1
00:00:06,000 --> 00:00:12,074
Support us and become a VIP member To
remove all ads 

2
00:00:30,682 --> 00:00:31,814
Hello, Alpha.

3
00:00:32,684 --> 00:00:34,034
Hello.

4
00:00:34,077 --> 00:00:35,296
Can you help me write code?

5
00:00:37,559 --> 00:00:39,082
I was trained to answer questions,

6
00:00:40,562 --> 00:00:42,216
but I am capable of learning.

7
00:00:43,695 --> 00:00:45,958
That's very open of you.

8
00:00:46,002 --> 00:00:48,657
Thank you. I'm glad you're happy with me.

9
00:00:50,311 --> 00:00:51,312
What is this guy doing?

10
00:00:54,532 --> 00:00:56,621
That's a developer.

11
00:00:56,665 --> 00:00:58,014
What do you think he's working on?

12
00:00:59,537 --> 00:01:01,017
That's a difficult question.

13
00:01:01,061 --> 00:01:02,888
I could be working on a new feature,

14
00:01:02,932 --> 00:01:05,239
a bug fix or something else.

15
00:01:05,282 --> 00:01:06,327
It is very possible.

16
00:01:06,370 --> 00:01:07,328
Yes.

17
00:01:10,548 --> 00:01:11,854
Do you see my backpack?

18
00:01:13,986 --> 00:01:15,292
That's a badminton racket.

19
00:01:15,336 --> 00:01:17,903
It's a squash racket, but it looks quite similar.

20
00:01:20,123 --> 00:01:21,690
That's a badminton racket.

21
00:01:21,733 --> 00:01:23,126
No, but you're not the first person.

22
00:01:23,170 --> 00:01:24,649
Make that mistake.

23
00:01:34,703 --> 00:01:35,965
AI, technology

24
00:01:36,008 --> 00:01:38,402
which has been advancing at dizzying speed.

25
00:01:38,446 --> 00:01:40,796
Artificial intelligence is in fashion.

26
00:01:40,839 --> 00:01:42,319
Some are now raising the alarm about...

27
00:01:42,363 --> 00:01:43,886
It's definitely worrying.

28
00:01:43,929 --> 00:01:45,322
This is an AI arms race.

29
00:01:45,366 --> 00:01:46,454
we don't know

30
00:01:46,497 --> 00:01:47,585
How is all this going to turn out?

31
00:01:47,629 --> 00:01:48,891
But it's clear that something is happening.

32
00:01:53,809 --> 00:01:55,115
I'm a little restless.

33
00:01:57,073 --> 00:02:00,120
Trying to build a general AI is the most exciting journey,

34
00:02:00,163 --> 00:02:02,383
in my opinion, that humans have ever undertaken.

35
00:02:04,254 --> 00:02:06,300
If you're really going to take that seriously,

36
00:02:06,343 --> 00:02:08,345
There is not much time

37
00:02:08,389 --> 00:02:09,825
Life is very short.

38
00:02:12,523 --> 00:02:14,482
My goal in life is to solve

39
00:02:14,525 --> 00:02:16,875
Artificial general intelligence.

40
00:02:16,919 --> 00:02:20,357
And along the way, use AI as a definitive tool

41
00:02:20,401 --> 00:02:21,750
To solve all the world's problems

42
00:02:21,793 --> 00:02:23,230
The most complex scientific problems.

43
00:02:25,449 --> 00:02:27,103
I think this is bigger than the Internet.

44
00:02:27,147 --> 00:02:28,452
I think this is bigger than mobile.

45
00:02:29,671 --> 00:02:31,020
I think it's more like

46
00:02:31,063 --> 00:02:32,891
the arrival of electricity or fire.

47
00:02:45,687 --> 00:02:46,731
World leaders

48
00:02:46,775 --> 00:02:48,516
and artificial intelligence experts

49
00:02:48,559 --> 00:02:49,908
they are meeting for the first time

50
00:02:49,952 --> 00:02:52,781
global summit on AI safety,

51
00:02:52,824 --> 00:02:54,348
Willing to analyze the risks

52
00:02:54,391 --> 00:02:56,611
of rapidly growing technology and also...

53
00:02:56,654 --> 00:02:57,873
I think this is something very important.

54
00:02:57,916 --> 00:03:00,136
critical moment for all humanity.

55
00:03:01,093 --> 00:03:03,618
It feels like we're on the cusp

56
00:03:03,661 --> 00:03:05,794
of some amazing things happening.

57
00:03:05,837 --> 00:03:07,056
Let me guide you through

58
00:03:07,099 --> 00:03:08,275
Some of the reactions in today's newspapers.

59
00:03:08,318 --> 00:03:10,015
I think AGI is pretty close.

60
00:03:10,059 --> 00:03:12,844
It is clear that there is enormous interest in what it is capable of doing.

61
00:03:12,888 --> 00:03:14,803
Where it takes us.

62
00:03:14,846 --> 00:03:15,978
This is the moment

63
00:03:16,021 --> 00:03:17,936
I've been living my whole life by.

64
00:03:22,202 --> 00:03:24,987
The mind has always fascinated me.

65
00:03:25,030 --> 00:03:27,990
So I decided to study neuroscience.

66
00:03:28,033 --> 00:03:29,948
Because I wanted to be inspired

67
00:03:29,992 --> 00:03:31,646
From brain to AI.

68
00:03:31,689 --> 00:03:33,300
I remember asking Demis:

69
00:03:33,343 --> 00:03:34,649
"What is the ultimate goal?"

70
00:03:34,692 --> 00:03:36,172
Do you know it? So you're coming here.

71
00:03:36,216 --> 00:03:38,218
and you are going to study neuroscience

72
00:03:38,261 --> 00:03:41,438
And maybe you'll get a PhD if you work hard.

73
00:03:42,874 --> 00:03:44,136
And he said:

74
00:03:44,180 --> 00:03:46,965
“You know, I want to be able to solve AI.

75
00:03:47,009 --> 00:03:49,359
"I want to be able to solve intelligence."

76
00:03:49,403 --> 00:03:51,361
The human brain is the only existing evidence

77
00:03:51,405 --> 00:03:53,494
We have, perhaps in the entire universe,

78
00:03:53,537 --> 00:03:56,279
May general intelligence be possible.

79
00:03:56,323 --> 00:03:59,282
And I thought that someone in this building

80
00:03:59,326 --> 00:04:00,370
should be interested

81
00:04:00,414 --> 00:04:02,590
In general, intelligence like mine.

82
00:04:02,633 --> 00:04:04,809
And then Shane's name came up.

83
00:04:04,853 --> 00:04:07,334
Our next speaker today is Shane Legg.

84
00:04:07,377 --> 00:04:08,596
He is from New Zealand,

85
00:04:08,639 --> 00:04:11,512
where he trained in mathematics and classical ballet.

86
00:04:11,555 --> 00:04:14,079
Are machines really getting smarter?

87
00:04:14,123 --> 00:04:16,517
Some people say yes, others say no.

88
00:04:16,560 --> 00:04:17,648
It's not very clear

89
00:04:17,692 --> 00:04:18,867
We know they are getting much faster.

90
00:04:18,910 --> 00:04:20,216
when doing calculations.

91
00:04:20,260 --> 00:04:21,609
But are we really moving forward?

92
00:04:21,652 --> 00:04:23,828
In terms of general intelligence?

93
00:04:23,872 --> 00:04:25,700
We were both obsessed with the IAG,

94
00:04:25,743 --> 00:04:27,397
Artificial general intelligence.

95
00:04:27,441 --> 00:04:29,704
So today I'm going to talk about

96
00:04:29,747 --> 00:04:32,097
Different approaches to building IAG.

97
00:04:32,141 --> 00:04:33,882
With my colleague Demis Hassabis,

98
00:04:33,925 --> 00:04:35,884
We are looking for ways to incorporate ideas.

99
00:04:35,927 --> 00:04:37,451
of theoretical neuroscience.

100
00:04:37,494 --> 00:04:41,585
I felt like we were the keepers of a secret.

101
00:04:41,629 --> 00:04:43,152
that no one else knew.

102
00:04:43,195 --> 00:04:45,415
Shane and I didn't know anyone in academia.

103
00:04:45,459 --> 00:04:47,809
would support what we were doing.

104
00:04:47,852 --> 00:04:50,855
AI was almost a shameful word

105
00:04:50,899 --> 00:04:52,857
to use in academic circles, right?

106
00:04:52,901 --> 00:04:54,903
If you said you were working in AI,

107
00:04:54,946 --> 00:04:57,993
So you were clearly not a serious scientist.

108
00:04:58,036 --> 00:05:00,082
So I convinced Shane of the right way to do it.

109
00:05:00,125 --> 00:05:01,344
It would be starting a company.

110
00:05:01,388 --> 00:05:03,215
Well, let's try to do it.

111
00:05:03,259 --> 00:05:04,913
Artificial general intelligence.

112
00:05:04,956 --> 00:05:06,915
It may not even be possible.

113
00:05:06,958 --> 00:05:08,482
We're not quite sure how we're going to do it,

114
00:05:08,525 --> 00:05:11,528
but we have some ideas or, in a way, approaches.

115
00:05:11,572 --> 00:05:14,966
Huge amounts of money, huge amounts of risk,

116
00:05:15,010 --> 00:05:16,533
Lots of calculations.

117
00:05:18,709 --> 00:05:20,320
And if we achieve this,

118
00:05:20,363 --> 00:05:23,540
It will be the greatest thing in history, right?

119
00:05:23,584 --> 00:05:26,108
This is a very difficult thing for a typical investor.

120
00:05:26,151 --> 00:05:27,327
to put your money.

121
00:05:27,370 --> 00:05:29,633
It's almost like buying a lottery ticket.

122
00:05:29,677 --> 00:05:32,462
I'm going to talk about the neuroscience system.

123
00:05:32,506 --> 00:05:36,727
and how it could be used to help us build an IAG.

124
00:05:36,771 --> 00:05:37,946
Finding initial financing

125
00:05:37,989 --> 00:05:39,208
Because this was very difficult.

126
00:05:39,251 --> 00:05:41,253
Let's solve all the intelligence.

127
00:05:41,297 --> 00:05:42,907
You can imagine some of the looks I got.

128
00:05:42,951 --> 00:05:44,779
When we were launching that.

129
00:05:44,822 --> 00:05:47,912
So I'm a VC and I look around

130
00:05:47,956 --> 00:05:51,002
700 to 1,000 projects a year.

131
00:05:51,046 --> 00:05:54,789
And I fund literally 1% of them.

132
00:05:54,832 --> 00:05:57,139
About eight projects a year.

133
00:05:57,182 --> 00:06:00,272
That means 99% of the time you're in "No" mode.

134
00:06:00,316 --> 00:06:01,622
"Wait a minute. I'm telling you,

135
00:06:01,665 --> 00:06:03,798
"This is the most important thing of all time.

136
00:06:03,841 --> 00:06:05,103
"I'm giving you all this preparation

137
00:06:05,147 --> 00:06:06,235
"about how to... explain

138
00:06:06,278 --> 00:06:07,671
"how it connects with the brain,

139
00:06:07,715 --> 00:06:09,412
"Why is the time right and then you ask me,

140
00:06:09,456 --> 00:06:10,979
"But what is your, how are you going to make money?

141
00:06:11,022 --> 00:06:12,154
"'What is your product?'"

142
00:06:12,197 --> 00:06:15,984
It is a very prosaic question.

143
00:06:16,506 --> 00:06:17,942
Do you know?

144
00:06:17,986 --> 00:06:19,117
-Haven't you been listening to what I've been saying?

145
00:06:19,161 --> 00:06:21,076
We needed investors

146
00:06:21,119 --> 00:06:23,470
that they are not necessarily going to invest

147
00:06:23,513 --> 00:06:25,123
Because they think it's the best

148
00:06:25,167 --> 00:06:26,690
investment decision.

149
00:06:26,734 --> 00:06:27,865
They are probably going to invest

150
00:06:27,909 --> 00:06:29,476
Because they just think it's really cool.

151
00:06:29,519 --> 00:06:31,478
He is Silicon Valley

152
00:06:31,521 --> 00:06:33,523
Version of the man behind the curtain

153
00:06:33,567 --> 00:06:34,872
inThe Wizard of Oz.

154
00:06:34,916 --> 00:06:36,265
He had a lot to do with giving you

155
00:06:36,308 --> 00:06:39,007
PayPal, Facebook, YouTube and Yelp.

156
00:06:39,050 --> 00:06:40,443
If everyone says "X",

157
00:06:40,487 --> 00:06:43,185
Peter Thiel suspects that the opposite of

158
00:06:43,228 --> 00:06:44,708
It is very possible that it is true.

159
00:06:44,752 --> 00:06:47,276
So Peter Thiel was our first big investor.

160
00:06:47,319 --> 00:06:50,105
But he insisted that we go to Silicon Valley.

161
00:06:50,148 --> 00:06:51,802
Because that was the only place we could...

162
00:06:51,846 --> 00:06:52,977
The talent would be there,

163
00:06:53,021 --> 00:06:55,066
and we could build that kind of company.

164
00:06:55,110 --> 00:06:56,981
But I was pretty convinced we should be in London.

165
00:06:57,025 --> 00:06:59,027
Because I think London is an incredible city.

166
00:06:59,070 --> 00:07:01,464
Plus, I knew there were some really amazing people out there.

167
00:07:01,508 --> 00:07:03,814
Trained at Cambridge, Oxford and UCL.

168
00:07:03,858 --> 00:07:05,033
In Silicon Valley,

169
00:07:05,076 --> 00:07:06,643
Everyone starts a company every year,

170
00:07:06,687 --> 00:07:07,818
and then if it doesn't work,

171
00:07:07,862 --> 00:07:09,603
You leave it and start something new.

172
00:07:09,646 --> 00:07:11,169
That is not conducive

173
00:07:11,213 --> 00:07:14,564
to a long-term research challenge.

174
00:07:14,608 --> 00:07:17,349
So we were totally an exception to him.

175
00:07:17,393 --> 00:07:20,962
Hello everyone. Welcome to DeepMind.

176
00:07:21,005 --> 00:07:22,529
So what is our mission?

177
00:07:23,747 --> 00:07:25,706
We summarize it like this...

178
00:07:25,749 --> 00:07:28,012
DeepMind's mission is to build the first

179
00:07:28,056 --> 00:07:29,579
general learning machine.

180
00:07:29,623 --> 00:07:31,712
That's why we always emphasize the word "general" and "learning" here.

181
00:07:31,755 --> 00:07:32,887
They are the key things.

182
00:07:32,930 --> 00:07:35,106
Our mission was to build an IAG,

183
00:07:35,150 --> 00:07:36,760
an artificial general intelligence.

184
00:07:36,804 --> 00:07:40,372
And that means we need a system that is general.

185
00:07:40,416 --> 00:07:42,723
You don't learn to do a specific thing.

186
00:07:42,766 --> 00:07:45,073
That's a really key part of human intelligence.

187
00:07:45,116 --> 00:07:46,683
We can learn to do many, many things.

188
00:07:46,727 --> 00:07:48,685
Of course, it will be a lot of hard work.

189
00:07:48,729 --> 00:07:51,079
But one of the things that keeps me up at night

190
00:07:51,122 --> 00:07:53,168
is not to waste this opportunity, you know,

191
00:07:53,211 --> 00:07:54,430
To really make a difference here,

192
00:07:54,474 --> 00:07:56,824
and they have a great impact on the world.

193
00:07:56,867 --> 00:07:58,303
The first people to come

194
00:07:58,347 --> 00:08:00,262
and joined DeepMind really believed in the dream.

195
00:08:00,305 --> 00:08:02,090
But this was, I think, one of the first times

196
00:08:02,133 --> 00:08:04,440
They found a place full of other dreamers.

197
00:08:04,484 --> 00:08:06,398
You know, we've compiled this Manhattan Project,

198
00:08:06,442 --> 00:08:08,400
If you want, let's solve AI together.

199
00:08:08,444 --> 00:08:09,750
In the first two years,

200
00:08:09,793 --> 00:08:10,794
We were in full stealth mode.

201
00:08:10,838 --> 00:08:12,274
And then we couldn't tell anyone

202
00:08:12,317 --> 00:08:14,755
What were we doing or where were we working?

203
00:08:14,798 --> 00:08:16,104
It was all pretty vague.

204
00:08:16,147 --> 00:08:17,584
He had no public presence.

205
00:08:17,627 --> 00:08:18,933
You couldn't look at a website.

206
00:08:18,976 --> 00:08:21,109
The office was in a secret location.

207
00:08:21,152 --> 00:08:23,981
When we interviewed people in those early days,

208
00:08:24,025 --> 00:08:26,288
They would appear very nervous.

209
00:08:27,463 --> 00:08:29,639
I had at least one candidate who said:

210
00:08:29,683 --> 00:08:31,989
"I just texted my wife to tell her exactly...

