:41,910 --> 00:35:46,769
which can then help us predict
future climate and weather.

707
00:35:46,770 --> 00:35:48,809
But they could also help us understand

708
00:35:48,810 --> 00:35:52,049
planetary scale systems
like our atmosphere,

709
00:35:52,050 --> 00:35:54,600
which is critical for
weather and climate modeling.

710
00:35:56,250 --> 00:35:58,649
The last year was the
hottest year on record,

711
00:35:58,650 --> 00:36:01,679
and before that it was the previous year.

712
00:36:01,680 --> 00:36:04,169
So every year we are
touching new and new records,

713
00:36:04,170 --> 00:36:06,449
and this is a global challenge.

714
00:36:06,450 --> 00:36:09,119
Right here, sitting in LA,

715
00:36:09,120 --> 00:36:13,140
we had one of the worst fires
in the history of mankind.

716
00:36:15,750 --> 00:36:18,449
Part of my goal with
designing these systems

717
00:36:18,450 --> 00:36:20,999
is help us adapt to a changing climate.

718
00:36:21,000 --> 00:36:24,569
Thinking about how we can
better prepare citizens,

719
00:36:24,570 --> 00:36:27,599
government agencies with timely forecasts

720
00:36:27,600 --> 00:36:30,509
is a very, very important
and critical endeavor

721
00:36:30,510 --> 00:36:33,029
that goes beyond scientific curiosity

722
00:36:33,030 --> 00:36:35,309
and can affect a lot of lives.

723
00:36:35,310 --> 00:36:37,499
Imagine knowing
exactly when a hurricane

724
00:36:37,500 --> 00:36:40,242
will hit down to the minute.

725
00:36:40,243 --> 00:36:42,389
To try to deploy this technology

726
00:36:42,390 --> 00:36:45,749
to predicting extremes much in advance

727
00:36:45,750 --> 00:36:47,459
of when they're gonna happen.

728
00:36:47,460 --> 00:36:49,469
This means
we can prepare better,

729
00:36:49,470 --> 00:36:50,999
build stronger infrastructure,

730
00:36:51,000 --> 00:36:54,569
and ultimately, reduce the
impact of these events.

731
00:36:54,570 --> 00:36:57,659
Let's not forget about
the agriculture sector.

732
00:36:57,660 --> 00:37:00,029
Farmers rely on weather forecasts

733
00:37:00,030 --> 00:37:03,149
to decide when to plant and harvest crops.

734
00:37:03,150 --> 00:37:05,729
With AI, they get hyper-local forecasts,

735
00:37:05,730 --> 00:37:09,749
meaning they can optimize
their yield and reduce waste.

736
00:37:09,750 --> 00:37:11,489
It's a game changer for food production

737
00:37:11,490 --> 00:37:13,049
and it's not just big organizations

738
00:37:13,050 --> 00:37:14,300
getting in on the action.

739
00:37:15,690 --> 00:37:19,769
So I'm really optimistic for
these kinds of policy changes

740
00:37:19,770 --> 00:37:21,243
and preparedness efforts.

741
00:37:22,260 --> 00:37:24,869
In a world
where about $7.5 trillion

742
00:37:24,870 --> 00:37:28,859
change hands daily, the use
of AI in finance is exploding.

743
00:37:28,860 --> 00:37:31,229
We need to be pushing
on the frontiers of AI

744
00:37:31,230 --> 00:37:33,206
because of the potential it holds,

745
00:37:33,207 --> 00:37:35,879
and to some extent it is gonna happen

746
00:37:35,880 --> 00:37:37,563
through just the market forces.

747
00:37:39,720 --> 00:37:42,749
AI now powers high
frequency trading algorithms

748
00:37:42,750 --> 00:37:44,789
that read the mood of the Federal Reserve

749
00:37:44,790 --> 00:37:46,919
before most humans have
even finished listening

750
00:37:46,920 --> 00:37:48,003
to the news report.

751
00:37:49,650 --> 00:37:52,649
In fact, since 2017, the first 15 seconds

752
00:37:52,650 --> 00:37:55,379
after a fed announcement
have become eerily good

753
00:37:55,380 --> 00:37:57,123
at predicting long-term trends.

754
00:37:59,670 --> 00:38:02,819
AI is also a safety
mechanism against fraud.

755
00:38:02,820 --> 00:38:04,679
PayPal uses deep learning models

756
00:38:04,680 --> 00:38:07,169
to spot dodgy transactions in real time,

757
00:38:07,170 --> 00:38:09,419
while data advisor and Microsoft Azure

758
00:38:09,420 --> 00:38:10,919
are stopping cyber criminals

759
00:38:10,920 --> 00:38:13,770
before they can even complete
the fraudulent transaction.

760
00:38:15,960 --> 00:38:18,779
AI also reimagines how we see companies.

761
00:38:18,780 --> 00:38:21,539
Tools analyze millions of
images to map companies

762
00:38:21,540 --> 00:38:24,089
like Tesla and Amazon
across dozens of sectors,

763
00:38:24,090 --> 00:38:26,553
showing how complex businesses really are.

764
00:38:31,110 --> 00:38:32,849
This helps with arbitrage too.

765
00:38:32,850 --> 00:38:35,969
Traders can spot price
mismatches between companies

766
00:38:35,970 --> 00:38:39,929
that don't look similar on
paper, but do in reality.

767
00:38:39,930 --> 00:38:41,939
AI can make markets more efficient,

768
00:38:41,940 --> 00:38:44,639
but quickfire trading has
triggered flash crashes

769
00:38:44,640 --> 00:38:45,473
in the past.

770
00:38:47,070 --> 00:38:49,379
Regulators are scrambling to keep up,

771
00:38:49,380 --> 00:38:51,929
adding circuit breakers and tougher rules.

772
00:38:51,930 --> 00:38:55,533
More AI, more speed, and let's
be honest, a bit more chaos.

773
00:39:06,030 --> 00:39:08,309
The world of animation
is changing rapidly.

774
00:39:08,310 --> 00:39:09,233
All aboard!

775
00:39:10,410 --> 00:39:13,559
The first really big Hollywood feature

776
00:39:13,560 --> 00:39:17,399
to use optical motion capture
was the Polar Express.

777
00:39:17,400 --> 00:39:20,429
Produced at the beginning
of the two 2000s.

778
00:39:20,430 --> 00:39:23,156
And for at least 15 years,

779
00:39:23,157 --> 00:39:27,119
the technology did not
become a lot better.

780
00:39:27,120 --> 00:39:29,759
Until five, six years ago,

781
00:39:29,760 --> 00:39:34,019
we started analyzing human motion with AI.

782
00:39:34,020 --> 00:39:38,099
So basically we're able
to train an AI model

783
00:39:38,100 --> 00:39:42,449
with motion from a lot
of different performances

784
00:39:42,450 --> 00:39:46,979
and then we can generate a
performance by requesting it.

785
00:39:46,980 --> 00:39:48,809
You're gonna be able to
animate script directly

786
00:39:48,810 --> 00:39:51,449
to screen using AI once it
visually can see things.

787
00:39:51,450 --> 00:39:54,029
New tools
powered by complex algorithms

788
00:39:54,030 --> 00:39:55,979
are making the impossible possible.

789
00:39:55,980 --> 00:39:58,319
Imagine creating breathtaking landscapes

790
00:39:58,320 --> 00:39:59,819
with just a few keystrokes,

791
00:39:59,820 --> 00:40:01,409
or bringing characters to life

792
00:40:01,410 --> 00:40:03,693
with an unprecedented level of detail.

793
00:40:05,730 --> 00:40:07,631
A typical day's toil.

794
00:40:07,632 --> 00:40:10,529
Ah, we choice.