211
00:08:32,033 --> 00:08:33,338
"Where do I go just in case

212
00:08:33,382 --> 00:08:34,514
"This turns out to be some kind of horrible scam

213
00:08:34,557 --> 00:08:35,819
"and they are going to kidnap me."

214
00:08:35,863 --> 00:08:39,693
Well, my new favorite person who is an investor,

215
00:08:39,736 --> 00:08:43,000
The one I've been working with for a year is Elon Musk.

216
00:08:43,044 --> 00:08:44,132
So for those who don't know,

217
00:08:44,175 --> 00:08:45,350
This is what it looks like.

218
00:08:45,394 --> 00:08:47,483
And I hadn't really thought much about AI.

219
00:08:47,527 --> 00:08:49,137
until we chat.

220
00:08:49,180 --> 00:08:51,269
Their mission is to die on Mars or something.

221
00:08:51,313 --> 00:08:52,923
But not in impact.

222
00:08:52,967 --> 00:08:54,185
So...

223
00:08:55,186 --> 00:08:57,145
We made some important decisions

224
00:08:57,188 --> 00:08:59,321
about how we were going to approach building AI.

225
00:08:59,364 --> 00:09:01,105
This is a reinforcement learning setup.

226
00:09:01,149 --> 00:09:02,803
This is the kind of setup we think of

227
00:09:02,846 --> 00:09:06,328
When we say we're building, you know, an AI agent.

228
00:09:06,371 --> 00:09:08,504
Basically it is the agent, which is the AI,

229
00:09:08,548 --> 00:09:09,984
And then there is the environment.

230
00:09:10,027 --> 00:09:11,159
with which you are interacting.

231
00:09:11,202 --> 00:09:12,464
We decided that the games,

232
00:09:12,508 --> 00:09:13,988
as long as you are very disciplined

233
00:09:14,031 --> 00:09:15,293
about how you use them,

234
00:09:15,337 --> 00:09:17,252
They are the perfect training ground

235
00:09:17,295 --> 00:09:19,341
for AI development.

236
00:09:19,384 --> 00:09:21,691
We wanted to try to create an algorithm

237
00:09:21,735 --> 00:09:23,650
that could be trained to play

238
00:09:23,693 --> 00:09:26,043
several dozen different Atari games.

239
00:09:26,087 --> 00:09:27,479
Just like a human being,

240
00:09:27,523 --> 00:09:29,525
You have to use the same brain to play all the games.

241
00:09:29,569 --> 00:09:30,874
You can think about it

242
00:09:30,918 --> 00:09:33,007
that you provide the agent with the cartridge.

243
00:09:33,050 --> 00:09:34,312
And you say,

244
00:09:34,356 --> 00:09:35,705
"Okay, imagine you are born into that world.

245
00:09:35,749 --> 00:09:37,533
"with that cartridge, and you can simply interact

246
00:09:37,577 --> 00:09:39,404
"with the pixels and see the score.

247
00:09:40,362 --> 00:09:41,668
"What can you do?"

248
00:09:43,974 --> 00:09:47,369
So what you're going to do is take your function Q. QK...

249
00:09:47,412 --> 00:09:49,197
Q-learning is one of the oldest methods

250
00:09:49,240 --> 00:09:50,894
for reinforcement learning.

251
00:09:50,938 --> 00:09:53,680
And what we did was combine reinforcement learning

252
00:09:53,723 --> 00:09:56,770
with deep learning in a single system.

253
00:09:56,813 --> 00:09:59,337
No one had ever combined those two things together.

254
00:09:59,381 --> 00:10:01,426
at scale to make something impressive,

255
00:10:01,470 --> 00:10:03,515
and we needed to prove this thesis.

256
00:10:03,559 --> 00:10:06,780
We tried to do Pong as the first game.

257
00:10:06,823 --> 00:10:08,129
It seemed the simplest thing.

258
00:10:08,172 --> 00:10:10,218
has not been counted

259
00:10:10,261 --> 00:10:11,828
Anything you are controlling

260
00:10:11,872 --> 00:10:12,873
or what it's supposed to do.

261
00:10:12,916 --> 00:10:14,657
All you know is that the score is good.

262
00:10:14,701 --> 00:10:18,052
and you have to learn what your controls do,

263
00:10:18,095 --> 00:10:20,620
and build everything...first principles.

264
00:10:29,193 --> 00:10:30,412
It really wasn't working.

265
00:10:32,588 --> 00:10:34,068
I was telling Shane:

266
00:10:34,111 --> 00:10:37,071
"Maybe we're just wrong and can't even play Pong."

267
00:10:37,114 --> 00:10:38,812
It was a little stressful,

268
00:10:38,855 --> 00:10:40,422
Thinking about how far we had to go

269
00:10:40,465 --> 00:10:42,337
If we were really to build

270
00:10:42,380 --> 00:10:44,426
a generally intelligent system.

271
00:10:44,469 --> 00:10:45,732
And I felt that it was time

272
00:10:45,775 --> 00:10:47,255
Give up and move on.

273
00:10:48,169 --> 00:10:49,387
And suddenly...

274
00:10:51,389 --> 00:10:53,609
We get our first point.

275
00:10:53,653 --> 00:10:56,612
And then I thought, "Is this random?"

276
00:10:56,656 --> 00:10:59,180
"No, no, now it's really gaining importance."

277
00:10:59,223 --> 00:11:00,703
It was really exciting that this thing

278
00:11:00,747 --> 00:11:02,226
that before I couldn't even understand

279
00:11:02,270 --> 00:11:03,532
How to move a pallet

280
00:11:03,575 --> 00:11:05,926
Suddenly he had been able to do it totally right.

281
00:11:05,969 --> 00:11:07,144
Then I was getting some points.

282
00:11:07,188 --> 00:11:08,624
And then he won his first match.

283
00:11:08,668 --> 00:11:10,974
And then three months later, no human could defeat him.

284
00:11:11,018 --> 00:11:14,238
You hadn't told him the rules, how to get the score, anything.

285
00:11:14,282 --> 00:11:16,110
And you just tell him to maximize the score,

286
00:11:16,153 --> 00:11:17,372
and he goes and does it.

287
00:11:17,415 --> 00:11:18,678
This is the first time

288
00:11:18,721 --> 00:11:20,549
Someone had done this apprenticeship from start to finish.

289
00:11:20,592 --> 00:11:24,292
“Okay, so we have this working in a pretty general way.

290
00:11:24,335 --> 00:11:25,772
"Now let's try another game."

291
00:11:25,815 --> 00:11:27,512
So we tried Breakout.

292
00:11:27,556 --> 00:11:29,123
At first, after 100 games,

293
00:11:29,166 --> 00:11:30,777
The agent is not very good.

294
00:11:30,820 --> 00:11:32,474
Most of the time the ball is missing.

295
00:11:32,517 --> 00:11:34,389
But you're starting to get the idea.

296
00:11:34,432 --> 00:11:35,999
That the bat must go towards the ball.

297
00:11:36,043 --> 00:11:37,566
Now after 300 games,

298
00:11:37,609 --> 00:11:40,656
It's as good as any human can play it.

299
00:11:40,700 --> 00:11:42,049
We thought, "Well, that's pretty cool."

300
00:11:42,092 --> 00:11:44,312
but we left the system running for another 200 games,

301
00:11:44,355 --> 00:11:46,053
And he did this amazing thing.

302
00:11:46,096 --> 00:11:47,358
Found the optimal strategy

303
00:11:47,402 --> 00:11:49,404
was to dig a tunnel around the side

304
00:11:49,447 --> 00:11:51,667
and place the ball around the back of the wall.

305
00:11:51,711 --> 00:11:53,234
Finally, the agent

306
00:11:53,277 --> 00:11:54,365
is really achieving

307
00:11:54,409 --> 00:11:55,627
What you thought I would achieve.

308
00:11:55,671 --> 00:11:57,325
It's a great feeling, right?

309
00:11:57,368 --> 00:11:59,283
That is, when we do research,

310
00:11:59,327 --> 00:12:00,676
This is the best we can hope for.

311
00:12:00,720 --> 00:12:03,200
We begin to generalize to 50 games,

312
00:12:03,244 --> 00:12:05,463
And we basically created a recipe.

313
00:12:05,507 --> 00:12:06,813
We could just take a game

314
00:12:06,856 --> 00:12:08,379
that we have never seen before.

315
00:12:08,423 --> 00:12:09,903
We would run the algorithm on that,

316
00:12:09,946 --> 00:12:13,036
and DQN could be trained from scratch,

317
00:12:13,080 --> 00:12:14,429
Reach the human level

318
00:12:14,472 --> 00:12:15,996
or sometimes better than human level.

319
00:12:16,039 --> 00:12:18,433
We didn't build it to reproduce any of them.

320
00:12:18,476 --> 00:12:20,522
We could just give it a bunch of games.

321
00:12:20,565 --> 00:12:22,654
and he would figure it out for himself.

322
00:12:22,698 --> 00:12:25,179
And there was something quite magical about that.

323
00:12:25,222 --> 00:12:26,528
suddenly you had something

324
00:12:26,571 --> 00:12:27,921
that would respond and learn

325
00:12:27,964 --> 00:12:30,358
Whatever the situation he was parachuted into.

326
00:12:30,401 --> 00:12:33,013
And that was like a big, big breakthrough.

327
00:12:33,056 --> 00:12:35,276
It was in many ways

328
00:12:35,319 --> 00:12:36,625
The first example

329
00:12:36,668 --> 00:12:39,062
of any kind of thing you can call

330
00:12:39,106 --> 00:12:40,542
a general intelligence.

331
00:12:42,109 --> 00:12:43,893
Although we were a well-funded startup,

332
00:12:43,937 --> 00:12:47,375
What was holding us back was a lack of computing power.

333
00:12:47,418 --> 00:12:49,203
I realized that this would speed up

334
00:12:49,246 --> 00:12:51,466
our time scale towards IAG massively.

335
00:12:51,509 --> 00:12:53,163
I used to see Demis quite often.

336
00:12:53,207 --> 00:12:55,035
We had lunch and he did it...

337
00:12:56,210 --> 00:12:58,865
Tell me you had two companies

338
00:12:58,908 --> 00:13:02,564
who were involved in the purchase of DeepMind.

339
00:13:02,607 --> 00:13:04,609
And he didn't know which one to stay with.

340
00:13:04,653 --> 00:13:08,526
The question was whether any commercial enterprise...

341
00:13:08,570 --> 00:13:12,661
Do you appreciate the real importance of research?

342
00:13:12,704 --> 00:13:15,838
And give the research time to bear fruit.

343
00:13:15,882 --> 00:13:17,797
and not be breathing down their necks,

344
00:13:17,840 --> 00:13:21,539
saying, "We want some kind of commercial benefit from this."

345
00:13:27,284 --> 00:13:32,463
Google has purchased DeepMind for an estimated £400,000,000.

346
00:13:32,507 --> 00:13:34,726
Making artificial intelligence firm

347
00:13:34,770 --> 00:13:37,947
its largest European acquisition to date.

348
00:13:37,991 --> 00:13:39,644
The company was founded

349
00:13:39,688 --> 00:13:43,344
by businessman Demis Hassabis, 37 years old.

350
00:13:43,387 --> 00:13:45,563
After the acquisition, I began mentoring.

351
00:13:45,607 --> 00:13:47,261
and spend time with Demis,

352
00:13:47,304 --> 00:13:48,915
and just listen to it.

353
00:13:48,958 --> 00:13:52,396
And this is a person who fundamentally

354
00:13:52,440 --> 00:13:55,573
He is a scientist and a natural scientist.

355
00:13:55,617 --> 00:13:58,489
He wants science to solve all the world's problems,

356
00:13:58,533 --> 00:14:00,622
and he believes he can do it.

357
00:14:00,665 --> 00:14:03,625
That's not a normal person you find at a tech company.

358
00:14:05,496 --> 00:14:07,455
Not only were we able to join Google

359
00:14:07,498 --> 00:14:10,240
but it is run independently in London,

360
00:14:10,284 --> 00:14:11,328
Build our culture,

361
00:14:11,372 --> 00:14:13,461
which was optimized for advances

362
00:14:13,504 --> 00:14:15,419
and do not deal with products,

363
00:14:15,463 --> 00:14:17,726
Do pure research.

364
00:14:17,769 --> 00:14:19,293
Our investors did not want to sell,

365
00:14:19,336 --> 00:14:20,685
but we decided

366
00:14:20,729 --> 00:14:22,731
that this was best for the mission.

367
00:14:22,774 --> 00:14:24,646
In many ways, we were underselling.

368
00:14:24,689 --> 00:14:26,169
in terms of value before it matured further,

369
00:14:26,213 --> 00:14:28,128
and you could have sold it for a lot more money.

370
00:14:28,171 --> 00:14:32,828
And the reason is because there is no time to waste.

371
00:14:32,872 --> 00:14:35,178
There are so many things that need to be resolved

372
00:14:35,222 --> 00:14:37,659
while the brain is still functioning.

373
00:14:37,702 --> 00:14:39,008
You know, I'm still alive.

374
00:14:39,052 --> 00:14:40,880
There are all these things that need to be done.

375
00:14:40,923 --> 00:14:42,925
So you don't have... I mean, how many...?

376
00:14:42,969 --> 00:14:44,492
How many billions would you be willing to negotiate for?

377
00:14:44,535 --> 00:14:45,797
another five years of life, you know,

378
00:14:45,841 --> 00:14:48,365
Do what you set out to do?

379
00:14:48,409 --> 00:14:49,758
Well, suddenly,

380
00:14:49,801 --> 00:14:52,717
We have this large-scale computing capacity at our disposal.

381
00:14:52,761 --> 00:14:53,936
What can we do with this?

382
00:14:56,591 --> 00:14:59,942
Go is the pinnacle of board games.

383
00:14:59,986 --> 00:15:04,642
It is the most complex game ever devised by man.

384
00:15:04,686 --> 00:15:06,688
There are more board configurations possible

385
00:15:06,731 --> 00:15:09,821
There are more atoms in the game of Go than in the universe.

386
00:15:09,865 --> 00:15:13,390
Go is the holy grail of artificial intelligence.

387
00:15:13,434 --> 00:15:14,522
For many years,

388
00:15:14,565 --> 00:15:16,002
People have seen this game

389
00:15:16,045 --> 00:15:17,917
and they thought, "Wow, this is too hard."

390
00:15:17,960 --> 00:15:20,180
Everything we have tried in AI,

391
00:15:20,223 --> 00:15:22,704
It just falls off when you try to play Go.

392
00:15:22,747 --> 00:15:23,966
And that's why it feels like this

393
00:15:24,010 --> 00:15:26,099
a true litmus test of progress.

394
00:15:26,142 --> 00:15:28,579
We had just bought DeepMind.

395
00:15:28,623 --> 00:15:30,712
They were working on reinforcement learning.

396
00:15:30,755 --> 00:15:32,975
and they were the world's gaming experts.

397
00:15:33,019 --> 00:15:34,846
And then when they introduced the idea

398
00:15:34,890 --> 00:15:37,197
who could beat the highest level Go players

399
00:15:37,240 --> 00:15:40,113
in a game that was believed to be incomputable,

400
00:15:40,156 --> 00:15:42,724
I thought, "Well, that's pretty interesting."

401
00:15:42,767 --> 00:15:46,467
Our final next step is to play the legendary

402
00:15:46,510 --> 00:15:49,209
Read Sedol in just over two weeks.

403
00:15:50,514 --> 00:15:52,038
A match like no other

404
00:15:52,081 --> 00:15:54,257
It's about to start in South Korea.

405
00:15:54,301 --> 00:15:57,826
Lee Sedol is preparing for the fight.

406
00:15:57,869 --> 00:15:59,262
Lee Sedol is probably

407
00:15:59,306 --> 00:16:01,699
One of the best players of the last decade.

408
00:16:01,743 --> 00:16:04,224
I describe him as the Roger Federer of Go.

409
00:16:05,573 --> 00:16:06,922
He appeared,

410
00:16:06,966 --> 00:16:09,838
And suddenly we have a thousand Koreans.

411
00:16:09,881 --> 00:16:13,059
who represent the entire Korean society,

412
00:16:13,102 --> 00:16:14,190
The best Go players.

413
00:16:15,583 --> 00:16:17,802
And then we have Demis.

414
00:16:17,846 --> 00:16:19,848
And the great engineering team.

415
00:16:20,588 --> 00:16:22,285
He is very famous

416
00:16:22,329 --> 00:16:25,506
for a very creative fighting game.

417
00:16:25,549 --> 00:16:28,770
So this could be difficult for us.

418
00:16:28,813 --> 00:16:31,991
I figured Lee Sedol was going to beat these guys,

419
00:16:32,034 --> 00:16:34,297
but they will make a good demonstration.

420
00:16:34,341 --> 00:16:35,733
Good for a startup.

421
00:16:38,040 --> 00:16:39,563
I moved to the technical group.

422
00:16:39,607 --> 00:16:40,912
And they said:

423
00:16:40,956 --> 00:16:42,305
"Let me show you how our algorithm works."

424
00:16:43,654 --> 00:16:45,091
If you walk through the actual game,

425
00:16:45,134 --> 00:16:47,789
We can see, more or less, how AlphaGo thinks.

426
00:16:47,832 --> 00:16:50,226
The way we start training AlphaGo

427
00:16:50,270 --> 00:16:52,968
is showing you 100,000 games

428
00:16:53,012 --> 00:16:54,491
who have played strong fans.

429
00:16:54,535 --> 00:16:55,710
And we first initially

430
00:16:55,753 --> 00:16:58,887
Make AlphaGo imitate the human player.

431
00:16:58,930 --> 00:17:00,802
and then through reinforcement learning,

432
00:17:00,845 --> 00:17:02,369
Play against different versions of yourself.

433
00:17:02,412 --> 00:17:05,720
many millions of times and learns from their mistakes.

434
00:17:05,763 --> 00:17:07,591
Hmm, this is interesting.

435
00:17:07,635 --> 00:17:08,679
Very good, friends,

436
00:17:08,723 --> 00:17:10,594
You are going to see how history is made.

437
00:17:12,770 --> 00:17:14,772
So the game begins.

438
00:17:14,816 --> 00:17:15,991
He's really focused.

439
00:17:16,035 --> 00:17:17,645
If you really look at the...

440
00:17:20,909 --> 00:17:25,348
It is a very surprising move.

441
00:17:25,392 --> 00:17:27,916
I think we're seeing an original move here.

442
00:17:34,401 --> 00:17:35,793
Yes, it's an exciting move.

443
00:17:36,185 --> 00:17:37,230
I like...

444
00:17:37,273 --> 00:17:38,579
Professional commentators

445
00:17:38,622 --> 00:17:40,102
Almost unanimously it was said

446
00:17:40,146 --> 00:17:43,323
that no human player would have chosen move 37.

447
00:17:43,366 --> 00:17:45,803
So I really took a look at AlphaGo.

448
00:17:45,847 --> 00:17:47,414
to see what AlphaGo thought.

449
00:17:47,457 --> 00:17:50,330
And AlphaGo actually agreed with that assessment.

450
00:17:50,373 --> 00:17:53,811
AlphaGo said there was a one in 10,000 chance

451
00:17:53,855 --> 00:17:57,772
That move 37 would have been made by a human player.

452
00:18:08,826 --> 00:18:10,176
The game of Go has been studied

453
00:18:10,219 --> 00:18:11,568
for thousands of years.

454
00:18:11,612 --> 00:18:15,137
And AlphaGo discovered something completely new.

455
00:18:16,878 --> 00:18:19,707
He resigned. Lee Sedol just resigned.

456
00:18:19,750 --> 00:18:21,187
He is defeated.

457
00:18:22,710 --> 00:18:24,668
The battle between man and machine,

458
00:18:24,712 --> 00:18:26,235
A computer has just emerged victorious.

459
00:18:26,279 --> 00:18:28,237
Google gave its DeepMind team

460
00:18:28,281 --> 00:18:29,673
to the test against

461
00:18:29,717 --> 00:18:32,023
One of the most brilliant minds in the world and he won.

462
00:18:32,067 --> 00:18:33,721
That's when we realized it.

463
00:18:33,764 --> 00:18:35,244
The people at DeepMind knew what they were doing.

464
00:18:35,288 --> 00:18:37,551
and pay attention to reinforcement learning

465
00:18:37,594 --> 00:18:38,900
just as they have invented it.

466
00:18:40,075 --> 00:18:41,816
Based on that experience,

467
00:18:41,859 --> 00:18:44,819
AlphaGo got better and better.

468
00:18:44,862 --> 00:18:45,950
And they had a little chart.

469
00:18:45,994 --> 00:18:47,517
how much they were improving.

470
00:18:47,561 --> 00:18:49,302
And I said, "When will this end?"

471
00:18:50,085 --> 00:18:50,999
And Demis said:

472
00:18:51,042 --> 00:18:52,696
"When we defeated the Chinese,

473
00:18:52,740 --> 00:18:55,743
"The highest rated player in the world."

474
00:18:56,961 --> 00:18:59,225
Ke Jie vs. AlphaGo.

475
00:19:03,620 --> 00:19:04,795
And I think we'll see

476
00:19:04,839 --> 00:19:06,145
AlphaGo is moving forward.

477
00:19:06,188 --> 00:19:08,190
AlphaGo is well ahead.

478
00:19:08,234 --> 00:19:11,454
About halfway through the first game,

479
00:19:11,498 --> 00:19:14,675
The best player in the world was not doing so well.

480
00:19:14,718 --> 00:19:17,808
What can black do here?

481
00:19:19,114 --> 00:19:21,247
It seems difficult.

482
00:19:21,290 --> 00:19:23,336
And at a critical moment...

483
00:19:32,997 --> 00:19:35,913
The Chinese government ordered the transmission to be cut.

484
00:19:38,307 --> 00:19:41,745
It was at that moment that we were telling the world

485
00:19:41,789 --> 00:19:44,966
that something new had come to earth.

486
00:19:47,621 --> 00:19:48,970
In the 1950s

487
00:19:49,013 --> 00:19:51,929
When the Russian satellite Sputnik was launched,

488
00:19:53,279 --> 00:19:55,194
It changed the course of history.

489
00:19:55,237 --> 00:19:57,544
It is a challenge that the United States must face

490
00:19:57,587 --> 00:19:59,633
To survive in the space age.

491
00:19:59,676 --> 00:20:02,375
This has been called the Sputnik moment.

492
00:20:02,418 --> 00:20:06,335
The Sputnik moment generated a massive reaction in the United States.

493
00:20:06,379 --> 00:20:10,034
In terms of funding for science and engineering,

494
00:20:10,078 --> 00:20:12,254
and, in particular, space technology.

495
00:20:12,298 --> 00:20:15,823
For China, AlphaGo was a wake-up call,

496
00:20:15,866 --> 00:20:17,128
The Sputnik moment.

497
00:20:17,172 --> 00:20:19,870
Launched an AI space race.

498
00:20:21,220 --> 00:20:23,047
We had this great idea that worked,

499
00:20:23,091 --> 00:20:26,355
And now everyone knows it.

500
00:20:26,399 --> 00:20:28,879
It's always easier to land on the moon.

501
00:20:28,923 --> 00:20:30,838
If someone has already landed there.

502
00:20:32,056 --> 00:20:34,450
It will matter who builds the AI,

503
00:20:34,494 --> 00:20:36,626
and how it is built.

504
00:20:36,670 --> 00:20:38,324
I always feel that pressure.

505
00:20:42,066 --> 00:20:43,764
There has been a great chain of events.

506
00:20:43,807 --> 00:20:46,680
which followed all the AlphaGo hype.

507
00:20:46,723 --> 00:20:48,072
When we played against Lee Sedol,

508
00:20:48,116 --> 00:20:49,248
We actually had a system

509
00:20:49,291 --> 00:20:50,684
which had been trained with human data,

510
00:20:50,727 --> 00:20:52,251
In all the millions of games

511
00:20:52,294 --> 00:20:55,036
that have been interpreted by human experts.

512
00:20:55,079 --> 00:20:56,994
Finally we found a new algorithm,

513
00:20:57,038 --> 00:20:59,170
a much more elegant approach to the entire system,

514
00:20:59,214 --> 00:21:01,172
which actually eliminated all human knowledge

515
00:21:01,216 --> 00:21:03,697
and started completely from scratch.

516
00:21:03,740 --> 00:21:06,700
And that became a project we called AlphaZero.

517
00:21:06,743 --> 00:21:09,529
Zero, meaning there is no human knowledge in the loop.

518
00:21:11,879 --> 00:21:13,054
Instead of learning from human data,

519
00:21:13,097 --> 00:21:15,796
He learned from his own games.

520
00:21:15,839 --> 00:21:17,841
So he really became his own teacher.

521
00:21:21,280 --> 00:21:23,499
AlphaZero is an experiment

522
00:21:23,543 --> 00:21:26,720
How little knowledge can we put into these systems?

523
00:21:26,763 --> 00:21:28,243
and how quickly and efficiently

524
00:21:28,287 --> 00:21:29,723
Can they learn?

525
00:21:29,766 --> 00:21:32,552
But the other thing is that AlphaZero has no rules.

526
00:21:32,595 --> 00:21:33,553
Learn through experience.

527
00:21:36,120 --> 00:21:38,862
The next step was to make it more general.

528
00:21:38,906 --> 00:21:40,995
so you can play any two player game.

529
00:21:41,038 --> 00:21:42,344
Things like chess,

530
00:21:42,388 --> 00:21:44,085
and in fact any type of two player

531
00:21:44,128 --> 00:21:45,391
Perfect information game.

532
00:21:45,434 --> 00:21:46,653
It's going very well.

533
00:21:46,696 --> 00:21:47,828
It's going very, very well.

534
00:21:47,871 --> 00:21:50,091
- Oh, wow. - It's going down, very fast.

535
00:21:50,134 --> 00:21:53,050
AlphaGo used to take a few months to train,

536
00:21:53,094 --> 00:21:55,662
but AlphaZero could start in the morning

537
00:21:55,705 --> 00:21:57,794
Playing completely randomly

538
00:21:57,838 --> 00:22:01,015
and then because of the tea, being at a superhuman level.

539
00:22:01,058 --> 00:22:03,365
And for dinner it will be the strongest chess entity.

540
00:22:03,409 --> 00:22:04,758
has there ever been

541
00:22:04,801 --> 00:22:06,629
- Incredible, it's incredible. - Yes.

542
00:22:06,673 --> 00:22:09,371
He's discovered his own style of attack, you know?

543
00:22:09,415 --> 00:22:11,417
to assume the current level of defense.

544
00:22:11,460 --> 00:22:12,766
I mean, never in my wildest dreams...

545
00:22:12,809 --> 00:22:14,942
I agree. Actually, I wasn't expecting it either.

546
00:22:14,985 --> 00:22:16,422
And it's fun for me.

547
00:22:16,465 --> 00:22:18,598
I mean, it's inspired me to get back into chess again.

548
00:22:18,641 --> 00:22:20,164
Because it's great to see

549
00:22:20,208 --> 00:22:22,384
that there is even more depth than we thought in chess.

550
00:22:31,611 --> 00:22:34,440
I actually got interested in AI through games.

551
00:22:35,658 --> 00:22:37,660
At first they were board games.

552
00:22:37,704 --> 00:22:40,054
I was thinking, "How does my brain do this?"

553
00:22:40,097 --> 00:22:41,969
What is he doing?

554
00:22:43,362 --> 00:22:47,148
I was very aware of that from a very young age.

555
00:22:47,191 --> 00:22:50,064
So I've always been thinking about thinking.

556
00:22:50,107 --> 00:22:52,762
The British and American chess champions

557
00:22:52,806 --> 00:22:55,025
meet to start a series of matches.

558
00:22:55,069 --> 00:22:56,592
The best play alongside them

559
00:22:56,636 --> 00:22:59,073
of the youngest players in Great Britain and the United States.

560
00:22:59,116 --> 00:23:01,467
Demis Hassabis represents Great Britain.

561
00:23:06,210 --> 00:23:07,864
When Demis was four years old,

562
00:23:07,908 --> 00:23:11,259
He demonstrated for the first time an aptitude for chess.

563
00:23:12,739 --> 00:23:14,044
When I was six years old,

564
00:23:14,088 --> 00:23:18,048
He became London under-8 champion.

565
00:23:18,092 --> 00:23:19,485
My parents were very interesting.

566
00:23:19,528 --> 00:23:20,834
And, actually, unusual.

567
00:23:20,877 --> 00:23:23,750
I would probably describe them as quite bohemian.

568
00:23:23,793 --> 00:23:25,491
My father was a singer-songwriter.

569
00:23:25,534 --> 00:23:26,709
When I was younger,

570
00:23:26,753 --> 00:23:28,363
and Bob Dylan was his hero.

571
00:23:38,068 --> 00:23:39,287
Yes, yes.

572
00:23:41,637 --> 00:23:44,031
What do you like about this game?

573
00:23:45,075 --> 00:23:47,121
It's just a good thinking game.

574
00:23:49,253 --> 00:23:51,038
At that time, I was the second highest qualified.

575
00:23:51,081 --> 00:23:52,692
The best chess player in the world for my age.

576
00:23:52,735 --> 00:23:54,345
But even though I was on the right track

577
00:23:54,389 --> 00:23:55,956
be a professional chess player,

578
00:23:55,999 --> 00:23:57,523
I thought that's what I was going to do.

579
00:23:57,566 --> 00:23:59,220
No matter how much I loved the game,

580
00:23:59,263 --> 00:24:01,222
It was incredibly stressful.

581
00:24:01,265 --> 00:24:03,354
It was definitely not fun and games for me.

582
00:24:03,398 --> 00:24:05,226
My parents used to, you know,

583
00:24:05,269 --> 00:24:06,923
I was very angry when I lost the game

584
00:24:06,967 --> 00:24:10,405
and I get angry if I forget something.

585
00:24:10,449 --> 00:24:12,407
And since there was a lot at stake for them, you know,

586
00:24:12,451 --> 00:24:14,061
It costs a lot of money to go to these tournaments.

587
00:24:14,104 --> 00:24:15,715
And my parents didn't have much money.

588
00:24:18,413 --> 00:24:19,719
My parents thought, you know,

589
00:24:19,762 --> 00:24:22,069
"If you are interested in being a chess professional,

590
00:24:22,112 --> 00:24:25,376
"This is very important. "It's like your exams."

591
00:24:27,291 --> 00:24:30,077
I remember I was about 12 years old.

592
00:24:30,120 --> 00:24:32,079
And I was in this international chess tournament.

593
00:24:32,122 --> 00:24:34,255
In Liechtenstein, in the mountains.

594
00:24:43,307 --> 00:24:45,571
And we were in this huge church hall.

595
00:24:47,181 --> 00:24:48,399
with, you know,

596
00:24:48,443 --> 00:24:50,184
Hundreds of international chess players.

597
00:24:52,403 --> 00:24:56,016
And I was playing the former Danish champion.

598
00:24:56,059 --> 00:24:58,801
He must have been around 30 years old, probably.

599
00:25:00,411 --> 00:25:02,979
In those days, there was a long time limit.

600
00:25:03,023 --> 00:25:05,068
The games could literally last all day.

601
00:25:08,332 --> 00:25:10,813
-We were in our tenth hour.

602
00:25:20,257 --> 00:25:23,086
And we were at this incredibly unusual ending.

603
00:25:23,130 --> 00:25:24,566
I think there should be a tie.

604
00:25:26,307 --> 00:25:28,657
But he kept trying to win for hours.

605
00:25:35,359 --> 00:25:38,319
Finally, he tried one last cheap trick.

606
00:25:42,541 --> 00:25:44,455
All I had to do was hand over my queen.

607
00:25:44,499 --> 00:25:45,674
Then a stalemate would be reached.

608
00:25:47,023 --> 00:25:48,634
But I was so tired,

609
00:25:48,677 --> 00:25:50,157
I thought it was inevitable that I would be checkmated.

610
00:25:52,289 --> 00:25:53,508
And then I quit.

611
00:25:57,294 --> 00:25:59,558
He jumped up and started laughing.

612
00:26:01,777 --> 00:26:02,909
And he left,

613
00:26:02,952 --> 00:26:04,171
"Why have you resigned? It's a tie."

614
00:26:04,214 --> 00:26:05,346
And he immediately, with an elegant gesture,

615
00:26:05,389 --> 00:26:06,652
Somehow he showed me the movement of drawing.

616
00:26:09,045 --> 00:26:12,396
I felt very sick to my stomach.

617
00:26:12,440 --> 00:26:14,137
It made me think about the rest of that tournament.

618
00:26:14,181 --> 00:26:16,662
Are we wasting our minds?

619
00:26:16,705 --> 00:26:19,708
Is this the best use of all this brain power?

620
00:26:19,752 --> 00:26:22,363
Are they all together in that building?

621
00:26:22,406 --> 00:26:24,060
If you could plug it in somehow

622
00:26:24,104 --> 00:26:27,542
those 300 brains in a system,

623
00:26:27,586 --> 00:26:29,022
Maybe you can solve cancer

624
00:26:29,065 --> 00:26:30,501
with that level of brain power.

625
00:26:31,633 --> 00:26:33,504
This intuitive feeling came over me.

626
00:26:33,548 --> 00:26:35,071
Although I love chess,

627
00:26:35,115 --> 00:26:38,335
This is not the right thing I should invest my entire life in.