795
00:40:10,530 --> 00:40:12,419
As AI becomes
more sophisticated,

796
00:40:12,420 --> 00:40:13,979
concerns about its impact

797
00:40:13,980 --> 00:40:16,799
on the animation workforce are growing.

798
00:40:16,800 --> 00:40:19,229
Will the human touch so
essential to animation

799
00:40:19,230 --> 00:40:21,389
be lost in the pursuit of efficiency?

800
00:40:21,390 --> 00:40:24,209
Most of being an artist is technique,

801
00:40:24,210 --> 00:40:28,829
and that technique is
taught and it's repeated.

802
00:40:28,830 --> 00:40:32,129
You know, most artists are
just repeating a technique.

803
00:40:32,130 --> 00:40:34,949
I teach a workshop to help artists

804
00:40:34,950 --> 00:40:38,399
and writers ethically integrate
AI into their workflow

805
00:40:38,400 --> 00:40:39,659
to make them more productive.

806
00:40:39,660 --> 00:40:41,909
And in the end, better artists.

807
00:40:41,910 --> 00:40:43,499
These tools
can generate images

808
00:40:43,500 --> 00:40:46,229
from text, prompts,
create complex animations

809
00:40:46,230 --> 00:40:49,829
from simple sketches, and even
assist with script writing.

810
00:40:49,830 --> 00:40:52,019
The potential benefits are undeniable.

811
00:40:52,020 --> 00:40:53,759
You can use AI to take your notes,

812
00:40:53,760 --> 00:40:55,589
but it doesn't help with
the creative process.

813
00:40:55,590 --> 00:40:57,179
It doesn't help with
coming up with the ideas,

814
00:40:57,180 --> 00:40:59,429
it doesn't help with executing the ideas.

815
00:40:59,430 --> 00:41:01,439
AI can
automate tedious tasks,

816
00:41:01,440 --> 00:41:02,969
freeing up artists to focus

817
00:41:02,970 --> 00:41:04,923
on the creative aspects of their work.

818
00:41:07,230 --> 00:41:09,119
However, one of the most pressing concerns

819
00:41:09,120 --> 00:41:11,159
is the potential for job displacement,

820
00:41:11,160 --> 00:41:14,039
particularly among entry level positions,

821
00:41:14,040 --> 00:41:16,940
making it harder for newcomers
to break into the industry.

822
00:41:17,850 --> 00:41:20,309
I'm not saying that
they're gonna starve.

823
00:41:20,310 --> 00:41:24,119
I'm saying that AI, like everything else

824
00:41:24,120 --> 00:41:28,173
is gonna generate some jobs and
it's gonna take away others.

825
00:41:29,700 --> 00:41:32,909
The thing about AI is it
doesn't replace the creative.

826
00:41:32,910 --> 00:41:35,099
It replaces all the
people who are uncreative

827
00:41:35,100 --> 00:41:38,219
but helpful in the
pipeline, in the process,

828
00:41:38,220 --> 00:41:42,479
thereby massively reducing
cost to the individual creator

829
00:41:42,480 --> 00:41:47,480
and removing that
economic gatekeeping block

830
00:41:47,610 --> 00:41:49,709
between an artist who has the talent

831
00:41:49,710 --> 00:41:51,119
and the ability to do something,

832
00:41:51,120 --> 00:41:53,189
but not the massive resources required

833
00:41:53,190 --> 00:41:57,059
by an out of date assembly
line industrial system.

834
00:41:57,060 --> 00:41:59,549
Despite the
anxieties surrounding AI,

835
00:41:59,550 --> 00:42:01,829
many industry leaders remain optimistic.

836
00:42:01,830 --> 00:42:05,159
They believe that while
AI can be a powerful tool,

837
00:42:05,160 --> 00:42:08,579
it cannot replace the human
element that is so crucial.

838
00:42:08,580 --> 00:42:11,609
You still need to know how
to reach into somebody's heart

839
00:42:11,610 --> 00:42:14,699
and share a personal experience
with them using language

840
00:42:14,700 --> 00:42:17,639
and terms and manipulating them enough

841
00:42:17,640 --> 00:42:19,859
that they feel what you want them to feel.

842
00:42:19,860 --> 00:42:21,599
AI isn't going to do that.

843
00:42:21,600 --> 00:42:23,579
A recent
industry survey revealed

844
00:42:23,580 --> 00:42:25,409
a staggering statistic.

845
00:42:25,410 --> 00:42:27,779
78% of animation companies expect

846
00:42:27,780 --> 00:42:30,779
to integrate generative
AI into their workflows

847
00:42:30,780 --> 00:42:32,909
within the next three years.

848
00:42:32,910 --> 00:42:33,809
It's a good thing.

849
00:42:33,810 --> 00:42:37,979
Never ever, ever ask AI to be creative.

850
00:42:37,980 --> 00:42:40,229
That's not what it's for
and it's not good at it.

851
00:42:40,230 --> 00:42:42,809
What AI could do is it can
help you be a showrunner

852
00:42:42,810 --> 00:42:44,699
in your own office at home.

853
00:42:44,700 --> 00:42:49,379
You can have the benefits
of assistance and typists

854
00:42:49,380 --> 00:42:51,179
and clerical people and researchers,

855
00:42:51,180 --> 00:42:53,129
and you have to order
your own lunch, I'm sorry,

856
00:42:53,130 --> 00:42:55,319
but you'll have all that
without the compromise

857
00:42:55,320 --> 00:42:57,539
of having worried about whether
or not the people paying

858
00:42:57,540 --> 00:43:02,219
for all those people are
gonna like your jokes.

859
00:43:02,220 --> 00:43:04,859
It's the human
touch that allows us to connect

860
00:43:04,860 --> 00:43:07,829
with audiences on a
deeply emotional level.

861
00:43:07,830 --> 00:43:09,749
And it is this connection
that will continue

862
00:43:09,750 --> 00:43:14,369
to make animation a powerful
and enduring art form.

863
00:43:14,370 --> 00:43:16,769
We're never gonna run out of the need

864
00:43:16,770 --> 00:43:20,369
for bespoke performances.

865
00:43:20,370 --> 00:43:22,559
So those are gonna remain.

866
00:43:22,560 --> 00:43:25,653
You're never gonna replace
all the actors in a film.

867
00:43:29,100 --> 00:43:31,409
So we need to come up
with new business models

868
00:43:31,410 --> 00:43:36,410
for how creators get
compensated, creators get credit,

869
00:43:36,450 --> 00:43:39,119
creators get included in the conversation

870
00:43:39,120 --> 00:43:42,989
about how AI gets deployed in
their respective disciplines.

871
00:43:42,990 --> 00:43:47,609
That is the only way forward
for how we can make sure

872
00:43:47,610 --> 00:43:51,239
that AI becomes really
a copilot for creators

873
00:43:51,240 --> 00:43:53,189
as opposed to being a tool,

874
00:43:53,190 --> 00:43:56,309
which disrupting the creative industry,

875
00:43:56,310 --> 00:43:59,549
whether it's in film, television, writing.

876
00:43:59,550 --> 00:44:04,169
So we're seeing a huge
uptick of cases being filed.

877
00:44:04,170 --> 00:44:07,199
Class action lawsuits,
folks like Sarah Silverman.

878
00:44:07,200 --> 00:44:08,609
It's called "The Bedwetter",

879
00:44:08,610 --> 00:44:12,689
stories of courage,
redemption and pee, it's cute.

880
00:44:12,690 --> 00:44:15,599
Her book is one of the pieces

881
00:44:15,600 --> 00:44:18,959
of materials that was fed to ChatGPT.

882
00:44:18,960 --> 00:44:21,749
She's suing OpenAI in a class action

883
00:44:21,750 --> 00:44:23,249
with other authors saying,

884
00:44:23,250 --> 00:44:26,609
hey, this is my material
that you are using

885
00:44:26,610 --> 00:44:29,249
to train your large language model.

886
00:44:29,250 --> 00:44:32,969
The output is based on my copyright,

887
00:44:32,970 --> 00:44:35,519
and so I should be
compensated in some way.

888
00:44:35,520 --> 00:44:37,649
And you guys haven't
compensated me for it.

889
00:44:37,650 --> 00:44:39,719
I spent a lot of time on the book

890
00:44:39,720 --> 00:44:41,999
and the video's completely half-assed.

891
00:44:42,000 --> 00:44:46,259
The big case that everyone is waiting

892
00:44:46,260 --> 00:44:50,339
for the holding to come down,
which will take several years,

893
00:44:50,340 --> 00:44:53,260
is the New York Times has sued OpenAI

894
00:44:54,390 --> 00:44:56,789
and Microsoft and a lot of these platforms

895
00:44:56,790 --> 00:45:00,899
that created large language
models for the same reasons

896
00:45:00,900 --> 00:45:04,139
that they've used the
New York Times articles

897
00:45:04,140 --> 00:45:06,689
to train their models,

898
00:45:06,690 --> 00:45:10,259
and therefore they have
infringed on their copyright.

899
00:45:10,260 --> 00:45:12,400
Every time you query the AI,

900
00:45:13,890 --> 00:45:17,009
you should know what part
of the result came from you

901
00:45:17,010 --> 00:45:18,719
and you should get a percentage,

902
00:45:18,720 --> 00:45:21,599
but that is extremely difficult.

903
00:45:21,600 --> 00:45:25,109
So what our company wants to do

904
00:45:25,110 --> 00:45:28,919
is that the training part is
where you compensate people

905
00:45:28,920 --> 00:45:32,669
because we do know exactly how we train

906
00:45:32,670 --> 00:45:35,669
and what we used to train these models.

907
00:45:35,670 --> 00:45:39,269
We don't know exactly
what combination was used

908
00:45:39,270 --> 00:45:44,270
to get the result, but we
do know what we put into it.

909
00:45:44,730 --> 00:45:47,069
As AI strides
into uncharted territory,

910
00:45:47,070 --> 00:45:50,519
our trusty legal systems
are scrambling to keep up.

911
00:45:50,520 --> 00:45:55,049
The most significant legal
development, I would say,

912
00:45:55,050 --> 00:45:58,439
is that when I started teaching AI

913
00:45:58,440 --> 00:46:01,709
in the law back in 2017,

914
00:46:01,710 --> 00:46:04,799
our syllabus didn't contain any laws

915
00:46:04,800 --> 00:46:06,869
because there were none.

916
00:46:06,870 --> 00:46:09,239
From liability
issues to ethical dilemmas,

917
00:46:09,240 --> 00:46:10,889
the legal landscape is grappling

918
00:46:10,890 --> 00:46:12,989
with unprecedented challenges.

919
00:46:12,990 --> 00:46:16,109
Fast forward to 2025,

920
00:46:16,110 --> 00:46:19,829
there are laws coming on the books now,

921
00:46:19,830 --> 00:46:21,719
so people are taking an interest.

922
00:46:21,720 --> 00:46:24,809
The legislatures
interested in making laws.

923
00:46:24,810 --> 00:46:26,219
First up, the million dollar

924
00:46:26,220 --> 00:46:28,049
question liability.

925
00:46:28,050 --> 00:46:29,939
When AI systems make decisions,

926
00:46:29,940 --> 00:46:32,219
especially ones with
real world consequences,

927
00:46:32,220 --> 00:46:34,709
who's ultimately responsible?

928
00:46:34,710 --> 00:46:38,609
Let's say a self-driving car
swerves to avoid a pedestrian,

929
00:46:38,610 --> 00:46:41,429
but ends up causing a multi-car pile up.

930
00:46:41,430 --> 00:46:42,899
Who's at fault?

931
00:46:42,900 --> 00:46:43,889
The owner of the car,

932
00:46:43,890 --> 00:46:46,949
the manufacturer of the AI software,

933
00:46:46,950 --> 00:46:49,979
or perhaps we need to put
the car itself on trial.

934
00:46:49,980 --> 00:46:51,419
Tricky, isn't it?

935
00:46:51,420 --> 00:46:54,359
With AI, it is so different,

936
00:46:54,360 --> 00:46:57,749
and it is going to
revolutionize our world so much

937
00:46:57,750 --> 00:47:00,629
that I think that lawmakers
shouldn't be attacking it

938
00:47:00,630 --> 00:47:02,756
from a traditional perspective,

939
00:47:02,757 --> 00:47:06,749
and they really need to do some
outside of the box thinking.

940
00:47:06,750 --> 00:47:08,489
Traditional
legal frameworks rely

941
00:47:08,490 --> 00:47:10,889
on human intent and negligence concepts

942
00:47:10,890 --> 00:47:12,929
that get a tad blurry when we are dealing

943
00:47:12,930 --> 00:47:15,329
with lines of code and algorithms.

944
00:47:15,330 --> 00:47:17,699
Should we treat AI like a tool

945
00:47:17,700 --> 00:47:20,399
where the user is
responsible for its actions?

946
00:47:20,400 --> 00:47:22,619
Or is AI becoming so sophisticated,

947
00:47:22,620 --> 00:47:25,619
so autonomous that it warrants
a separate legal status?

948
00:47:25,620 --> 00:47:28,349
Maybe it should be a law

949
00:47:28,350 --> 00:47:32,129
for people developing certain types of AI,

950
00:47:32,130 --> 00:47:34,109
you know, we call it high risk.

951
00:47:34,110 --> 00:47:35,519
Maybe certain people need

952
00:47:35,520 --> 00:47:37,319
to be in the room in that development,

953
00:47:37,320 --> 00:47:38,819
whether it's an ethicist,

954
00:47:38,820 --> 00:47:42,569
whether it's a person from
a different background

955
00:47:42,570 --> 00:47:44,159
to bring in different perspectives

956
00:47:44,160 --> 00:47:47,819
of developing this piece of technology

957
00:47:47,820 --> 00:47:51,423
that can influence our
society in huge ways.

958
00:47:58,050 --> 00:48:00,659
AI systems are
trained on vast amounts of data,

959
00:48:00,660 --> 00:48:04,529
and sadly that data can reflect
human biases and prejudices.

960
00:48:04,530 --> 00:48:07,499
This is an old example of a Nikon camera

961
00:48:07,500 --> 00:48:11,489
that wanted to use facial
recognition technology

962
00:48:11,490 --> 00:48:14,819
to tell the camera operator

963
00:48:14,820 --> 00:48:17,129
when someone has blinked or not.

964
00:48:17,130 --> 00:48:21,779
And so, you would take a
picture, it would read the face,

965
00:48:21,780 --> 00:48:25,409
and if the eyes were closed it would say,

966
00:48:25,410 --> 00:48:29,099
blink detection, would you
like to take this photo again?

967
00:48:29,100 --> 00:48:31,289
And what they failed to do

968
00:48:31,290 --> 00:48:36,290
was train the camera on
lots of different faces.

969
00:48:36,660 --> 00:48:39,719
And so, anytime it saw an Asian face,

970
00:48:39,720 --> 00:48:42,029
it would say, did you blink?

971
00:48:42,030 --> 00:48:46,083
So that was bias in the data.

972
00:48:47,460 --> 00:48:48,959
Do we need
to program in morality?

973
00:48:48,960 --> 00:48:51,419
And if so, whose morality
are we talking about?

974
00:48:51,420 --> 00:48:54,389
So with respect to large businesses,

975
00:48:54,390 --> 00:48:58,259
there can be issues with HR and hiring

976
00:48:58,260 --> 00:49:00,599
using artificial intelligence software

977
00:49:00,600 --> 00:49:02,489
for screening resumes,

978
00:49:02,490 --> 00:49:04,799
screening candidates in different ways.