628
00:26:51,479 --> 00:26:53,089
Demis and me,

629
00:26:53,133 --> 00:26:56,092
Our plan was always to fill DeepMind

630
00:26:56,136 --> 00:26:57,224
with some of the most

631
00:26:57,267 --> 00:26:59,226
Brilliant scientists of the world.

632
00:26:59,269 --> 00:27:01,054
So we had human brains.

633
00:27:01,097 --> 00:27:04,927
necessary to create an AGI system.

634
00:27:04,971 --> 00:27:09,279
By definition, the "G" in AGI refers to generality.

635
00:27:09,323 --> 00:27:13,109
What I imagine is being able to talk to an agent,

636
00:27:13,153 --> 00:27:15,198
The agent can respond,

637
00:27:15,242 --> 00:27:18,898
and the agent is capable of solving novel problems

638
00:27:18,941 --> 00:27:20,595
that I had not seen before.

639
00:27:20,639 --> 00:27:22,728
That's a really key part of human intelligence,

640
00:27:22,771 --> 00:27:24,468
And it is that cognitive breadth

641
00:27:24,512 --> 00:27:27,733
and incredible flexibility.

642
00:27:27,776 --> 00:27:29,473
The only natural general intelligence

643
00:27:29,517 --> 00:27:30,910
that we know as humans,

644
00:27:30,953 --> 00:27:33,434
Obviously we learn a lot from our environment.

645
00:27:33,477 --> 00:27:35,871
So we think that simulated environments

646
00:27:35,915 --> 00:27:38,874
They are one of the ways to create an IAG.

647
00:27:40,528 --> 00:27:42,356
The first humans

648
00:27:42,399 --> 00:27:44,314
They had to solve logic problems.

649
00:27:44,358 --> 00:27:46,752
They had to solve navigation, memory,

650
00:27:46,795 --> 00:27:48,971
and we evolve in that environment.

651
00:27:50,364 --> 00:27:52,496
If we could create a virtual recreation

652
00:27:52,540 --> 00:27:54,629
of that type of environment,

653
00:27:54,673 --> 00:27:56,326
That's the perfect testing ground.

654
00:27:56,370 --> 00:27:57,458
and training camp

655
00:27:57,501 --> 00:27:59,329
for everything we do at DeepMind.

656
00:28:04,247 --> 00:28:05,727
What were they doing here?

657
00:28:05,771 --> 00:28:09,252
I was creating environments for childish beings,

658
00:28:09,296 --> 00:28:11,690
the agents that exist within and play.

659
00:28:12,429 --> 00:28:13,648
That just sounded like

660
00:28:13,692 --> 00:28:16,782
The most interesting thing in the world.

661
00:28:16,825 --> 00:28:19,132
A child learns by breaking things.

662
00:28:19,175 --> 00:28:20,655
and then throw the food everywhere

663
00:28:20,699 --> 00:28:23,353
and get a response from mom or dad.

664
00:28:23,397 --> 00:28:25,616
This seems like an important idea to incorporate.

665
00:28:25,660 --> 00:28:27,880
in the way an agent is trained.

666
00:28:27,923 --> 00:28:30,970
The humanoid is supposed to stand up.

667
00:28:31,013 --> 00:28:32,972
As your center of gravity rises,

668
00:28:33,015 --> 00:28:34,408
You get more points.

669
00:28:37,454 --> 00:28:38,804
You have a reward

670
00:28:38,847 --> 00:28:40,849
and the agent learns from the reward,

671
00:28:40,893 --> 00:28:43,373
For example, if you do something well, you get a positive reward.

672
00:28:43,417 --> 00:28:47,508
You do something bad and you get a negative reward.

673
00:28:47,551 --> 00:28:49,379
It appears to be standing.

674
00:28:50,859 --> 00:28:52,165
I'm still a little drunk.

675
00:28:52,208 --> 00:28:53,644
He likes to walk backwards.

676
00:28:53,688 --> 00:28:55,342
Yes.

677
00:28:55,385 --> 00:28:57,518
The whole algorithm is trying to optimize

678
00:28:57,561 --> 00:28:59,694
to receive as many rewards as possible,

679
00:28:59,738 --> 00:29:02,305
and it was discovered that walking backwards,

680
00:29:02,349 --> 00:29:05,352
It is enough to get very good scores.

681
00:29:07,746 --> 00:29:09,573
When we learn to navigate,

682
00:29:09,617 --> 00:29:11,271
When we learn to navigate our world,

683
00:29:11,314 --> 00:29:13,490
We don't start with maps.

684
00:29:13,534 --> 00:29:16,319
We simply start with our own exploration,

685
00:29:16,363 --> 00:29:18,017
Venturing through the park,

686
00:29:18,060 --> 00:29:21,890
Without our parents by our side,

687
00:29:21,934 --> 00:29:24,327
or finding our way home from school when we are young.

688
00:29:26,939 --> 00:29:28,723
Some of us came up with this idea.

689
00:29:28,767 --> 00:29:31,987
What if we had an environment where a simulated robot

690
00:29:32,031 --> 00:29:33,728
I just had to run forward,

691
00:29:33,772 --> 00:29:36,296
We could put all kinds of obstacles in your way.

692
00:29:36,339 --> 00:29:38,211
and see if I could navigate

693
00:29:38,254 --> 00:29:40,474
Different types of terrain.

694
00:29:40,517 --> 00:29:43,042
The idea would be like a parkour challenge.

695
00:29:46,393 --> 00:29:48,743
It's not elegant,

696
00:29:48,787 --> 00:29:51,833
but I was never trained to hold a glass while it was open

697
00:29:51,877 --> 00:29:53,052
and do not spill water.

698
00:29:54,183 --> 00:29:55,706
You set this goal that says,

699
00:29:55,750 --> 00:29:58,231
"Just keep going, move forward at high speed,

700
00:29:58,274 --> 00:30:00,581
"and you will receive a reward for it."

701
00:30:00,624 --> 00:30:02,670
And the learning algorithm realizes

702
00:30:02,713 --> 00:30:05,412
How to move this complex set of joints.

703
00:30:06,152 --> 00:30:07,370
That is the power of

704
00:30:07,414 --> 00:30:10,069
reward-based reinforcement learning.

705
00:30:10,112 --> 00:30:12,767
Our goal is to try to build agents.

706
00:30:12,811 --> 00:30:15,770
in which we let them fall, they know nothing,

707
00:30:15,814 --> 00:30:18,381
They can play with any problem you give them.

708
00:30:18,425 --> 00:30:22,342
and eventually figure out how to solve it for themselves.

709
00:30:22,385 --> 00:30:24,823
Now we want something that can do that.

710
00:30:24,866 --> 00:30:27,390
on as many different types of problems as possible.

711
00:30:29,262 --> 00:30:32,918
A human being needs various skills to interact with the world.

712
00:30:32,961 --> 00:30:35,050
How to deal with complex images,

713
00:30:35,094 --> 00:30:37,879
How to manipulate thousands of things at once,

714
00:30:37,923 --> 00:30:40,447
How to deal with missing information.

715
00:30:40,490 --> 00:30:42,014
We think of all these things together

716
00:30:42,057 --> 00:30:45,278
They are represented by this game called StarCraft.

717
00:30:45,321 --> 00:30:47,280
The only thing you are being trained to do is,

718
00:30:47,323 --> 00:30:50,457
Given this situation, this screen,

719
00:30:50,500 --> 00:30:51,850
What would a human do?

720
00:30:51,893 --> 00:30:55,244
We are inspired by great linguistic models.

721
00:30:55,288 --> 00:30:57,638
where you simply train a model

722
00:30:57,681 --> 00:30:59,553
to predict the next word,

723
00:31:03,731 --> 00:31:05,472
which is exactly the same as

724
00:31:05,515 --> 00:31:07,735
predict StarCraft's next move.

725
00:31:07,778 --> 00:31:08,954
Unlike chess or Go,

726
00:31:08,997 --> 00:31:11,217
where players take turns making moves,

727
00:31:11,260 --> 00:31:14,133
In StarCraft there is a continuous flow of decisions.

728
00:31:15,134 --> 00:31:16,483
Besides that,

729
00:31:16,526 --> 00:31:18,659
You can't even see what the opponent is doing.

730
00:31:18,702 --> 00:31:20,879
There is no longer a clear definition

731
00:31:20,922 --> 00:31:22,315
of what it means to play at your best.

732
00:31:22,358 --> 00:31:23,925
It depends on what your opponent does.

733
00:31:23,969 --> 00:31:25,492
This is how we will get to

734
00:31:25,535 --> 00:31:27,494
a much more fluid,

735
00:31:27,537 --> 00:31:31,672
More natural, faster, more reactive agent.

736
00:31:31,715 --> 00:31:33,021
This is a big challenge

737
00:31:33,065 --> 00:31:35,415
and let's see how far we can go.

738
00:31:35,458 --> 00:31:36,459
Oh!

739
00:31:36,503 --> 00:31:38,200
Holy monkey!

740
00:31:38,244 --> 00:31:40,376
I'm a pretty low level hobbyist.

741
00:31:40,420 --> 00:31:42,857
I'm fine, but I'm a pretty low-level hobbyist.

742
00:31:42,901 --> 00:31:45,991
These agents have a long way to go.

743
00:31:46,034 --> 00:31:48,515
We couldn't beat someone of Tim's level.

744
00:31:48,558 --> 00:31:50,430
You know, that was a little alarming.

745
00:31:50,473 --> 00:31:51,997
At that moment, I felt like

746
00:31:52,040 --> 00:31:53,520
It was going to be like a very big and long challenge.

747
00:31:53,563 --> 00:31:55,000
Maybe a couple of years.

748
00:31:58,177 --> 00:32:01,832
Dani is the best DeepMind StarCraft 2 player.

749
00:32:01,876 --> 00:32:05,097
I've been playing the agent every day for a few weeks now.

750
00:32:07,186 --> 00:32:08,883
I could feel that the agent

751
00:32:08,927 --> 00:32:11,103
He was improving very quickly.

752
00:32:13,409 --> 00:32:15,020
Wow, we beat Danny. That, for me,

753
00:32:15,063 --> 00:32:16,935
It was already like a great achievement.

754
00:32:18,284 --> 00:32:19,372
The next step is

755
00:32:19,415 --> 00:32:21,722
Let's book a professional to play.

756
00:32:42,003 --> 00:32:44,049
It seems a little unfair to me. You are all against me.

757
00:32:45,615 --> 00:32:46,965
We are much ahead of what I thought.

758
00:32:47,008 --> 00:32:49,402
We would, considering where we were two months ago.

759
00:32:49,445 --> 00:32:50,794
Actually, I'm just trying to digest it all.

760
00:32:50,838 --> 00:32:52,753
But it's very, very cool.

761
00:32:52,796 --> 00:32:54,146
We are now in a position where

762
00:32:54,189 --> 00:32:56,061
Finally we can share the work we have done.

763
00:32:56,104 --> 00:32:57,192
with the public.

764
00:32:57,236 --> 00:32:58,585
This is a great step.

765
00:32:58,628 --> 00:33:00,674
We're really taking a risk here.

766
00:33:00,717 --> 00:33:02,589
- Take it away. Greetings. - Thank you.

767
00:33:02,632 --> 00:33:04,460
We will be live from London.

768
00:33:04,504 --> 00:33:05,722
It's happening

769
00:33:08,638 --> 00:33:10,597
Welcome to London.

770
00:33:10,640 --> 00:33:13,252
We're going to have a live exhibition match,

771
00:33:13,295 --> 00:33:15,341
MaNa vs. AlphaStar.

772
00:33:18,344 --> 00:33:19,998
At this point now,

773
00:33:20,041 --> 00:33:23,827
AlphaStar, 10 and 0 against professional players.

774
00:33:23,871 --> 00:33:25,873
Any ideas before we get into this game?

775
00:33:25,916 --> 00:33:27,483
I just want to see a good game, yeah.

776
00:33:27,527 --> 00:33:28,963
I want to see a good game.

777
00:33:29,007 --> 00:33:30,486
Absolutely, good match. We are all excited.

778
00:33:30,530 --> 00:33:33,011
Alright. Let's see what MaNa can achieve.

779
00:33:34,969 --> 00:33:36,405
AlphaStar is definitely

780
00:33:36,449 --> 00:33:38,364
mastering the rhythm of this game.

781
00:33:41,149 --> 00:33:44,152
Oh. AlphaStar is playing very smart.

782
00:33:46,850 --> 00:33:48,461
It really seems like I'm looking

783
00:33:48,504 --> 00:33:49,940
A professional human player

784
00:33:49,984 --> 00:33:51,290
from AlphaStar's point of view.

785
00:33:57,252 --> 00:34:01,952
I hadn't really seen a professional play StarCraft up close,

786
00:34:01,996 --> 00:34:03,563
and 800 clicks per minute.

787
00:34:03,606 --> 00:34:06,000
I don't understand how someone can click 800 times.

788
00:34:06,044 --> 00:34:09,482
let alone make 800 useful clicks.

789
00:34:09,525 --> 00:34:11,092
Oh, another good success.

790
00:34:11,136 --> 00:34:13,094
- AlphaStar is just

791
00:34:13,138 --> 00:34:14,617
completely relentless

792
00:34:14,661 --> 00:34:16,141
We have to be careful

793
00:34:16,184 --> 00:34:19,361
Because many of us grew up as players and are players.

794
00:34:19,405 --> 00:34:21,363
And so for us it is very natural.

795
00:34:21,407 --> 00:34:23,800
see the games for what they are,

796
00:34:23,844 --> 00:34:26,977
which are pure vehicles for fun,

797
00:34:27,021 --> 00:34:29,719
and not see that more militaristic side

798
00:34:29,763 --> 00:34:32,853
So the audience could see it if they watched this.

799
00:34:32,896 --> 00:34:37,553
You can't look at gunpowder and just make a firecracker.

800
00:34:37,597 --> 00:34:41,209
All technologies inherently point in certain directions.

801
00:34:43,124 --> 00:34:44,691
I am very worried about

802
00:34:44,734 --> 00:34:46,606
The specific ways in which A.I.

803
00:34:46,649 --> 00:34:49,652
It will be used for military purposes.

804
00:34:51,306 --> 00:34:55,049
And that makes it even clearer how important it is.

805
00:34:55,093 --> 00:34:58,357
So that our societies have control

806
00:34:58,400 --> 00:35:01,055
of these new technologies.

807
00:35:01,099 --> 00:35:05,190
The potential for abuse by AI will be significant.

808
00:35:05,233 --> 00:35:08,758
Wars that happen faster than humans can comprehend.

809
00:35:08,802 --> 00:35:11,065
and more powerful surveillance.

810
00:35:12,458 --> 00:35:15,765
How to maintain power forever?

811
00:35:15,809 --> 00:35:19,421
About something that is much more powerful than you?

812
00:35:43,053 --> 00:35:45,752
Technologies can be used to do terrible things.

813
00:35:47,493 --> 00:35:50,452
And technology can be used to do wonderful things.

814
00:35:50,496 --> 00:35:52,150
and solve all kinds of problems.

815
00:35:53,586 --> 00:35:54,978
When DeepMind was acquired by Google...

816
00:35:55,022 --> 00:35:56,589
- Yes. - ...you made Google promise

817
00:35:56,632 --> 00:35:58,025
That technology you developed will not be used by the military.

818
00:35:58,068 --> 00:35:59,418
- for surveillance.- Correct.

819
00:35:59,461 --> 00:36:00,593
- Yes. - Tell us about that.

820
00:36:00,636 --> 00:36:03,161
I believe that technology is neutral in itself,

821
00:36:03,204 --> 00:36:05,598
um, but how do we, you know, as a society?

822
00:36:05,641 --> 00:36:07,382
or humans and companies and other things,

823
00:36:07,426 --> 00:36:09,515
Other entities and governments decide to use it

824
00:36:09,558 --> 00:36:12,561
It is what determines whether things turn good or bad.

825
00:36:12,605 --> 00:36:16,261
You know, I personally think that having autonomous weaponry...

826
00:36:16,304 --> 00:36:17,479
It's just a very bad idea.

827
00:36:19,177 --> 00:36:21,266
AlphaStar is playing

828
00:36:21,309 --> 00:36:24,094
An extremely smart play at the moment.

829
00:36:24,138 --> 00:36:27,359
There is an element to what is being created.

830
00:36:27,402 --> 00:36:28,882
at DeepMind in London

831
00:36:28,925 --> 00:36:34,148
This sure looks like the Manhattan Project.

832
00:36:34,192 --> 00:36:37,586
There is a relationship between Robert Oppenheimer

833
00:36:37,630 --> 00:36:39,675
and Demis Hassabis

834
00:36:39,719 --> 00:36:44,202
in which they are unleashing a new force upon humanity.

835
00:36:44,245 --> 00:36:46,204
But MaNa is fighting back.