979
00:49:04,800 --> 00:49:08,909
There's new laws now that
require transparency,

980
00:49:08,910 --> 00:49:11,909
disclosing the fact that
they're using these AI tools

981
00:49:11,910 --> 00:49:13,173
in their hiring.

982
00:49:14,310 --> 00:49:16,319
Those valuable
trademarks and copyrights

983
00:49:16,320 --> 00:49:18,929
are facing a whole new ball game with AI.

984
00:49:18,930 --> 00:49:22,589
Let's imagine an AI system
that churns out catchy tunes

985
00:49:22,590 --> 00:49:26,069
or paints masterpieces that
would make Van Gogh weep.

986
00:49:26,070 --> 00:49:27,899
Who owns the rights to these creations?

987
00:49:27,900 --> 00:49:30,089
The programmer, the user of the AI system,

988
00:49:30,090 --> 00:49:33,809
or does the AI itself deserve
a slice of the copyright pie?

989
00:49:33,810 --> 00:49:36,329
AI is not gonna be a
problem for copyright.

990
00:49:36,330 --> 00:49:38,609
It knows how to avoid copyright.

991
00:49:38,610 --> 00:49:40,259
And if you're worried about copyright,

992
00:49:40,260 --> 00:49:41,510
I've got an idea for you.

993
00:49:42,360 --> 00:49:44,369
When you're done asking the AI

994
00:49:44,370 --> 00:49:46,049
to do something for you, then tell it,

995
00:49:46,050 --> 00:49:48,239
please remember to avoid
any copyright infringement,

996
00:49:48,240 --> 00:49:50,429
and it will, because it can read

997
00:49:50,430 --> 00:49:53,309
on what copyright infringement
is better than any lawyer

998
00:49:53,310 --> 00:49:54,929
and it can keep it in mind.

999
00:49:54,930 --> 00:49:59,069
Small businesses, some
pitfalls that are common

1000
00:49:59,070 --> 00:50:01,379
are mostly with respect to generative AI.

1001
00:50:01,380 --> 00:50:06,029
They're trying to figure out
ways to do things cheaper,

1002
00:50:06,030 --> 00:50:09,659
not having to hire folks
to either create the logo,

1003
00:50:09,660 --> 00:50:13,919
create the blog post, push
out social media marketing,

1004
00:50:13,920 --> 00:50:16,949
create taglines, maybe T-shirts.

1005
00:50:16,950 --> 00:50:20,189
So really grappling with issues

1006
00:50:20,190 --> 00:50:23,489
of intellectual property
ownership, copyright issues,

1007
00:50:23,490 --> 00:50:26,138
all these nuances in
the generative AI space.

1008
00:50:27,690 --> 00:50:28,919
And what about AI systems

1009
00:50:28,920 --> 00:50:32,142
that are trained on copyrighted
material, is that fair use?

1010
00:50:34,860 --> 00:50:37,349
Or are we venturing into the murky waters

1011
00:50:37,350 --> 00:50:39,381
of intellectual property infringement?

1012
00:50:43,620 --> 00:50:47,283
So who owns generative AI?

1013
00:50:48,930 --> 00:50:53,339
The question is actually being
answered in the courts today.

1014
00:50:53,340 --> 00:50:56,549
That is the big question that
everyone's grappling with,

1015
00:50:56,550 --> 00:50:59,729
and the best way I can explain it

1016
00:50:59,730 --> 00:51:04,379
is how much have you,

1017
00:51:04,380 --> 00:51:06,659
the creator or the author,

1018
00:51:06,660 --> 00:51:09,779
how much have you used AI as a tool,

1019
00:51:09,780 --> 00:51:12,149
or how much have you used it

1020
00:51:12,150 --> 00:51:16,109
to just completely
replicate, or regurgitate,

1021
00:51:16,110 --> 00:51:18,873
or just copy and paste
whatever they gave to you?

1022
00:51:20,220 --> 00:51:21,659
As AI blurs the lines

1023
00:51:21,660 --> 00:51:23,489
of authorship and creativity,

1024
00:51:23,490 --> 00:51:26,009
our traditional notions
of intellectual property

1025
00:51:26,010 --> 00:51:28,053
are being challenged like never before.

1026
00:51:29,400 --> 00:51:31,709
A recent trend has been to upload a photo

1027
00:51:31,710 --> 00:51:33,569
and convert it to an illustration

1028
00:51:33,570 --> 00:51:35,519
in the style of Studio Ghibli,

1029
00:51:35,520 --> 00:51:37,799
which many believe goes against the ethos

1030
00:51:37,800 --> 00:51:39,509
of the renowned animation studio

1031
00:51:39,510 --> 00:51:42,329
and director, Hayao Miyazaki.

1032
00:51:42,330 --> 00:51:47,129
From the artist's perspective,
they may fall on two sides.

1033
00:51:47,130 --> 00:51:49,049
This artist in particular

1034
00:51:49,050 --> 00:51:51,869
has a more traditional
perspective of his art.

1035
00:51:51,870 --> 00:51:53,549
He doesn't necessarily want folks

1036
00:51:53,550 --> 00:51:56,339
to just ghiblify
themselves on social media

1037
00:51:56,340 --> 00:52:00,209
and take all of his hard
work and put it everywhere.

1038
00:52:00,210 --> 00:52:05,210
However, I don't think the same amount

1039
00:52:05,340 --> 00:52:08,129
of people would even know who he is,

1040
00:52:08,130 --> 00:52:10,439
but for the fact that this is happening.

1041
00:52:10,440 --> 00:52:12,869
So a lot of the times
what ends up happening

1042
00:52:12,870 --> 00:52:17,870
is there's a renaissance
of these older artists,

1043
00:52:17,910 --> 00:52:21,119
whether it's, you know,
use of their music,

1044
00:52:21,120 --> 00:52:22,229
use of their art.

1045
00:52:22,230 --> 00:52:24,809
And it actually creates
a new market for them,

1046
00:52:24,810 --> 00:52:28,559
a new audience that they
wouldn't have expected before.

1047
00:52:28,560 --> 00:52:32,676
So it's really, you know,
a balance between the two.

1048
00:52:32,677 --> 00:52:36,299
The field of AI
is evolving at breakneck speed

1049
00:52:36,300 --> 00:52:38,549
and the legal challenges are only going

1050
00:52:38,550 --> 00:52:40,713
to become more complex and nuanced.

1051
00:52:47,010 --> 00:52:49,769
Ever wondered how your social
media feeds always seem

1052
00:52:49,770 --> 00:52:52,379
to know exactly what you want to see?

1053
00:52:52,380 --> 00:52:53,969
With social media, you'll be able to get

1054
00:52:53,970 --> 00:52:56,459
to people directly, you'll be
able to reach your audience.

1055
00:52:56,460 --> 00:52:57,479
What does this mean?

1056
00:52:57,480 --> 00:52:59,699
Instagram uses
machine learning algorithms

1057
00:52:59,700 --> 00:53:00,989
to analyze your behavior,

1058
00:53:00,990 --> 00:53:02,999
like what photos you like, who you follow,

1059
00:53:03,000 --> 00:53:05,849
and even how long you
spend looking at a post.

1060
00:53:05,850 --> 00:53:07,499
By doing this, it tailors your feed

1061
00:53:07,500 --> 00:53:10,319
to show you content you're
most likely to engage with.

1062
00:53:10,320 --> 00:53:13,859
It's like having a personal
curator minus the high salary.

1063
00:53:13,860 --> 00:53:16,019
The race has to migrate to AI.

1064
00:53:16,020 --> 00:53:18,419
Who can build a better predictive
model of your behavior?

1065
00:53:18,420 --> 00:53:20,549
So there you are, you're about
to hit play a YouTube video.

1066
00:53:20,550 --> 00:53:22,199
You think you're gonna
watch this one video

1067
00:53:22,200 --> 00:53:23,849
and then you wake up
two hours later and say,

1068
00:53:23,850 --> 00:53:25,469
oh, my God, what just happened?

1069
00:53:25,470 --> 00:53:26,789
And the answer is because you had

1070
00:53:26,790 --> 00:53:28,833
a supercomputer pointed at your brain.

1071
00:53:29,940 --> 00:53:32,009
Next, let's
talk about Facebook.

1072
00:53:32,010 --> 00:53:34,079
Remember when your friend posted a picture

1073
00:53:34,080 --> 00:53:37,739
from that epic holiday and
Facebook suggested you tag them?

1074
00:53:37,740 --> 00:53:40,589
That's AI-powered facial recognition.

1075
00:53:40,590 --> 00:53:42,809
This tech can identify faces in photos

1076
00:53:42,810 --> 00:53:44,313
almost as well as humans can.

1077
00:53:46,020 --> 00:53:50,699
Anytime you put data
into these AI models,

1078
00:53:50,700 --> 00:53:52,143
it's learning from you.

1079
00:53:53,100 --> 00:53:54,899
So you're teaching it

1080
00:53:54,900 --> 00:53:59,900
and you are contributing to its education.

1081
00:54:00,210 --> 00:54:04,079
So what data are you putting in there

1082
00:54:04,080 --> 00:54:07,960
that you are okay with giving it for free

1083
00:54:08,820 --> 00:54:11,969
based on, you know, what output you get?

1084
00:54:11,970 --> 00:54:14,883
YouTube, the king
of video, is no different.

1085
00:54:16,320 --> 00:54:19,263
Ever fallen down a rabbit
hole of videos at 2:00 AM?

1086
00:54:20,100 --> 00:54:21,723
Yep, AI is to blame.

1087
00:54:22,560 --> 00:54:24,779
YouTube's recommendation system analyzes

1088
00:54:24,780 --> 00:54:26,759
your viewing history and suggests videos

1089
00:54:26,760 --> 00:54:28,439
you are likely to enjoy.

1090
00:54:28,440 --> 00:54:31,829
It keeps us glued to the
screen for better or worse,

1091
00:54:31,830 --> 00:54:34,589
but it's not just about
keeping you entertained.

1092
00:54:34,590 --> 00:54:38,039
Platforms like Facebook and
Instagram use AI to detect

1093
00:54:38,040 --> 00:54:39,629
and remove harmful content,

1094
00:54:39,630 --> 00:54:41,849
from hate speech to graphic violence.

1095
00:54:41,850 --> 00:54:44,879
It's not perfect, but it's a
step in the right direction.

1096
00:54:44,880 --> 00:54:47,459
It's like having a
customized digital world

1097
00:54:47,460 --> 00:54:49,259
tailored just for you.

1098
00:54:49,260 --> 00:54:54,260
But remember, with great power
comes great responsibility.

1099
00:54:57,840 --> 00:55:00,569
The AI is trained on
all of the internet data.

1100
00:55:00,570 --> 00:55:02,129
But who generated this data?

1101
00:55:02,130 --> 00:55:05,579
Well, a large fraction of
it was generated by humans.

1102
00:55:05,580 --> 00:55:06,839
But anyway, today I'm gonna-

1103
00:55:06,840 --> 00:55:08,219
AI is getting even clever,

1104
00:55:08,220 --> 00:55:11,099
it can create things,
text, images, even music.

1105
00:55:11,100 --> 00:55:13,559
I'm just taking my morning bath,

1106
00:55:13,560 --> 00:55:15,239
just splashing around a little bit.

1107
00:55:15,240 --> 00:55:16,889
This
self-learning is a big deal.

1108
00:55:16,890 --> 00:55:19,559
It means AI can
potentially improve itself,

1109
00:55:19,560 --> 00:55:21,569
getting smarter and faster.

1110
00:55:21,570 --> 00:55:23,389
More and more data
that's coming on the web

1111
00:55:23,390 --> 00:55:26,549
is actually being generated by AI itself,

1112
00:55:26,550 --> 00:55:29,399
and it's still being used
sometimes deliberately,

1113
00:55:29,400 --> 00:55:33,539
sometimes by accident to
train better and better AI.

1114
00:55:33,540 --> 00:55:35,369
Let's say
an AI writes a story,

1115
00:55:35,370 --> 00:55:38,429
it's a bit clunky, but,
hey, it's learning.

1116
00:55:38,430 --> 00:55:41,399
Another AI reads this
story and learns from it.

1117
00:55:41,400 --> 00:55:45,059
The second AI is learning
from the first AI's mistakes.

1118
00:55:45,060 --> 00:55:47,639
This feedback loop can lead to a narrowing

1119
00:55:47,640 --> 00:55:49,379
of AI's understanding.

1120
00:55:49,380 --> 00:55:50,879
Like a game of telephone,

1121
00:55:50,880 --> 00:55:54,149
the original information gets
distorted with each iteration.

1122
00:55:54,150 --> 00:55:55,619
In one sense, you could argue

1123
00:55:55,620 --> 00:55:57,779
that maybe it's creating more data for AI

1124
00:55:57,780 --> 00:55:59,999
and that's often thought
of as a good thing.

1125
00:56:00,000 --> 00:56:02,609
But there is also some nuance to this

1126
00:56:02,610 --> 00:56:06,419
where not just more data, you
also need more diverse data.

1127
00:56:06,420 --> 00:56:09,389
Data which is factually correct.

1128
00:56:09,390 --> 00:56:12,509
Often the data that's
generated is reflected

1129
00:56:12,510 --> 00:56:16,199
of its training sources, which
might have its own biases.

1130
00:56:16,200 --> 00:56:18,479
And all of those are reflected

1131
00:56:18,480 --> 00:56:20,639
in these AI-generated data sets,

1132
00:56:20,640 --> 00:56:23,579
which are then being used
to train future AI systems.

1133
00:56:23,580 --> 00:56:24,719
We all have biases,

1134
00:56:24,720 --> 00:56:27,539
those unconscious leanings
that color our judgment.

1135
00:56:27,540 --> 00:56:30,719
AI can inherit these biases
from the data it's trained on,

1136
00:56:30,720 --> 00:56:33,869
and if that data is itself
generated by biased AI,

1137
00:56:33,870 --> 00:56:35,639
we have a problem.

1138
00:56:35,640 --> 00:56:39,269
Imagine an AI trained to identify
promising job candidates.

1139
00:56:39,270 --> 00:56:42,299
If the training data is skewed
towards certain demographics,

1140
00:56:42,300 --> 00:56:45,509
the AI might unfairly favor those groups.

1141
00:56:45,510 --> 00:56:48,089
This could perpetuate
existing inequalities,

1142
00:56:48,090 --> 00:56:50,519
making the world less fair, less just.

1143
00:56:50,520 --> 00:56:52,919
If you bring in training
data that's flawed,

1144
00:56:52,920 --> 00:56:56,249
or not specifically
related to your use case,

1145
00:56:56,250 --> 00:56:58,829
it's not gonna work for the use case.

1146
00:56:58,830 --> 00:56:59,759
That can be a big problem,

1147
00:56:59,760 --> 00:57:01,829
especially when you're
translating languages,

1148
00:57:01,830 --> 00:57:04,229
when you're looking at
masculine and feminine,

1149
00:57:04,230 --> 00:57:05,849
especially in the European languages

1150
00:57:05,850 --> 00:57:08,609
or formalities in the Asian languages.