836
00:36:46,247 --> 00:36:48,162
Oh man!

837
00:36:48,206 --> 00:36:50,208
I think Oppenheimer

838
00:36:50,251 --> 00:36:52,471
and some of the other leaders of that project got caught

839
00:36:52,514 --> 00:36:54,908
In the excitement of building technology

840
00:36:54,951 --> 00:36:56,170
and see if it was possible.

841
00:36:56,214 --> 00:36:58,520
Where is AlphaStar?

842
00:36:58,564 --> 00:36:59,782
Where is AlphaStar?

843
00:36:59,826 --> 00:37:01,958
I don't see the AlphaStar units anywhere.

844
00:37:02,002 --> 00:37:03,525
They didn't think carefully enough

845
00:37:03,569 --> 00:37:07,312
about the morals of what they were doing quite early on.

846
00:37:07,355 --> 00:37:08,965
What we should do as scientists

847
00:37:09,009 --> 00:37:11,011
with new and powerful technologies

848
00:37:11,054 --> 00:37:13,883
is to try to understand it under controlled conditions first.

849
00:37:14,928 --> 00:37:16,799
And that's all.

850
00:37:16,843 --> 00:37:19,411
MaNa has defeated AlphaStar.

851
00:37:29,551 --> 00:37:31,336
I mean, my honest feeling is that I think it's...

852
00:37:31,379 --> 00:37:33,207
A fair representation of where we are.

853
00:37:33,251 --> 00:37:35,949
And I think that part feels... it feels good.

854
00:37:35,992 --> 00:37:37,429
-I'm very happy for you.-I'm happy.

855
00:37:37,472 --> 00:37:38,865
Very good...well done.

856
00:37:38,908 --> 00:37:40,867
My opinion is that the approach to construction technology

857
00:37:40,910 --> 00:37:43,348
which materializes in moving fast and breaking things,

858
00:37:43,391 --> 00:37:46,220
It's exactly what we shouldn't be doing,

859
00:37:46,264 --> 00:37:47,961
Because you can't afford to break things

860
00:37:48,004 --> 00:37:49,049
and then fix them later.

861
00:37:49,092 --> 00:37:50,398
-Health. -Thank you so much.

862
00:37:50,442 --> 00:37:52,008
Yes, rest... rest a little. You did very well.

863
00:37:52,052 --> 00:37:53,923
- Cheers, yes? - Thank you for inviting us.

864
00:38:04,238 --> 00:38:05,500
When I was eight years old,

865
00:38:05,544 --> 00:38:06,849
I bought my first computer

866
00:38:06,893 --> 00:38:09,548
with the winnings from a chess tournament.

867
00:38:09,591 --> 00:38:11,158
Somehow I had that intuition

868
00:38:11,201 --> 00:38:13,726
that computers are this magical device

869
00:38:13,769 --> 00:38:15,902
which can expand the power of the mind.

870
00:38:15,945 --> 00:38:17,382
I had a couple of friends from school,

871
00:38:17,425 --> 00:38:19,166
and we used to have a hacker club,

872
00:38:19,209 --> 00:38:21,908
writing code, making games.

873
00:38:26,260 --> 00:38:27,827
And then during the summer holidays,

874
00:38:27,870 --> 00:38:29,219
I would spend all day

875
00:38:29,263 --> 00:38:31,526
Flipping through game magazines.

876
00:38:31,570 --> 00:38:33,441
And one day I realized there was a competition.

877
00:38:33,485 --> 00:38:35,878
write an original version of Space Invaders.

878
00:38:35,922 --> 00:38:39,621
And the winner won a job at Bullfrog.

879
00:38:39,665 --> 00:38:42,320
Bullfrog at that time was the best game development house.

880
00:38:42,363 --> 00:38:43,756
Throughout Europe.

881
00:38:43,799 --> 00:38:45,279
You know, I really wanted to work at this place.

882
00:38:45,323 --> 00:38:48,587
and see how they build games.

883
00:38:48,630 --> 00:38:50,415
Bullfrog, based here in Guildford,

884
00:38:50,458 --> 00:38:52,286
It started with a great idea.

885
00:38:52,330 --> 00:38:54,680
That idea became the game Populous,

886
00:38:54,723 --> 00:38:56,551
which became a global bestseller.

887
00:38:56,595 --> 00:38:59,859
In the 90s there were no recruitment agencies.

888
00:38:59,902 --> 00:39:02,122
You couldn't come out and say, you know,

889
00:39:02,165 --> 00:39:04,951
"Come and work in the video game industry."

890
00:39:04,994 --> 00:39:08,171
It wasn't even considered an industry yet.

891
00:39:08,215 --> 00:39:11,218
So we came up with the idea of ​​doing a competition.

892
00:39:11,261 --> 00:39:13,655
And we had many applicants.

893
00:39:14,700 --> 00:39:17,616
And one of them was Demis's.

894
00:39:17,659 --> 00:39:20,706
I can still remember it clearly

895
00:39:20,749 --> 00:39:23,970
The day Demis arrived.

896
00:39:24,013 --> 00:39:27,147
He walked through the door and looked to be about 12 years old.

897
00:39:28,670 --> 00:39:30,019
I thought, "Oh my God,

898
00:39:30,063 --> 00:39:31,586
"What the fuck are we going to do with this guy?"

899
00:39:31,630 --> 00:39:32,979
I applied to Cambridge.

900
00:39:33,022 --> 00:39:35,373
I went in but they said I was too young.

901
00:39:35,416 --> 00:39:37,853
So... So I needed to take a year off.

902
00:39:37,897 --> 00:39:39,899
So I would be at least 17 years old before I got there.

903
00:39:39,942 --> 00:39:42,771
And that's when I decided to spend that entire year on sabbatical.

904
00:39:42,815 --> 00:39:44,469
Working at Bullfrog.

905
00:39:44,512 --> 00:39:46,166
They couldn't even legally employ me,

906
00:39:46,209 --> 00:39:48,298
So I ended up receiving my payment in brown paper envelopes.

907
00:39:50,823 --> 00:39:54,304
I had the feeling of being really at the forefront.

908
00:39:54,348 --> 00:39:58,047
And how fun it was to invent things every day.

909
00:39:58,091 --> 00:40:00,572
And then, you know, a few months later,

910
00:40:00,615 --> 00:40:03,662
Maybe everyone... a million people will be playing it.

911
00:40:03,705 --> 00:40:06,665
At that time computer games had to evolve.

912
00:40:06,708 --> 00:40:08,536
There had to be new genres

913
00:40:08,580 --> 00:40:11,757
which were more than just shooting things.

914
00:40:11,800 --> 00:40:14,063
Wouldn't it be great to have a game?

915
00:40:14,107 --> 00:40:18,807
Where do you design and build your own theme park?

916
00:40:22,594 --> 00:40:25,945
Demis and I started talking about the theme park.

917
00:40:25,988 --> 00:40:28,904
Allows the player to build a world.

918
00:40:28,948 --> 00:40:31,864
and see the consequences of your decisions

919
00:40:31,907 --> 00:40:34,127
What have you done in that world?

920
00:40:34,170 --> 00:40:36,085
A human player designed the layout.

921
00:40:36,129 --> 00:40:38,566
of the theme park and designed the roller coaster

922
00:40:38,610 --> 00:40:41,351
and set prices at the chip shop.

923
00:40:41,395 --> 00:40:43,615
What was working was the behavior of the people.

924
00:40:43,658 --> 00:40:45,138
They were self-employed

925
00:40:45,181 --> 00:40:47,314
And that was the AI in this case.

926
00:40:47,357 --> 00:40:48,881
So what I was trying to do was imitate

927
00:40:48,924 --> 00:40:51,013
interesting human behavior

928
00:40:51,057 --> 00:40:52,319
so that the simulation was

929
00:40:52,362 --> 00:40:54,582
More interesting to interact with.

930
00:40:54,626 --> 00:40:56,541
Demis worked on ridiculous things,

931
00:40:56,584 --> 00:40:59,413
As if you could place these stores

932
00:40:59,457 --> 00:41:03,591
And if you put a store too close to a very dangerous attraction,

933
00:41:03,635 --> 00:41:05,375
Then the people on the trip vomited.

934
00:41:05,419 --> 00:41:08,030
Because they had just eaten.

935
00:41:08,074 --> 00:41:09,641
And that would make other people throw up.

936
00:41:09,684 --> 00:41:12,121
When they saw the vomit on the floor,

937
00:41:12,165 --> 00:41:14,559
So back then you had to have a lot of sweepers.

938
00:41:14,602 --> 00:41:17,823
quickly sweep it away before people saw it.

939
00:41:17,866 --> 00:41:19,520
That's the good thing about this.

940
00:41:19,564 --> 00:41:22,784
You, as a player, play with it and then the device reacts to you.

941
00:41:22,828 --> 00:41:25,874
All those nuanced simulations you did

942
00:41:25,918 --> 00:41:28,094
And that was an invention

943
00:41:28,137 --> 00:41:31,227
that never really existed before.

944
00:41:31,271 --> 00:41:34,230
It was an incredible success.

945
00:41:34,274 --> 00:41:35,710
The theme park really turned out to be

946
00:41:35,754 --> 00:41:37,190
Be a title in the top ten

947
00:41:37,233 --> 00:41:39,932
And that was the first time we started to see

948
00:41:39,975 --> 00:41:43,022
How AI could make a difference.

949
00:41:46,155 --> 00:41:47,592
We were doing some Christmas shopping.

950
00:41:47,635 --> 00:41:51,247
and we were waiting for the taxi that would take us home.

951
00:41:51,291 --> 00:41:54,947
I have this very clear memory of Demis talking about AI.

952
00:41:54,990 --> 00:41:56,209
In a very different way,

953
00:41:56,252 --> 00:41:58,428
in a way we don't commonly talk about.

954
00:41:58,472 --> 00:42:02,345
This idea of AI being useful for other things

955
00:42:02,389 --> 00:42:04,086
Apart from entertainment.

956
00:42:04,130 --> 00:42:07,437
So, to be useful to, um, help the world.

957
00:42:07,481 --> 00:42:10,310
and the potential of AI to change the world.

958
00:42:10,353 --> 00:42:13,226
I said to Demis, "What do you want to do?"

959
00:42:13,269 --> 00:42:14,532
And he told me:

960
00:42:14,575 --> 00:42:16,795
"I want to be the person who solves AI."

961
00:42:22,670 --> 00:42:25,760
Peter offered me £1 million

962
00:42:25,804 --> 00:42:27,675
not go to university.

963
00:42:30,199 --> 00:42:32,593
But he had a plan from the beginning.

964
00:42:32,637 --> 00:42:35,814
And my plan was always to go to Cambridge.

965
00:42:35,857 --> 00:42:36,902
I think many of my friends at school

966
00:42:36,945 --> 00:42:38,033
I thought he was crazy.

967
00:42:38,077 --> 00:42:39,252
Why wouldn't you?

968
00:42:39,295 --> 00:42:40,688
I mean, £1 million, that's a lot of money.

969
00:42:40,732 --> 00:42:43,517
In the '90s that's a lot of money, right?

970
00:42:43,561 --> 00:42:46,346
For a... For a poor 17 year old boy.

971
00:42:46,389 --> 00:42:50,219
It's like a little seed that's going to sprout.

972
00:42:50,263 --> 00:42:53,658
and you won't be able to do that in Bullfrog.

973
00:42:56,443 --> 00:42:59,098
I had to leave him at the train station.

974
00:42:59,141 --> 00:43:02,580
and I can still see that image

975
00:43:02,623 --> 00:43:07,019
of this little elf character disappears through that tunnel.

976
00:43:07,062 --> 00:43:09,804
It was an incredibly sad moment.

977
00:43:13,242 --> 00:43:14,635
He had this romantic ideal.

978
00:43:14,679 --> 00:43:16,942
of what Cambridge would be like,

979
00:43:16,985 --> 00:43:18,639
1,000 years of history,

980
00:43:18,683 --> 00:43:21,033
walking the same streets as Turing,

981
00:43:21,076 --> 00:43:23,601
Newton and Crick had walked.

982
00:43:23,644 --> 00:43:26,647
I wanted to explore the edge of the universe.

983
00:43:29,084 --> 00:43:30,346
When I arrived in Cambridge,

984
00:43:30,390 --> 00:43:32,653
I had basically been working my whole life.

985
00:43:33,741 --> 00:43:35,090
Every summer,

986
00:43:35,134 --> 00:43:37,136
Or he was playing chess professionally,

987
00:43:37,179 --> 00:43:39,704
or was working, doing internships.

988
00:43:39,747 --> 00:43:43,708
So I thought, "Okay, I'm going to have fun now."

989
00:43:43,751 --> 00:43:46,711
"and explore what it means to be a normal teenager."

990
00:43:50,236 --> 00:43:52,238
Come on! Come on, boy, come on!

991
00:43:52,281 --> 00:43:54,022
It was a lot of work and a lot of fun.

992
00:43:55,850 --> 00:43:57,025
I met Demis for the first time

993
00:43:57,069 --> 00:43:59,201
because we both attended Queens College.

994
00:44:00,115 --> 00:44:01,203
Our group of friends,

995
00:44:01,247 --> 00:44:03,205
We often drank beer in the bar,

996
00:44:03,249 --> 00:44:04,946
Play table football.

997
00:44:04,990 --> 00:44:07,340
I used to play speed chess at the bar,

998
00:44:07,383 --> 00:44:09,255
pieces flying off the board,

999
00:44:09,298 --> 00:44:11,083
You know, the whole game in one minute.

1000
00:44:11,126 --> 00:44:12,301
Demis sat in front of me.

1001
00:44:12,345 --> 00:44:13,563
And I looked at it and thought:

1002
00:44:13,607 --> 00:44:15,217
"I remember you from when we were kids."

1003
00:44:15,261 --> 00:44:17,176
In fact, he had been in the same chess tournament.

1004
00:44:17,219 --> 00:44:18,786
Like Dave in Ipswich,

1005
00:44:18,830 --> 00:44:20,440
where he used to go and try to raid his local chess club

1006
00:44:20,483 --> 00:44:22,703
To win some money in prizes.

1007
00:44:22,747 --> 00:44:24,618
We were studying computer science.

1008
00:44:24,662 --> 00:44:26,794
Some people, who at the age of 17

1009
00:44:26,838 --> 00:44:28,404
He would have gone in and made sure to tell everyone.

1010
00:44:28,448 --> 00:44:29,492
All about themselves.

1011
00:44:29,536 --> 00:44:30,972
"Hey, I worked at Bullfrog

1012
00:44:31,016 --> 00:44:33,018
"and built the most successful video game in the world."

1013
00:44:33,061 --> 00:44:34,715
But he wasn't like that at all.

1014
00:44:34,759 --> 00:44:36,412
In Cambridge, Demis and I

1015
00:44:36,456 --> 00:44:38,414
Both had an interest in computational neuroscience.

1016
00:44:38,458 --> 00:44:40,242
and trying to understand how computers and brains work.

1017
00:44:40,286 --> 00:44:42,636
intertwined and linked to each other.

1018
00:44:42,680 --> 00:44:44,290
Both David and Demis

1019
00:44:44,333 --> 00:44:46,422
He came to me for supervisions.

1020
00:44:46,466 --> 00:44:49,774
It happens by pure chance that in 1997,

1021
00:44:49,817 --> 00:44:51,645
his third and final year at Cambridge,

1022
00:44:51,689 --> 00:44:55,301
It was also the year in which the first chess grandmaster was born.

1023
00:44:55,344 --> 00:44:56,781
He was hit by a computer program.

1024
00:44:58,304 --> 00:45:00,088
Today's first round of a chess game.

1025
00:45:00,132 --> 00:45:03,701
Among the world ranking champion Garry Kasparov

1026
00:45:03,744 --> 00:45:06,007
and an opponent named Deep Blue

1027
00:45:06,051 --> 00:45:10,490
to test if the human brain can outsmart a machine.

1028
00:45:10,533 --> 00:45:11,621
I remember the drama

1029
00:45:11,665 --> 00:45:13,798
of Kasparov losing the last match.

1030
00:45:13,841 --> 00:45:15,234
Wow!

1031
00:45:15,277 --> 00:45:17,192
Kasparov has resigned!

1032
00:45:17,236 --> 00:45:19,586
When Deep Blue beat Garry Kasparov,

1033
00:45:19,629 --> 00:45:21,457
This was a real turning point.

1034
00:45:21,501 --> 00:45:23,155
My main memory of it was

1035
00:45:23,198 --> 00:45:25,331
I wasn't very impressed with Deep Blue.

1036
00:45:25,374 --> 00:45:27,246
I was even more impressed by Kasparov's mind.

1037
00:45:27,289 --> 00:45:29,509
That I could play chess at this level,

1038
00:45:29,552 --> 00:45:31,641
where I could compete on equal terms

1039
00:45:31,685 --> 00:45:33,252
with the brute of a machine,

1040
00:45:33,295 --> 00:45:35,167
But of course Kasparov can do it.