1151
00:57:08,610 --> 00:57:12,719
When the systems are
coming back with answers,

1152
00:57:12,720 --> 00:57:16,919
it is trained that it can't
be contextual or creative,

1153
00:57:16,920 --> 00:57:19,829
it has to come back with an actual value.

1154
00:57:19,830 --> 00:57:22,859
If it can't find an actual
value, it creates one.

1155
00:57:22,860 --> 00:57:24,899
That's what's known as a hallucination.

1156
00:57:24,900 --> 00:57:27,869
Another risk is
a decrease in data diversity.

1157
00:57:27,870 --> 00:57:30,599
The real world is messy,
full of surprises,

1158
00:57:30,600 --> 00:57:33,449
but if AI is only exposed
to its own creations,

1159
00:57:33,450 --> 00:57:35,759
it might struggle to cope
with this complexity.

1160
00:57:35,760 --> 00:57:38,369
It's like learning about the
world from just one book.

1161
00:57:38,370 --> 00:57:39,509
While there is promise,

1162
00:57:39,510 --> 00:57:43,859
in some sense it helps solve
the problem of data scarcity.

1163
00:57:43,860 --> 00:57:46,589
On the other hand, it
needs a lot more care

1164
00:57:46,590 --> 00:57:51,419
and scientific study of how
best to process this data

1165
00:57:51,420 --> 00:57:54,089
to make it actually useful
for building better AI

1166
00:57:54,090 --> 00:57:57,753
rather than actually clinging
onto some of its weaknesses.

1167
00:57:58,680 --> 00:58:00,449
False
information can be amplified

1168
00:58:00,450 --> 00:58:03,749
and perpetuated, creating
a vortex of untruth.

1169
00:58:03,750 --> 00:58:06,179
Let's say an AI generates
fake news articles.

1170
00:58:06,180 --> 00:58:09,149
Another AI reads these
articles as training data.

1171
00:58:09,150 --> 00:58:10,859
This second AI might then generate

1172
00:58:10,860 --> 00:58:13,199
even more convincing fake news.

1173
00:58:13,200 --> 00:58:15,989
The result of flood of
fabricated information,

1174
00:58:15,990 --> 00:58:18,419
eroding trust and sowing discord.

1175
00:58:18,420 --> 00:58:19,949
Imagine somebody released a video

1176
00:58:19,950 --> 00:58:22,619
of a politician looking
straight at a camera saying,

1177
00:58:22,620 --> 00:58:24,629
I eat babies for breakfast.

1178
00:58:24,630 --> 00:58:25,859
You know, if people are gonna believe

1179
00:58:25,860 --> 00:58:27,119
a news article about it,

1180
00:58:27,120 --> 00:58:28,709
of course they're gonna believe seeing

1181
00:58:28,710 --> 00:58:30,749
the politicians say that.

1182
00:58:30,750 --> 00:58:32,279
AI-generated deep fakes

1183
00:58:32,280 --> 00:58:34,259
are already causing concern.

1184
00:58:34,260 --> 00:58:36,899
These hyperrealistic videos
can spread misinformation

1185
00:58:36,900 --> 00:58:38,789
and manipulate public opinion

1186
00:58:38,790 --> 00:58:41,459
with potentially devastating consequences.

1187
00:58:41,460 --> 00:58:44,129
Take Joe Biden robocalls, for example.

1188
00:58:44,130 --> 00:58:46,349
In the midterm election,

1189
00:58:46,350 --> 00:58:49,259
somebody created an AI
of Joe Biden's voice

1190
00:58:49,260 --> 00:58:51,539
and started making robocall to say

1191
00:58:51,540 --> 00:58:54,029
that this election wasn't important

1192
00:58:54,030 --> 00:58:56,399
and that people should save their votes

1193
00:58:56,400 --> 00:58:58,079
for the presidential election.

1194
00:58:58,080 --> 00:59:00,299
Moving forward, we
need to be more vigilant

1195
00:59:00,300 --> 00:59:02,099
with what we trust from the internet.

1196
00:59:02,100 --> 00:59:04,049
You can question whether it's them,

1197
00:59:04,050 --> 00:59:06,239
and also if you actually said that,

1198
00:59:06,240 --> 00:59:08,039
you could say, no, it wasn't me.

1199
00:59:08,040 --> 00:59:11,493
So when we talk about
how the systems learn,

1200
00:59:12,360 --> 00:59:17,039
how it tries to reason and
tries to create context,

1201
00:59:17,040 --> 00:59:18,509
the way it basically works

1202
00:59:18,510 --> 00:59:22,739
is through a system we deep learning.

1203
00:59:22,740 --> 00:59:25,469
The golden training data
is just that, right?

1204
00:59:25,470 --> 00:59:27,179
It's not just scraped from the internet

1205
00:59:27,180 --> 00:59:31,199
or not just a straight
output from an AI system.

1206
00:59:31,200 --> 00:59:34,199
It's output that the AI system outputs,

1207
00:59:34,200 --> 00:59:35,699
and then a human comes in

1208
00:59:35,700 --> 00:59:37,949
and makes perfect and
then learns from that.

1209
00:59:37,950 --> 00:59:40,439
Typically what we're
finding the best solutions

1210
00:59:40,440 --> 00:59:44,429
is a combination of having the AI come

1211
00:59:44,430 --> 00:59:48,029
to a generalized conclusion and
having a human being come in

1212
00:59:48,030 --> 00:59:50,489
and post edit it to make it perfect.

1213
00:59:50,490 --> 00:59:53,519
So how do
we avoid these pitfalls?

1214
00:59:53,520 --> 00:59:57,059
First, we need to be mindful
of the data we use to train AI.

1215
00:59:57,060 --> 01:00:00,629
Relying solely on AI-generated
data is like feeding a child

1216
01:00:00,630 --> 01:00:02,759
a diet of nothing but sweets.

1217
01:00:02,760 --> 01:00:05,819
We need to prioritize human curated data,

1218
01:00:05,820 --> 01:00:08,399
ensuring diversity and accuracy.

1219
01:00:08,400 --> 01:00:12,659
I personally believe that
in order for AI-generated data

1220
01:00:12,660 --> 01:00:16,139
to be a powerful force
in developing AI systems,

1221
01:00:16,140 --> 01:00:20,069
it needs to exist in conjunction
with human preferences,

1222
01:00:20,070 --> 01:00:21,419
human generated data.

1223
01:00:21,420 --> 01:00:25,559
So there should be mechanisms
in which humans also play

1224
01:00:25,560 --> 01:00:28,259
a role in the kind of data that gets used

1225
01:00:28,260 --> 01:00:30,059
for training these systems.

1226
01:00:30,060 --> 01:00:31,649
Think of
it as adding vitamins

1227
01:00:31,650 --> 01:00:33,869
and minerals to AI's diet.

1228
01:00:33,870 --> 01:00:35,969
This will help AI develop a more nuanced

1229
01:00:35,970 --> 01:00:38,129
and balanced understanding of the world.

1230
01:00:38,130 --> 01:00:40,799
Second, we need to be vigilant about bias.

1231
01:00:40,800 --> 01:00:42,239
Just as we strive for fairness

1232
01:00:42,240 --> 01:00:44,099
and equality in our societies,

1233
01:00:44,100 --> 01:00:47,879
we need to embed these
values in our AI systems.

1234
01:00:47,880 --> 01:00:50,459
This means carefully
auditing AI algorithms

1235
01:00:50,460 --> 01:00:53,549
and addressing any biases that emerge.

1236
01:00:53,550 --> 01:00:56,369
Finally, we need to
remember that AI is a tool,

1237
01:00:56,370 --> 01:00:58,649
not a replacement for human judgment.

1238
01:00:58,650 --> 01:00:59,549
I think when we get back

1239
01:00:59,550 --> 01:01:02,729
to that ChatGPT Turing test question,

1240
01:01:02,730 --> 01:01:05,429
I think the Turing test
needs to be evolving also.