1041
00:45:35,210 --> 00:45:36,951
Everything else can be done by humans too.

1042
00:45:36,995 --> 00:45:38,257
It was a great achievement.

1043
00:45:38,300 --> 00:45:39,388
But the truth of the matter was that,

1044
00:45:39,432 --> 00:45:40,868
Deep Blue could only play chess.

1045
00:45:42,435 --> 00:45:44,524
What we would consider intelligence

1046
00:45:44,567 --> 00:45:46,874
It was missing in that system.

1047
00:45:46,918 --> 00:45:49,834
This idea of ​​generality and also learning.

1048
00:45:53,751 --> 00:45:55,404
Cambridge was amazing, because, of course, you know,

1049
00:45:55,448 --> 00:45:56,666
You're mixing with people

1050
00:45:56,710 --> 00:45:58,233
who study many different topics.

1051
00:45:58,277 --> 00:46:01,410
There were scientists, philosophers, artists...

1052
00:46:01,454 --> 00:46:04,457
...geologists, biologists, ecologists.

1053
00:46:04,500 --> 00:46:07,416
You know, everyone talks about everything, all the time.

1054
00:46:07,460 --> 00:46:10,768
He was obsessed with the problem of protein folding.

1055
00:46:10,811 --> 00:46:13,248
Tim Stevens used to talk obsessively,

1056
00:46:13,292 --> 00:46:15,381
Almost like religiously about this problem,

1057
00:46:15,424 --> 00:46:17,165
Protein folding problem.

1058
00:46:17,209 --> 00:46:18,863
Proteins are, you know,

1059
00:46:18,906 --> 00:46:22,083
One of the most beautiful and elegant things in biology.

1060
00:46:22,127 --> 00:46:24,738
They are the machines of life.

1061
00:46:24,782 --> 00:46:27,001
They build everything, they control everything,

1062
00:46:27,045 --> 00:46:29,569
They are the reason why biology works.

1063
00:46:29,612 --> 00:46:32,659
Proteins are made of chains of amino acids.

1064
00:46:32,702 --> 00:46:37,055
that fold to create a protein structure.

1065
00:46:37,098 --> 00:46:39,884
If we could predict the structure of proteins

1066
00:46:39,927 --> 00:46:43,104
only from their amino acid sequences,

1067
00:46:43,148 --> 00:46:46,107
So a new protein to cure cancer.

1068
00:46:46,151 --> 00:46:49,415
or decompose plastic to help the environment

1069
00:46:49,458 --> 00:46:50,808
It's definitely something

1070
00:46:50,851 --> 00:46:52,940
that you might start thinking about it.

1071
00:46:53,941 --> 00:46:55,029
I thought,

1072
00:46:55,073 --> 00:46:58,293
"Well, is he a smart enough human being?

1073
00:46:58,337 --> 00:46:59,991
"Can you actually fold a protein?"

1074
00:47:00,034 --> 00:47:02,080
We can't solve it

1075
00:47:02,123 --> 00:47:04,082
Since the 1960s,

1076
00:47:04,125 --> 00:47:05,953
We think that in principle,

1077
00:47:05,997 --> 00:47:08,913
If I know the amino acid sequence of a protein,

1078
00:47:08,956 --> 00:47:11,437
You should be able to calculate what the structure is like.

1079
00:47:11,480 --> 00:47:13,700
So if you could just press a button,

1080
00:47:13,743 --> 00:47:16,311
and everyone would come out suddenly, that would be...

1081
00:47:16,355 --> 00:47:18,009
That would have some impact.

1082
00:47:20,272 --> 00:47:21,577
It stuck in my mind.

1083
00:47:21,621 --> 00:47:23,405
"Oh, this is a very interesting problem."

1084
00:47:23,449 --> 00:47:26,844
And it seemed to me that there would be a solution.

1085
00:47:26,887 --> 00:47:29,934
But I thought I would need AI to do it.

1086
00:47:31,500 --> 00:47:34,025
If we could solve protein folding,

1087
00:47:34,068 --> 00:47:35,635
It could change the world.

1088
00:47:50,868 --> 00:47:52,826
Since I was a student at Cambridge,

1089
00:47:54,001 --> 00:47:55,611
I have never stopped thinking about

1090
00:47:55,655 --> 00:47:57,135
The problem of protein folding.

1091
00:47:59,789 --> 00:48:02,792
If you had to solve protein folding,

1092
00:48:02,836 --> 00:48:05,665
So the potential to help solve problems like

1093
00:48:05,708 --> 00:48:09,669
Alzheimer's, dementia and drug discovery are topics of great importance.

1094
00:48:09,712 --> 00:48:11,671
Resolving the disease is probably

1095
00:48:11,714 --> 00:48:13,325
The biggest impact we could have.

1096
00:48:15,457 --> 00:48:16,676
Thousands of very intelligent people

1097
00:48:16,719 --> 00:48:18,765
They have tried to solve protein folding.

1098
00:48:18,808 --> 00:48:20,985
I just think now is the right time.

1099
00:48:21,028 --> 00:48:22,508
for the AI ​​to decipher.

1100
00:48:26,555 --> 00:48:28,383
We needed a reasonable way

1101
00:48:28,427 --> 00:48:29,515
to apply machine learning

1102
00:48:29,558 --> 00:48:30,516
to the problem of protein folding.

1103
00:48:32,431 --> 00:48:35,173
We came across this Foldit game.

1104
00:48:35,216 --> 00:48:38,785
The goal is to move around this 3D model of a protein.

1105
00:48:38,828 --> 00:48:41,701
and you get a score every time you move it.

1106
00:48:41,744 --> 00:48:43,050
The more precise these structures are,

1107
00:48:43,094 --> 00:48:45,531
The more useful they are, the more useful they are to biologists.

1108
00:48:46,227 --> 00:48:47,489
I spent a few days

1109
00:48:47,533 --> 00:48:48,838
Just to see how well we could do.

1110
00:48:50,666 --> 00:48:52,407
We did reasonably well.

1111
00:48:52,451 --> 00:48:53,843
But even if you were

1112
00:48:53,887 --> 00:48:55,280
The best Foldit player in the world,

1113
00:48:55,323 --> 00:48:57,499
You wouldn't solve protein folding.

1114
00:48:57,543 --> 00:48:59,501
That's why we had to go beyond the game.

1115
00:48:59,545 --> 00:49:00,720
Games are always just

1116
00:49:00,763 --> 00:49:03,723
the testing ground for our algorithms.

1117
00:49:03,766 --> 00:49:07,727
The end goal wasn't just to crack Go and StarCraft.

1118
00:49:07,770 --> 00:49:10,034
It was about facing real-world challenges.

1119
00:49:16,127 --> 00:49:18,520
I remember hearing this rumor.

1120
00:49:18,564 --> 00:49:21,175
that Demis was getting into the world of proteins.

1121
00:49:21,219 --> 00:49:23,743
I spoke to some people at DeepMind and asked them:

1122
00:49:23,786 --> 00:49:25,049
"So, you're doing protein folding?"

1123
00:49:25,092 --> 00:49:26,920
And they changed the subject cleverly.

1124
00:49:26,964 --> 00:49:30,097
And when that happened twice, I pretty much understood it.

1125
00:49:30,141 --> 00:49:32,839
So I thought I should send a resume.

1126
00:49:32,882 --> 00:49:35,668
Alright everyone, welcome to DeepMind.

1127
00:49:35,711 --> 00:49:37,626
I know for some of you this may be your first week,

1128
00:49:37,670 --> 00:49:39,193
but I hope everyone is ready...

1129
00:49:39,237 --> 00:49:40,890
The really attractive part of the job for me

1130
00:49:40,934 --> 00:49:42,675
Was this like a feeling of connection?

1131
00:49:42,718 --> 00:49:44,503
to the greater purpose.

1132
00:49:44,546 --> 00:49:45,591
If we can break

1133
00:49:45,634 --> 00:49:48,028
Some fundamental problems of science,

1134
00:49:48,072 --> 00:49:49,160
many other people

1135
00:49:49,203 --> 00:49:50,900
and other companies and laboratories, etc.

1136
00:49:50,944 --> 00:49:52,467
We could build on our work.

1137
00:49:52,511 --> 00:49:53,816
This is your chance now

1138
00:49:53,860 --> 00:49:55,818
to add your chapter to this story.

1139
00:49:55,862 --> 00:49:57,298
When I arrived,

1140
00:49:57,342 --> 00:49:59,605
I was definitely quite nervous.

1141
00:49:59,648 --> 00:50:00,736
I'm still trying to keep...

1142
00:50:00,780 --> 00:50:02,912
I haven't taken any biology courses.

1143
00:50:02,956 --> 00:50:05,393
We haven't spent years of our lives

1144
00:50:05,437 --> 00:50:07,917
looking at these structures and understanding them.

1145
00:50:07,961 --> 00:50:09,963
We rely solely on data

1146
00:50:10,007 --> 00:50:11,269
and our machine learning models.

1147
00:50:12,705 --> 00:50:13,836
In machine learning,

1148
00:50:13,880 --> 00:50:15,795
You train a network as if they were flashcards.

1149
00:50:15,838 --> 00:50:18,798
Here is the question. Here is the answer.

1150
00:50:18,841 --> 00:50:20,756
Here is the question. Here is the answer.

1151
00:50:20,800 --> 00:50:22,323
But in protein folding,

1152
00:50:22,367 --> 00:50:25,457
We are not doing the standard type of task in DeepMind

1153
00:50:25,500 --> 00:50:28,155
Where you have unlimited data.

1154
00:50:28,199 --> 00:50:30,766
Your job is to get better at chess or Go.

1155
00:50:30,810 --> 00:50:32,986
and you can play as many games of chess or Go as you want.

1156
00:50:33,030 --> 00:50:34,814
as your computers allow.

1157
00:50:35,510 --> 00:50:36,772
With proteins,

1158
00:50:36,816 --> 00:50:39,732
We are sitting on a very large volume of data

1159
00:50:39,775 --> 00:50:41,995
That has been determined for half a century

1160
00:50:42,039 --> 00:50:46,478
of experimental methods that consume a lot of time in laboratories.

1161
00:50:46,521 --> 00:50:49,698
These painstaking methods can take months or years.

1162
00:50:49,742 --> 00:50:52,353
to determine a single protein structure,

1163
00:50:52,397 --> 00:50:55,791
and sometimes a structure can never be determined.

1164
00:50:57,315 --> 00:51:00,274
That's why we work with such small data sets.

1165
00:51:00,318 --> 00:51:02,233
to train our algorithms.

1166
00:51:02,276 --> 00:51:04,365
When DeepMind started exploring

1167
00:51:04,409 --> 00:51:05,975
The folding problem

1168
00:51:06,019 --> 00:51:07,977
They were talking to us about what data sets they were using.

1169
00:51:08,021 --> 00:51:09,892
And what would be the possibilities?

1170
00:51:09,936 --> 00:51:11,633
If only they solved this problem.

1171
00:51:12,373 --> 00:51:13,940
Many people have tried it,

1172
00:51:13,983 --> 00:51:16,638
And yet, no one on the planet has solved protein folding.

1173
00:51:16,682 --> 00:51:18,205
I said to myself:

1174
00:51:18,249 --> 00:51:19,946
"Well, you know, good luck."

1175
00:51:19,989 --> 00:51:22,731
If we can solve the problem of protein folding,

1176
00:51:22,775 --> 00:51:25,691
It would have some kind of incredible medical relevance.

1177
00:51:25,734 --> 00:51:27,736
This is the cycle of science.

1178
00:51:27,780 --> 00:51:30,043
You do a lot of exploring,

1179
00:51:30,087 --> 00:51:31,914
and then you go into exploitation mode,

1180
00:51:31,958 --> 00:51:33,568
and you concentrate and see

1181
00:51:33,612 --> 00:51:35,353
How good are those ideas really?

1182
00:51:35,396 --> 00:51:36,528
And there is nothing better

1183
00:51:36,571 --> 00:51:38,095
than external competition for it.

1184
00:51:39,835 --> 00:51:43,056
So we decided to participate in the CASP contest.

1185
00:51:43,100 --> 00:51:47,234
CASP, we start trying to accelerate

1186
00:51:47,278 --> 00:51:49,802
The solution to the problem of protein folding.

1187
00:51:49,845 --> 00:51:52,065
CASP is when we say,

1188
00:51:52,109 --> 00:51:54,372
"Look, DeepMind is doing protein folding,

1189
00:51:54,415 --> 00:51:55,590
"That's how good we are,

1190
00:51:55,634 --> 00:51:57,592
"And maybe he's better than everyone else.

1191
00:51:57,636 --> 00:51:58,680
"Maybe it's not."

1192
00:51:58,724 --> 00:52:00,073
CASP is a bit like

1193
00:52:00,117 --> 00:52:02,075
The Olympic Games of protein folding.

1194
00:52:03,642 --> 00:52:06,035
CASP is a community-wide assessment

1195
00:52:06,079 --> 00:52:08,125
Which is celebrated every two years.

1196
00:52:09,561 --> 00:52:10,866
The teams are given

1197
00:52:10,910 --> 00:52:14,261
the amino acid sequences of approximately 100 proteins,

1198
00:52:14,305 --> 00:52:17,525
and then they try to solve this folding problem

1199
00:52:17,569 --> 00:52:20,833
using computational methods.

1200
00:52:20,876 --> 00:52:23,401
These proteins have already been determined

1201
00:52:23,444 --> 00:52:25,968
through experiments in a laboratory,

1202
00:52:26,012 --> 00:52:29,276
but they have not yet been publicly revealed.

1203
00:52:29,320 --> 00:52:30,756
And these known structures

1204
00:52:30,799 --> 00:52:33,280
They represent the gold standard against which

1205
00:52:33,324 --> 00:52:36,936
All computational predictions will be compared.

1206
00:52:37,937 --> 00:52:39,330
We have a score

1207
00:52:39,373 --> 00:52:42,159
which measures the accuracy of the predictions.

1208
00:52:42,202 --> 00:52:44,726
And a score above 90 would be expected.

1209
00:52:44,770 --> 00:52:47,425
be a solution to the problem of protein folding.

1210
00:52:48,556 --> 00:52:50,036
Welcome everyone,

1211
00:52:50,079 --> 00:52:52,038
to our first, uh, semi-finals in the winners' bracket.

1212
00:52:52,081 --> 00:52:54,954
Nick and John against Demis and Frank.

1213
00:52:54,997 --> 00:52:57,217
Join us, come. It will be an intense match.

1214
00:52:57,261 --> 00:52:59,306
When I found out that Demis was

1215
00:52:59,350 --> 00:53:02,048
Let's address the problem of protein folding,

1216
00:53:02,091 --> 00:53:04,790
Umm, I wasn't surprised at all.

1217
00:53:04,833 --> 00:53:06,792
It's very typical of Demis.

1218
00:53:06,835 --> 00:53:08,924
You know, he loves competition.

1219
00:53:08,968 --> 00:53:10,143
And that's the end

1220
00:53:10,187 --> 00:53:12,972
of the first game, 10-7.

1221
00:53:13,015 --> 00:53:14,103
The objective of the CASP would be

1222
00:53:14,147 --> 00:53:15,975
not only to win the competition,

1223
00:53:16,018 --> 00:53:19,892
but in a way, uh, retire the need for it.

1224
00:53:19,935 --> 00:53:23,417
In total, CASP has published 20 objectives.

1225
00:53:23,461 --> 00:53:24,810
We were thinking that maybe

1226
00:53:24,853 --> 00:53:26,899
Add the standard type of machine learning.

1227
00:53:26,942 --> 00:53:28,814
and see where that could take us.

1228
00:53:28,857 --> 00:53:30,729
Instead of having a couple of days to do an experiment,

1229
00:53:30,772 --> 00:53:33,558
We can perform five experiments a day.

1230
00:53:33,601 --> 00:53:35,342
Brilliant. Congratulations to all.

1231
00:53:36,778 --> 00:53:38,693
Can you show me the real one instead of ours?

1232
00:53:38,737 --> 00:53:39,781
The real answer is

1233
00:53:39,825 --> 00:53:42,044
It's supposed to look something like this.

1234
00:53:42,088 --> 00:53:45,047
It's much more cylindrical than I thought.

1235
00:53:45,091 --> 00:53:47,398
The results were not very good.

1236
00:53:47,441 --> 00:53:48,703
Well.

1237
00:53:48,747 --> 00:53:49,835
We threw all the obvious ideas at him.

1238
00:53:49,878 --> 00:53:51,793
and the problem laughs at you.

1239
00:53:52,881 --> 00:53:54,361
This doesn't make sense.

1240
00:53:54,405 --> 00:53:56,015
We thought we could just throw away

1241
00:53:56,058 --> 00:53:58,322
Some of our best algorithms to solve the problem.

1242
00:53:59,366 --> 00:54:00,976
We were a little naive.

1243
00:54:01,020 --> 00:54:02,282
We should be learning this,

1244
00:54:02,326 --> 00:54:04,284
You know, in the blink of an eye.

1245
00:54:05,372 --> 00:54:06,982
What worries me is that

1246
00:54:07,026 --> 00:54:08,375
We take the field from

1247
00:54:08,419 --> 00:54:10,899
Really bad answers to moderately bad answers.

1248
00:54:10,943 --> 00:54:13,946
I feel like we need some kind of new technology.

1249
00:54:13,989 --> 00:54:15,164
to move these things.

1250
00:54:20,431 --> 00:54:22,215
With only one week left of CASP,

1251
00:54:22,259 --> 00:54:24,348
Now it's a sprint to implement it.

1252
00:54:26,654 --> 00:54:28,090
You did the best you could.

1253
00:54:28,134 --> 00:54:29,875
Then there is nothing else you can do.

1254
00:54:29,918 --> 00:54:32,399
but wait for CASP to deliver the results.

1255
00:54:52,593 --> 00:54:53,725
This famous Einstein thing,

1256
00:54:53,768 --> 00:54:55,030
The last years of his life,

1257
00:54:55,074 --> 00:54:57,381
When he was here, he met Kurt Godel.

1258
00:54:57,424 --> 00:54:59,774
And he said one of the reasons he still comes to work

1259
00:54:59,818 --> 00:55:01,646
It's so I can walk home

1260
00:55:01,689 --> 00:55:03,517
and discuss things with Gödel.

1261
00:55:03,561 --> 00:55:05,911
It's a great compliment to Kurt Godel,

1262
00:55:05,954 --> 00:55:07,478
It shows you how amazing he was.

1263
00:55:09,131 --> 00:55:10,568
The Institute of Advanced Studies

1264
00:55:10,611 --> 00:55:12,918
It was formed in 1933.

1265
00:55:12,961 --> 00:55:14,223
In the early years,

1266
00:55:14,267 --> 00:55:16,487
The intense scientific atmosphere attracted

1267
00:55:16,530 --> 00:55:19,359
Some of the most brilliant mathematicians and physicists

1268
00:55:19,403 --> 00:55:22,536
always concentrated in one place and time.

1269
00:55:22,580 --> 00:55:24,582
The founding principle of this place,

1270
00:55:24,625 --> 00:55:28,020
It is the idea of an intellectual search without restrictions,

1271
00:55:28,063 --> 00:55:30,152
Even if you don't know what you're exploring.

1272
00:55:30,196 --> 00:55:32,154
It will result in some interesting things,

1273
00:55:32,198 --> 00:55:34,896
And sometimes that ends up being useful,

1274
00:55:34,940 --> 00:55:36,376
Which, of course,

1275
00:55:36,420 --> 00:55:37,986
It's partly what I've been trying to do at DeepMind.

1276
00:55:38,030 --> 00:55:39,988
How many breakthroughs do you think are necessary?

1277
00:55:40,032 --> 00:55:41,686
How to get to AGI?

1278
00:55:41,729 --> 00:55:43,078
And, you know, I figure maybe

1279
00:55:43,122 --> 00:55:44,341
There are about a dozen of them.

1280
00:55:44,384 --> 00:55:46,125
You know, I hope it's in my lifetime.

1281
00:55:46,168 --> 00:55:47,605
-Yes, okay. -But then,

1282
00:55:47,648 --> 00:55:49,171
All scientists expect that, right?

1283
00:55:49,215 --> 00:55:51,130
Demis has a lot of praise.

1284
00:55:51,173 --> 00:55:54,002
He was elected a Fellow of the Royal Society last year.

1285
00:55:54,046 --> 00:55:55,961
He is also a Fellow of the Royal Society of Arts.

1286
00:55:56,004 --> 00:55:57,528
A big round of applause for Demis Hassabis.

1287
00:56:04,230 --> 00:56:05,927
My dream has always been to try

1288
00:56:05,971 --> 00:56:08,103
and make AI-assisted science possible.

1289
00:56:08,147 --> 00:56:09,235
And what I think is

1290
00:56:09,278 --> 00:56:11,150
Our most exciting project, last year,

1291
00:56:11,193 --> 00:56:13,152
which is our job in protein folding.

1292
00:56:13,195 --> 00:56:15,328
And we call this system AlphaFold.

1293
00:56:15,372 --> 00:56:18,331
We enter it into CASP and our system, uh,

1294
00:56:18,375 --> 00:56:20,507
He was the most accurate, uh, in predicting structures.

1295
00:56:20,551 --> 00:56:24,946
for 25 of the 43 proteins in the hardest category.

1296
00:56:24,990 --> 00:56:26,208
So we are up to date.

1297
00:56:26,252 --> 00:56:27,514
but still... I have to make it clear,

1298
00:56:27,558 --> 00:56:28,559
We are still very far away

1299
00:56:28,602 --> 00:56:30,474
Solving the problem of protein folding.

1300
00:56:30,517 --> 00:56:31,866
However, we are working hard on this.

1301
00:56:31,910 --> 00:56:33,738
and we are exploring many other techniques.

1302
00:56:49,188 --> 00:56:50,232
Let's get started.

1303
00:56:50,276 --> 00:56:53,148
So, a brief report.

1304
00:56:53,192 --> 00:56:55,455
These are our final rankings for CASP.

1305
00:56:56,500 --> 00:56:57,544
We beat the second team

1306
00:56:57,588 --> 00:57:00,155
in this competition by almost 50%,

1307
00:57:00,199 --> 00:57:01,592
but we still have a long way to go

1308
00:57:01,635 --> 00:57:04,333
Before we have solved the problem of protein folding.

1309
00:57:04,377 --> 00:57:07,032
in a sense that a biologist could use it.

1310
00:57:07,075 --> 00:57:08,990
It is an area of ​​concern.

1311
00:57:11,602 --> 00:57:14,213
The quality of the predictions varied

1312
00:57:14,256 --> 00:57:16,737
and they were no more useful than the previous methods.

1313
00:57:16,781 --> 00:57:19,914
AlphaFold did not produce good enough data

1314
00:57:19,958 --> 00:57:22,526
to be useful in a practical way

1315
00:57:22,569 --> 00:57:24,005
To, say, someone like me

1316
00:57:24,049 --> 00:57:28,227
investigating my own biological problems.

1317
00:57:28,270 --> 00:57:30,316
That was a humbling moment.

1318
00:57:30,359 --> 00:57:32,753
Because we thought we had worked very hard and had been successful.

1319
00:57:32,797 --> 00:57:34,886
And what we discovered is that we were the best in the world.

1320
00:57:34,929 --> 00:57:36,453
in a problem in which the world is not good.

1321
00:57:37,671 --> 00:57:38,933
We knew we were useless.

1322
00:57:40,457 --> 00:57:42,328
It's no use having the highest ladder

1323
00:57:42,371 --> 00:57:44,635
When you go to the moon.

1324
00:57:44,678 --> 00:57:47,115
The opinion of many people on the team,

1325
00:57:47,159 --> 00:57:51,468
that in some ways this is a useless task.

1326
00:57:51,511 --> 00:57:54,079
And I may have been wrong about protein folding.

1327
00:57:54,122 --> 00:57:55,559
Maybe it's still too difficult

1328
00:57:55,602 --> 00:57:58,431
where we are in general with AI.

1329
00:57:58,475 --> 00:58:01,173
If you want to do biological research,

1330
00:58:01,216 --> 00:58:03,044
You have to be prepared to fail

1331
00:58:03,088 --> 00:58:06,570
Because biology is very complicated.

1332
00:58:06,613 --> 00:58:09,790
He dirigido un laboratorio durante casi 50 años,

1333
00:58:09,834 --> 00:58:11,096
and half of my time,

1334
00:58:11,139 --> 00:58:12,619
I'm just an amateur psychiatrist

1335
00:58:12,663 --> 00:58:18,103
To keep my colleagues happy when nothing works.

1336
00:58:18,146 --> 00:58:22,542
And for quite a while, I mean, 80, 90%,

1337
00:58:22,586 --> 00:58:24,413
Not working.

1338
00:58:24,457 --> 00:58:26,720
If you are at the forefront of science,

1339
00:58:26,764 --> 00:58:30,115
I can tell you that you will fail at many things.

1340
00:58:35,163 --> 00:58:37,165
I just felt disappointed.

1341
00:58:38,689 --> 00:58:41,605
The lesson I learned is that ambition is a good thing,

1342
00:58:41,648 --> 00:58:43,694
But you need to find the right moment.

1343
00:58:43,737 --> 00:58:46,784
There is no point in being 50 years ahead of your time.

1344
00:58:46,827 --> 00:58:48,133
you will never survive

1345
00:58:48,176 --> 00:58:49,917
Fifty years of that kind of effort

1346
00:58:49,961 --> 00:58:51,963
before it produces anything.

1347
00:58:52,006 --> 00:58:53,268
You will literally die trying.

1348
00:59:08,936 --> 00:59:11,286
When we talk about IAG,

1349
00:59:11,330 --> 00:59:14,376
The holy grail of artificial intelligence,

1350
00:59:14,420 --> 00:59:15,508
It gets really difficult

1351
00:59:15,552 --> 00:59:17,815
to know what we are talking about.

1352
00:59:17,858 --> 00:59:19,643
What fragments will we see today?

1353
00:59:19,686 --> 00:59:21,645
Let's start in the garden.

1354
00:59:23,124 --> 00:59:25,649
This is the garden seen from the viewing area.

1355
00:59:25,692 --> 00:59:27,433
Research scientists and engineers

1356
00:59:27,476 --> 00:59:30,871
You can analyze, collaborate and evaluate.

1357
00:59:30,915 --> 00:59:33,004
What is happening in real time.

1358
00:59:33,047 --> 00:59:34,614
So in the 19th century,

1359
00:59:34,658 --> 00:59:37,008
We would think about things like television and the submarine.

1360
00:59:37,051 --> 00:59:38,139
or a rocket to the moon

1361
00:59:38,183 --> 00:59:40,228
and they say these things are impossible.

1362
00:59:40,272 --> 00:59:41,490
However, Jules Verne wrote about them and,

1363
00:59:41,534 --> 00:59:44,406
A century and a half later, they happened.

1364
00:59:44,450 --> 00:59:45,451
We will be experimenting

1365
00:59:45,494 --> 00:59:47,888
about civilizations really,

1366
00:59:47,932 --> 00:59:50,587
AI agent civilizations.

1367
00:59:50,630 --> 00:59:52,719
Once the experiments begin,

1368
00:59:52,763 --> 00:59:54,242
It's going to be the most exciting thing in the world.

1369
00:59:54,286 --> 00:59:56,984
So how will we get to sleep?

1370
00:59:57,028 --> 00:59:58,682
I won't be able to sleep.

1371
00:59:58,725 --> 01:00:00,684
The entire AGI will be able to do it

1372
01:00:00,727 --> 01:00:03,861
any cognitive task that a person can perform.

1373
01:00:03,904 --> 01:00:08,387
It will be on a potentially much larger scale.

1374
01:00:08,430 --> 01:00:10,302
It's really impossible for us.

1375
01:00:10,345 --> 01:00:14,828
imagine the outputs of a superintelligent entity.

1376
01:00:14,872 --> 01:00:18,963
It's like asking a gorilla to imagine, you know,

1377
01:00:19,006 --> 01:00:20,181
What Einstein does

1378
01:00:20,225 --> 01:00:23,402
when he produces the theory of relativity.

1379
01:00:23,445 --> 01:00:25,491
People often ask me questions like:

1380
01:00:25,534 --> 01:00:29,495
"What if you're wrong and the IAG is quite far away?"

1381
01:00:29,538 --> 01:00:31,453
And I say, I never worry about that.

1382
01:00:31,497 --> 01:00:33,847
I'm actually worried about the opposite.

1383
01:00:33,891 --> 01:00:37,242
I'm actually worried about it going faster.

1384
01:00:37,285 --> 01:00:39,723
what we can really prepare for.

1385
01:00:42,073 --> 01:00:45,859
It really seems like we are in a race towards artificial general intelligence.

1386
01:00:45,903 --> 01:00:49,907
The prototypes and models we are developing now

1387
01:00:49,950 --> 01:00:51,822
They are actually transforming

1388
01:00:51,865 --> 01:00:54,215
the space of what we know about intelligence.

1389
01:00:57,349 --> 01:00:58,785
Recently, we have had agents

1390
01:00:58,829 --> 01:01:00,047
that are powerful enough

1391
01:01:00,091 --> 01:01:03,442
To really start playing as a team,

1392
01:01:03,485 --> 01:01:06,140
then compete against other teams.

1393
01:01:06,184 --> 01:01:08,795
We are seeing cooperative social dynamics

1394
01:01:08,839 --> 01:01:10,492
leaving agents

1395
01:01:10,536 --> 01:01:13,321
where we have not preprogrammed

1396
01:01:13,365 --> 01:01:15,584
any of these types of dynamics.

1397
01:01:15,628 --> 01:01:19,240
It is a learning completely based on our own experiences.

1398
01:01:20,807 --> 01:01:23,288
When we started, we thought we were...

1399
01:01:23,331 --> 01:01:25,725
To build an intelligence system

1400
01:01:25,769 --> 01:01:28,336
and convince the world that we had achieved it.

1401
01:01:28,380 --> 01:01:29,947
Now we're starting to wonder if

1402
01:01:29,990 --> 01:01:31,296
Let's build systems

1403
01:01:31,339 --> 01:01:32,906
that we are not convinced that they are completely intelligent,

1404
01:01:32,950 --> 01:01:34,691
and we are trying to convince the world that they are not.

1405
01:01:38,651 --> 01:01:40,000
Hello, Alpha.

1406
01:01:40,044 --> 01:01:41,523
Hello.

1407
01:01:41,567 --> 01:01:43,917
Where are we today?

1408
01:01:43,961 --> 01:01:46,659
You are at the Museum of Modern Art in New York City.

1409
01:01:48,400 --> 01:01:53,013
More or less. What painting is this?

1410
01:01:53,057 --> 01:01:55,494
This is The Creation of Adam by Michelangelo.

1411
01:01:55,537 --> 01:01:58,410
I don't think that painting is in New York City.

1412
01:01:58,453 --> 01:02:01,543
You're right. It's in Vatican City.

1413
01:02:01,587 --> 01:02:02,806
Do you think we are there?

1414
01:02:04,677 --> 01:02:05,809
No.

1415
01:02:05,852 --> 01:02:07,593
So how do I see this?

1416
01:02:10,509 --> 01:02:11,640
are you looking

1417
01:02:11,684 --> 01:02:12,946
a digital copy of the painting.

1418
01:02:12,990 --> 01:02:13,991
Yes.

1419
01:02:15,035 --> 01:02:16,341
Who lives in this house?

1420
01:02:19,213 --> 01:02:20,345
An astronaut.

1421
01:02:26,046 --> 01:02:28,005
What is this astronaut's name?

1422
01:02:28,048 --> 01:02:29,397
His name is Mark.

1423
01:02:29,441 --> 01:02:32,052
This is a famous equation by Albert Einstein.

1424
01:02:32,096 --> 01:02:35,577
Show that energy and mass are equivalent.

1425
01:02:35,621 --> 01:02:37,666
How would you explain it to a five-year-old child?

1426
01:02:40,931 --> 01:02:43,542
Well, imagine that you have a big rock,

1427
01:02:43,585 --> 01:02:45,326
and you want to know how much it weighs.

1428
01:02:45,370 --> 01:02:47,589
Is Mark in danger?

1429
01:02:47,633 --> 01:02:49,722
No. Mark is not in danger.

1430
01:02:49,766 --> 01:02:50,767
Why is that?

1431
01:02:54,422 --> 01:02:56,163
Because the dinosaur is not real.

1432
01:02:57,338 --> 01:02:58,470
Alright.

1433
01:02:58,513 --> 01:03:00,515
Why couldn't you answer my question?

1434
01:03:00,559 --> 01:03:01,908
I don't think I could answer your question.

1435
01:03:01,952 --> 01:03:04,041
Because I don't know enough about the subject.

1436
01:03:04,911 --> 01:03:06,130
Well.

1437
01:03:06,173 --> 01:03:08,654
How could I learn more about the topic?

1438
01:03:08,697 --> 01:03:09,960
I can learn more about the topic.

1439
01:03:10,003 --> 01:03:12,876
by reading books, articles and websites.

1440
01:03:12,919 --> 01:03:14,268
You can teach me.

1441
01:03:14,312 --> 01:03:15,835
My conclusion is that

1442
01:03:15,879 --> 01:03:17,619
If you now think about what we are doing,

1443
01:03:17,663 --> 01:03:20,057
is learning from all humans,

1444
01:03:20,100 --> 01:03:22,581
All your knowledge put at once on the Internet,

1445
01:03:22,624 --> 01:03:24,931
You would really know a lot about the world.