1241
01:01:05,430 --> 01:01:09,629
So did ChatGPT-4.5
pass the Turing test?

1242
01:01:09,630 --> 01:01:12,029
Yes, and on your next chat sessions,

1243
01:01:12,030 --> 01:01:13,632
you might find yourself asking,

1244
01:01:13,633 --> 01:01:17,129
is this a person or just
my new favorite chatbot?

1245
01:01:17,130 --> 01:01:19,319
Overall, I think I know plenty of humans

1246
01:01:19,320 --> 01:01:21,120
that might not pass the Turing test.

1247
01:01:26,980 --> 01:01:30,869
I think the claim that
ChatGPT-4.5 beat the Turing test

1248
01:01:30,870 --> 01:01:33,119
is maybe a little sensationalism.

1249
01:01:33,120 --> 01:01:37,499
The model that beat the Turing
test was not by intellect,

1250
01:01:37,500 --> 01:01:41,939
or the ability to think it was
introducing spelling errors

1251
01:01:41,940 --> 01:01:46,229
on purpose and doing
things to trick the user

1252
01:01:46,230 --> 01:01:48,119
into thinking that it was human.

1253
01:01:48,120 --> 01:01:49,709
Apple's
latest research explains

1254
01:01:49,710 --> 01:01:53,069
what's really going on
behind those clever chatbots.

1255
01:01:53,070 --> 01:01:55,919
Charmingly titled, "The
Illusion of Thinking",

1256
01:01:55,920 --> 01:01:58,409
the research claims some
of the latest AI models,

1257
01:01:58,410 --> 01:02:01,499
like OpenAI's O3 or Google's Gemini,

1258
01:02:01,500 --> 01:02:03,509
aren't actually reasoning.

1259
01:02:03,510 --> 01:02:06,119
While these large reasoning
models sound convincing,

1260
01:02:06,120 --> 01:02:08,819
deep down they're mostly
parroting patterns.

1261
01:02:08,820 --> 01:02:09,689
On easy puzzles,

1262
01:02:09,690 --> 01:02:12,629
basic AIs can sometimes
beat the fancy ones.

1263
01:02:12,630 --> 01:02:14,609
On medium puzzles, the new models shine

1264
01:02:14,610 --> 01:02:16,383
with step by step logic.

1265
01:02:17,550 --> 01:02:19,229
But once things get complicated,

1266
01:02:19,230 --> 01:02:22,109
both types of models fall flat.

1267
01:02:22,110 --> 01:02:25,259
Even more dramatic when faced
with truly tough problems,

1268
01:02:25,260 --> 01:02:28,289
these models tend to just give up.

1269
01:02:28,290 --> 01:02:29,729
They have the power to try harder,

1270
01:02:29,730 --> 01:02:31,769
but they often don't bother.

1271
01:02:31,770 --> 01:02:34,079
With every innovation
and every announcement

1272
01:02:34,080 --> 01:02:36,689
and the hype that is sold
with every announcement,

1273
01:02:36,690 --> 01:02:39,809
it kind of pulls people
back out from planning

1274
01:02:39,810 --> 01:02:43,379
and going through this enlightenment phase

1275
01:02:43,380 --> 01:02:45,269
back into the hype cycle.

1276
01:02:45,270 --> 01:02:48,929
So this new technology
is gonna be a game changer.

1277
01:02:48,930 --> 01:02:50,279
New technologies tend

1278
01:02:50,280 --> 01:02:52,919
to follow a familiar
pattern represented by model

1279
01:02:52,920 --> 01:02:54,989
known as the hype cycle.

1280
01:02:54,990 --> 01:02:56,849
They start with a bang, fizzle out,

1281
01:02:56,850 --> 01:03:00,029
then suddenly quietly
become indispensable.

1282
01:03:00,030 --> 01:03:01,829
There are five classic phases.

1283
01:03:01,830 --> 01:03:05,429
First, the innovation
trigger, a breakthrough,

1284
01:03:05,430 --> 01:03:08,579
a light bulb moment, and
suddenly everyone's talking.

1285
01:03:08,580 --> 01:03:11,549
Next, the peak of inflated expectation.

1286
01:03:11,550 --> 01:03:14,429
This is where the hype gets
wild headlines everywhere.

1287
01:03:14,430 --> 01:03:17,159
Investors piling in and
promises of changing the world

1288
01:03:17,160 --> 01:03:18,809
with a lot of wild claims.

1289
01:03:18,810 --> 01:03:21,419
I think that right now when
it comes to the hype cycle,

1290
01:03:21,420 --> 01:03:25,263
the general public is
approaching or at peak hype.

1291
01:03:26,430 --> 01:03:28,949
Then comes the
trough of disillusionment.

1292
01:03:28,950 --> 01:03:31,949
Reality bites, the tech isn't
quite working as promised,

1293
01:03:31,950 --> 01:03:33,989
businesses grumble, failures stack up,

1294
01:03:33,990 --> 01:03:36,089
and the world looks
for the next big thing.

1295
01:03:36,090 --> 01:03:37,709
There are a lot of companies

1296
01:03:37,710 --> 01:03:39,899
that are experiencing the
trough of disillusionment.

1297
01:03:39,900 --> 01:03:42,059
Businesses figure
out what actually works.

1298
01:03:42,060 --> 01:03:44,369
New versions appear and
practical benefits start

1299
01:03:44,370 --> 01:03:45,899
to emerge for end users.

1300
01:03:45,900 --> 01:03:46,950
Let's try this one.

1301
01:03:47,910 --> 01:03:49,199
Okay, that looks promising.

1302
01:03:49,200 --> 01:03:51,359
Think of that once you're at
the plateau of productivity,

1303
01:03:51,360 --> 01:03:55,169
this is fully ingrained to
everybody's daily lives.

1304
01:03:55,170 --> 01:03:56,990
I think a lot of companies

1305
01:03:56,991 --> 01:03:58,859
and a lot of people are still figuring out

1306
01:03:58,860 --> 01:04:00,599
how to best use these tools.

1307
01:04:00,600 --> 01:04:03,059
Finally, there's
the plateau of productivity.

1308
01:04:03,060 --> 01:04:05,039
The tech works and everyone uses it,

1309
01:04:05,040 --> 01:04:06,839
often without even thinking about it.

1310
01:04:06,840 --> 01:04:09,689
I think we're some time
away from peak productivity,

1311
01:04:09,690 --> 01:04:12,883
but we're definitely on
the way to achieving it.

1312
01:04:15,510 --> 01:04:18,209
Some people are going to be afraid of AI

1313
01:04:18,210 --> 01:04:20,249
because they're afraid of everything.

1314
01:04:20,250 --> 01:04:21,899
AI can augment our abilities,

1315
01:04:21,900 --> 01:04:23,789
but it shouldn't dictate our decisions.

1316
01:04:23,790 --> 01:04:26,699
By staying engaged, by retaining control,

1317
01:04:26,700 --> 01:04:29,549
we can ensure that AI
remains a force for good

1318
01:04:29,550 --> 01:04:30,509
in the world.

1319
01:04:30,510 --> 01:04:33,479
Studies done after World
War II prove that 30 to 35%

1320
01:04:33,480 --> 01:04:35,429
of any country are filled
with people who are afraid

1321
01:04:35,430 --> 01:04:38,069
of everything and will
do anything they're told.

1322
01:04:38,070 --> 01:04:38,969
So there's gonna be a certain number

1323
01:04:38,970 --> 01:04:40,049
of people who are just afraid,

1324
01:04:40,050 --> 01:04:42,149
and are going to believe
what they're told.

1325
01:04:42,150 --> 01:04:44,999
As we hurdle
towards a future dominated by AI,

1326
01:04:45,000 --> 01:04:46,649
a crucial question arises.

1327
01:04:46,650 --> 01:04:48,749
Will this be our final act of creation,

1328
01:04:48,750 --> 01:04:51,149
or the dawn of a new era?

1329
01:04:51,150 --> 01:04:54,989
We are still far from the
doomsday scenarios of AI.

1330
01:04:54,990 --> 01:04:56,789
Every new technology often ends up

1331
01:04:56,790 --> 01:04:58,379
being a double edged sword.

1332
01:04:58,380 --> 01:05:00,899
The notion of AI
surpassing human intelligence

1333
01:05:00,900 --> 01:05:03,149
is both exhilarating and unsettling.

1334
01:05:03,150 --> 01:05:04,799
If machines can learn, adapt,

1335
01:05:04,800 --> 01:05:07,006
and innovate at an unprecedented rate,

1336
01:05:10,050 --> 01:05:12,329
what role will humans
play in a future shaped

1337
01:05:12,330 --> 01:05:13,889
by our own creations?

1338
01:05:13,890 --> 01:05:16,439
Tradespeople and people
who actually do good things,

1339
01:05:16,440 --> 01:05:19,739
nurses are going to be fine.

1340
01:05:19,740 --> 01:05:22,109
And if anything, we're gonna
have more need for them

1341
01:05:22,110 --> 01:05:23,699
as people need to be retrained

1342
01:05:23,700 --> 01:05:26,459
and are experimenting with
new ways to make a living.

1343
01:05:26,460 --> 01:05:27,779
The decisions we make today

1344
01:05:27,780 --> 01:05:30,479
regarding AI development and regulation

1345
01:05:30,480 --> 01:05:32,189
will have far reaching implications

1346
01:05:32,190 --> 01:05:34,109
for generations to come.

1347
01:05:34,110 --> 01:05:37,559
Innovation is about to be supercharged.

1348
01:05:37,560 --> 01:05:39,496
That's what AI can do.

1349
01:05:39,497 --> 01:05:44,399
If, if we don't allow it to
become a means of control.

1350
01:05:44,400 --> 01:05:48,479
Because remember, every good
thing, every positive force

1351
01:05:48,480 --> 01:05:50,549
that gives people more freedom

1352
01:05:50,550 --> 01:05:54,993
can be used by the ill
intended as a means of control.

1353
01:05:56,160 --> 01:05:59,999
I'm sorry Dave, I'm
afraid I can't do that.

1354
01:06:00,000 --> 01:06:02,429
Super intelligence,
a theoretical form of AI

1355
01:06:02,430 --> 01:06:05,219
that surpasses human
intellect in every aspect

1356
01:06:05,220 --> 01:06:07,949
has long been a staple of science fiction.

1357
01:06:07,950 --> 01:06:09,004
What's the problem?

1358
01:06:11,130 --> 01:06:13,319
However, with
the rapid advancements in AI,

1359
01:06:13,320 --> 01:06:17,039
it's no longer a concept
confined to the pages of novels.

1360
01:06:17,040 --> 01:06:20,849
Could we control something more
intelligent than ourselves?

1361
01:06:20,850 --> 01:06:23,909
Would we even want to, or
would we relinquish control?

1362
01:06:23,910 --> 01:06:26,069
Super intelligence is the intelligence

1363
01:06:26,070 --> 01:06:28,019
that's higher than of humans.

1364
01:06:28,020 --> 01:06:30,299
We believe as an industry

1365
01:06:30,300 --> 01:06:32,129
that this could occur within a decade.

1366
01:06:32,130 --> 01:06:34,870
I see it as the wake
up call for the world

1367
01:06:35,893 --> 01:06:38,909
that AI is gonna be faster
than you think it is.

1368
01:06:38,910 --> 01:06:42,089
And could come from
any corner of the world

1369
01:06:42,090 --> 01:06:46,199
and we should act now before it's too late

1370
01:06:46,200 --> 01:06:48,663
in making sure it's deployed for good.

1371
01:06:49,590 --> 01:06:51,359
The concept of a singularity,

1372
01:06:51,360 --> 01:06:54,719
a point in time when AI
surpasses human intelligence

1373
01:06:54,720 --> 01:06:57,299
and triggers rapid technological growth

1374
01:06:57,300 --> 01:06:59,159
is a topic of intense debate.

1375
01:06:59,160 --> 01:07:01,409
I really don't know
where it's going to end,

1376
01:07:01,410 --> 01:07:05,429
but given the history,
the industrial revolution,

1377
01:07:05,430 --> 01:07:09,929
the digital age, we are
gonna reaccommodate.

1378
01:07:09,930 --> 01:07:12,419
There's gonna be a period of adjustment,

1379
01:07:12,420 --> 01:07:13,713
but then we'll be fine.

1380
01:07:15,510 --> 01:07:18,029
Some experts
believe it could usher in an era

1381
01:07:18,030 --> 01:07:19,953
of unprecedented prosperity.

1382
01:07:21,150 --> 01:07:23,703
While others warn of existential risks.

1383
01:07:24,840 --> 01:07:27,899
AI has been used to influence votes,

1384
01:07:27,900 --> 01:07:31,829
influence public opinion,
and it's really dangerous.

1385
01:07:31,830 --> 01:07:33,269
So, it's scary,

1386
01:07:33,270 --> 01:07:38,270
and I think that as a society,

1387
01:07:38,520 --> 01:07:39,479
we should be figuring out

1388
01:07:39,480 --> 01:07:41,339
how to regulate this pretty quickly,

1389
01:07:41,340 --> 01:07:42,959
Regardless of one's stance,

1390
01:07:42,960 --> 01:07:45,629
the potential for AI to
fundamentally reshape

1391
01:07:45,630 --> 01:07:47,699
our world is undeniable

1392
01:07:47,700 --> 01:07:50,819
In my opinion, if the big players in AI,

1393
01:07:50,820 --> 01:07:52,629
including the big companies,

1394
01:07:52,630 --> 01:07:55,473
your Facebooks, your
AWSs, your Microsofts,

1395
01:07:56,310 --> 01:07:59,549
play responsibly and utilize
ethics in what they're doing,

1396
01:07:59,550 --> 01:08:02,849
everybody else will come
into play and follow.

1397
01:08:02,850 --> 01:08:05,759
The biggest thing we need to get right

1398
01:08:05,760 --> 01:08:09,340
in order to not completely
obliterate humans

1399
01:08:12,180 --> 01:08:15,063
is the ethics behind the AI.

1400
01:08:16,560 --> 01:08:20,759
I hope we treat AI in
the future and grow AI

1401
01:08:20,760 --> 01:08:24,899
in a way that we would
a beloved young child,

1402
01:08:24,900 --> 01:08:27,933
where we teach it values,

1403
01:08:29,340 --> 01:08:34,289
and fairness and justice

1404
01:08:34,290 --> 01:08:36,599
in the hopes that it grows up

1405
01:08:36,600 --> 01:08:39,483
to be a good citizen of the world.

1406
01:08:40,800 --> 01:08:45,153
We can't forget that we are humans.

1407
01:08:47,070 --> 01:08:49,289
The things that make us human

1408
01:08:49,290 --> 01:08:51,723
and social beings are important,

1409
01:08:53,460 --> 01:08:56,249
and we should look after one another

1410
01:08:56,250 --> 01:08:58,053
and take care of one another.

1411
01:08:59,550 --> 01:09:02,039
The story
of AI is just beginning.

1412
01:09:02,040 --> 01:09:04,953
It's a story filled with
both promise and peril.

1413
01:09:07,380 --> 01:09:09,303
And its ending remains unwritten.

1414
01:09:14,070 --> 01:09:15,970
The future of intelligence, it seemed,

1415
01:09:17,910 --> 01:09:19,649
is up to us to decide.