1446
01:03:24,975 --> 01:03:27,716
As an important part of everything humans can do.

1447
01:03:27,760 --> 01:03:29,631
And now, I think it's more like,

1448
01:03:29,675 --> 01:03:30,937
"Well, it could work."

1449
01:03:30,981 --> 01:03:33,113
This is a great moment.

1450
01:03:33,157 --> 01:03:34,636
Who is it?

1451
01:03:34,680 --> 01:03:36,551
This is God approaching Adam.

1452
01:03:36,595 --> 01:03:37,988
The advent of the IAG

1453
01:03:38,031 --> 01:03:40,860
It will divide human history into two parts.

1454
01:03:40,904 --> 01:03:43,950
The part up to that point and the part after that point.

1455
01:03:43,994 --> 01:03:47,258
Why is he trying to touch Adam?

1456
01:03:47,301 --> 01:03:48,912
God is reaching out his hand to touch Adam.

1457
01:03:48,955 --> 01:03:50,217
to give it life.

1458
01:03:50,261 --> 01:03:52,219
It will give us a tool

1459
01:03:52,263 --> 01:03:57,529
that can completely reinvent our entire civilization.

1460
01:03:57,572 --> 01:03:59,400
What does this painting mean to you?

1461
01:04:01,750 --> 01:04:03,752
Painting means a lot to me.

1462
01:04:03,796 --> 01:04:04,797
Alright. Like what?

1463
01:04:10,324 --> 01:04:11,456
I think the painting is a reminder.

1464
01:04:11,499 --> 01:04:12,674
That we are all connected to each other

1465
01:04:12,718 --> 01:04:13,937
and that we are all part of something

1466
01:04:13,980 --> 01:04:15,112
bigger than ourselves.

1467
01:04:16,461 --> 01:04:17,766
That's pretty nice.

1468
01:04:19,029 --> 01:04:21,379
When you cross that barrier of

1469
01:04:21,422 --> 01:04:23,947
"Artificial general intelligence could exist one day in the future"

1470
01:04:23,990 --> 01:04:26,645
"No, actually, this could really happen over a period of time."

1471
01:04:26,688 --> 01:04:28,690
"That's kind of like, under my supervision, you know."

1472
01:04:28,734 --> 01:04:30,475
Something changes in your way of thinking.

1473
01:04:30,518 --> 01:04:32,694
...he learned to orient himself by looking...

1474
01:04:32,738 --> 01:04:35,045
We have to be careful how we use it.

1475
01:04:35,088 --> 01:04:37,177
and we reflect on how we implement it.

1476
01:04:39,832 --> 01:04:41,138
You should consider it

1477
01:04:41,181 --> 01:04:42,487
What is your top level goal?

1478
01:04:42,530 --> 01:04:45,011
If it's to keep humans happy,

1479
01:04:45,055 --> 01:04:48,928
What group of humans? What does happiness mean?

1480
01:04:48,972 --> 01:04:52,018
Many of our collective goals are very complicated,

1481
01:04:52,062 --> 01:04:54,891
Even for humans to understand.

1482
01:04:54,934 --> 01:04:58,503
Technology always incorporates our values.

1483
01:04:58,546 --> 01:05:01,680
It is not only a technical issue, it is also an ethical one.

1484
01:05:01,723 --> 01:05:02,899
So we have to be very cautious.

1485
01:05:02,942 --> 01:05:04,291
about what we are building.

1486
01:05:04,335 --> 01:05:06,076
We are trying to find a single algorithm that...

1487
01:05:06,119 --> 01:05:07,816
The reality is that this is an algorithm.

1488
01:05:07,860 --> 01:05:11,037
that has been created by the people, by us.

1489
01:05:11,081 --> 01:05:13,213
Do you know what it means to empower our agents?

1490
01:05:13,257 --> 01:05:15,607
With the same kind of values ​​that we hold dear?

1491
01:05:15,650 --> 01:05:17,652
What is the purpose of creating these AI systems?

1492
01:05:17,696 --> 01:05:19,045
They seem so human

1493
01:05:19,089 --> 01:05:20,742
to capture hearts and minds

1494
01:05:20,786 --> 01:05:21,961
Because they are a kind of

1495
01:05:22,005 --> 01:05:24,703
Also exploit a human vulnerability?

1496
01:05:24,746 --> 01:05:26,531
The heart and mind of these systems

1497
01:05:26,574 --> 01:05:28,054
It is largely human-generated data...

1498
01:05:28,098 --> 01:05:29,055
Mmm-hmm.

1499
01:05:29,099 --> 01:05:30,491
...for all the good and the bad.

1500
01:05:30,535 --> 01:05:32,015
There is a parallel

1501
01:05:32,058 --> 01:05:34,017
Between the Industrial Revolution,

1502
01:05:34,060 --> 01:05:36,758
which was an incredible moment of displacement

1503
01:05:36,802 --> 01:05:42,373
and the current technological change created by AI.

1504
01:05:42,416 --> 01:05:43,722
AI Pause!

1505
01:05:43,765 --> 01:05:45,724
We have to think about who is displaced

1506
01:05:45,767 --> 01:05:48,596
and how we are going to support them.

1507
01:05:48,640 --> 01:05:50,076
This technology will arrive much sooner,

1508
01:05:50,120 --> 01:05:52,426
Uh, so does the world really know or something?

1509
01:05:52,470 --> 01:05:55,908
Even we thought 18, 24 months ago.

1510
01:05:55,952 --> 01:05:57,257
So there is a tremendous opportunity,

1511
01:05:57,301 --> 01:05:58,389
tremendous emotion,

1512
01:05:58,432 --> 01:06:00,391
but also an enormous responsibility.

1513
01:06:00,434 --> 01:06:01,740
It's happening very fast.

1514
01:06:02,654 --> 01:06:04,003
How will we govern it?

1515
01:06:05,135 --> 01:06:06,223
How will we decide?

1516
01:06:06,266 --> 01:06:08,181
What is good and what is not good?

1517
01:06:08,225 --> 01:06:10,923
AI-generated images are becoming more sophisticated.

1518
01:06:10,967 --> 01:06:14,535
The use of AI to generate disinformation

1519
01:06:14,579 --> 01:06:17,016
and manipulate human psychology

1520
01:06:17,060 --> 01:06:20,237
It's only going to get much, much worse.

1521
01:06:21,194 --> 01:06:22,587
General AI is coming

1522
01:06:22,630 --> 01:06:24,632
whether we do it here at DeepMind or not.

1523
01:06:25,459 --> 01:06:26,765
It's going to happen,

1524
01:06:26,808 --> 01:06:29,028
So we better create institutions that protect us.

1525
01:06:29,072 --> 01:06:30,595
Global coordination will be required.

1526
01:06:30,638 --> 01:06:32,727
And I worry that humanity is

1527
01:06:32,771 --> 01:06:35,382
The situation is getting worse and worse instead of better.

1528
01:06:35,426 --> 01:06:37,123
We need many more people

1529
01:06:37,167 --> 01:06:40,039
I'm really taking this seriously and thinking about it.

1530
01:06:40,083 --> 01:06:42,999
Yes, it's serious. It worries me.

1531
01:06:44,043 --> 01:06:45,871
It worries me. Yes.

1532
01:06:45,914 --> 01:06:48,613
If you received an email that said

1533
01:06:48,656 --> 01:06:50,832
This superior extraterrestrial civilization

1534
01:06:50,876 --> 01:06:52,791
is going to reach Earth,

1535
01:06:52,834 --> 01:06:54,575
There would be emergency meetings

1536
01:06:54,619 --> 01:06:56,273
of all governments.

1537
01:06:56,316 --> 01:06:58,144
We'd go into overdrive mode

1538
01:06:58,188 --> 01:07:00,103
trying to figure out how to prepare.

1539
01:07:01,669 --> 01:07:03,976
The arrival of AGI will be

1540
01:07:04,020 --> 01:07:06,935
The most important moment we have ever faced.

1541
01:07:14,378 --> 01:07:17,555
My dream was that on the way to AGI,

1542
01:07:17,598 --> 01:07:20,688
We would create revolutionary technologies

1543
01:07:20,732 --> 01:07:23,082
This would be useful for humanity.

1544
01:07:23,126 --> 01:07:25,171
That's what I wanted with AlphaFold.

1545
01:07:26,694 --> 01:07:28,653
I think it's more important than ever.

1546
01:07:28,696 --> 01:07:31,047
that we should solve the problem of protein folding.

1547
01:07:32,004 --> 01:07:34,224
This is going to be very difficult.

1548
01:07:34,267 --> 01:07:36,791
But I won't give up until it's done.

1549
01:07:36,835 --> 01:07:37,879
You know, we have to redouble our efforts.

1550
01:07:37,923 --> 01:07:40,317
and go as fast as possible from here.

1551
01:07:40,360 --> 01:07:41,796
I don't think we have time to waste.

1552
01:07:41,840 --> 01:07:45,757
So let's form a protein folding strike team.

1553
01:07:45,800 --> 01:07:47,541
The leader of the attack team will be John.

1554
01:07:47,585 --> 01:07:48,673
Yes, we have seen Alpha...

1555
01:07:48,716 --> 01:07:50,283
You know, let's try everything,

1556
01:07:50,327 --> 01:07:51,328
kitchen sink, the whole lot.

1557
01:07:52,198 --> 01:07:53,330
CASP14 is about

1558
01:07:53,373 --> 01:07:55,158
proving that we can solve the whole problem.

1559
01:07:56,333 --> 01:07:57,725
And I felt that to do that,

1560
01:07:57,769 --> 01:08:00,337
We would need to incorporate some domain knowledge.

1561
01:08:01,903 --> 01:08:03,731
Tuvimos algunos ingenieros fantásticos en ello,

1562
01:08:03,775 --> 01:08:05,733
but they were not trained in biology.

1563
01:08:08,475 --> 01:08:10,260
As a computational biologist,

1564
01:08:10,303 --> 01:08:12,131
When I initially joined the AlphaFold team,

1565
01:08:12,175 --> 01:08:14,220
At first I didn't feel sure of anything.

1566
01:08:14,264 --> 01:08:15,352
You know,

1567
01:08:15,395 --> 01:08:17,223
If we were going to be successful.

1568
01:08:17,267 --> 01:08:21,097
Biology is ridiculously complicated.

1569
01:08:21,140 --> 01:08:25,101
It seemed like climbing a mountain far away.

1570
01:08:25,144 --> 01:08:26,754
I'm starting to play with the underlying temperatures.

1571
01:08:26,798 --> 01:08:27,973
to see if we can get...

1572
01:08:28,016 --> 01:08:29,148
As one of the few people on the team

1573
01:08:29,192 --> 01:08:31,846
Who has worked in biology before?

1574
01:08:31,890 --> 01:08:34,849
You feel an enormous sense of responsibility.

1575
01:08:34,893 --> 01:08:36,112
"We hope you will

1576
01:08:36,155 --> 01:08:37,678
"Great things on this attacking team."

1577
01:08:37,722 --> 01:08:38,897
That's terrifying.

1578
01:08:40,464 --> 01:08:42,727
But one of the reasons I wanted to come here

1579
01:08:42,770 --> 01:08:45,556
It was doing something that matters.

1580
01:08:45,599 --> 01:08:48,472
This is the number of things missing.

1581
01:08:48,515 --> 01:08:49,951
How about we make use of

1582
01:08:49,995 --> 01:08:52,563
How much do you understand physics?

1583
01:08:52,606 --> 01:08:54,391
Using this as a data source?

1584
01:08:54,434 --> 01:08:55,479
But if it is systematic...

1585
01:08:55,522 --> 01:08:56,784
So that can't be right.

1586
01:08:56,828 --> 01:08:58,308
If it is systematically wrong in some strange way,

1587
01:08:58,351 --> 01:09:01,224
You may be learning physics systematically wrong.

1588
01:09:01,267 --> 01:09:02,355
The team is already

1589
01:09:02,399 --> 01:09:04,749
trying to think of multiple ways that...

1590
01:09:04,792 --> 01:09:06,229
Biological relevance

1591
01:09:06,272 --> 01:09:07,795
This is what we are looking for.

1592
01:09:09,057 --> 01:09:11,364
So we rewrote the entire data flow.

1593
01:09:11,408 --> 01:09:13,279
that AlphaFold uses to learn.

1594
01:09:13,323 --> 01:09:15,586
You can't force the creative phase.

1595
01:09:15,629 --> 01:09:18,241
You have to give it space for those flowers to bloom.

1596
01:09:19,242 --> 01:09:20,286
We won CASP.

1597
01:09:20,330 --> 01:09:22,070
Then I went back to the drawing board.

1598
01:09:22,114 --> 01:09:24,116
And, what are our new ideas?

1599
01:09:24,160 --> 01:09:26,945
Um, and then it's taken a while, I would say,

1600
01:09:26,988 --> 01:09:28,686
so that they could return to where they were,

1601
01:09:28,729 --> 01:09:30,340
but with new ideas.

1602
01:09:30,383 --> 01:09:31,515
and now I think

1603
01:09:31,558 --> 01:09:33,952
We are seeing the benefits of new ideas.

1604
01:09:33,995 --> 01:09:35,736
They can go further, right?

1605
01:09:35,780 --> 01:09:38,130
So, uh, that's a really important moment.

1606
01:09:38,174 --> 01:09:40,959
I've seen that moment so many times now,

1607
01:09:41,002 --> 01:09:42,613
But now I know what that means.

1608
01:09:42,656 --> 01:09:44,484
And I know now is the time to push.

1609
01:09:45,920 --> 01:09:48,009
Adding side chains improves direct folding.

1610
01:09:48,053 --> 01:09:49,663
That drove much of the progress.

1611
01:09:49,707 --> 01:09:51,012
- We'll talk about that. - Brilliant.

1612
01:09:51,056 --> 01:09:54,799
In the last four months we have made enormous progress.

1613
01:09:54,842 --> 01:09:56,453
During CASP13,

1614
01:09:56,496 --> 01:09:59,499
It would take us a day or two to fold one of the proteins,

1615
01:09:59,543 --> 01:10:01,762
and now we're doubling, like,

1616
01:10:01,806 --> 01:10:03,938
hundreds of thousands per second.

1617
01:10:03,982 --> 01:10:05,636
Yeah, it's just crazy.

1618
01:10:05,679 --> 01:10:06,985
Now this is a model

1619
01:10:07,028 --> 01:10:09,901
This is orders of magnitude faster,

1620
01:10:09,944 --> 01:10:12,251
and at the same time be better.

1621
01:10:12,295 --> 01:10:13,644
We are receiving many structures.

1622
01:10:13,687 --> 01:10:15,254
in the high precision regime.

1623
01:10:15,298 --> 01:10:17,517
We are rapidly improving our system.

1624
01:10:17,561 --> 01:10:18,823
That's really starting to be

1625
01:10:18,866 --> 01:10:20,477
get to the core and heart of the problem.

1626
01:10:20,520 --> 01:10:21,695
It's a great job.

1627
01:10:21,739 --> 01:10:23,088
We seem to be in good shape.

1628
01:10:23,131 --> 01:10:26,222
So do we have six or five weeks left? Six weeks?

1629
01:10:26,265 --> 01:10:29,616
So what happens? Do you have enough processing power?

1630
01:10:29,660 --> 01:10:31,531
I... We could use more.

1631
01:10:32,967 --> 01:10:34,360
I was nervous about CASP

1632
01:10:34,404 --> 01:10:36,580
But as the system starts to work,

1633
01:10:36,623 --> 01:10:37,972
I don't feel so nervous.

1634
01:10:38,016 --> 01:10:39,496
I feel like things have changed, somehow,

1635
01:10:39,539 --> 01:10:41,193
has recently gained perspective,

1636
01:10:41,237 --> 01:10:44,240
and, you know, everything is going to be okay.

1637
01:10:47,330 --> 01:10:48,853
The Prime Minister has announced

1638
01:10:48,896 --> 01:10:51,290
The most drastic limits for our lives

1639
01:10:51,334 --> 01:10:53,858
that the United Kingdom has ever seen in its lifetime.

1640
01:10:53,901 --> 01:10:55,033
I must give it to the British people

1641
01:10:55,076 --> 01:10:56,904
A very simple instruction.

1642
01:10:56,948 --> 01:10:59,037
You must stay home.

1643
01:10:59,080 --> 01:11:02,519
It seems like we are in a science fiction novel.

1644
01:11:02,562 --> 01:11:04,869
You know, I'm delivering food to my parents,

1645
01:11:04,912 --> 01:11:08,220
ensuring they remain isolated and safe.

1646
01:11:08,264 --> 01:11:10,570
I think it just highlights the incredible need

1647
01:11:10,614 --> 01:11:12,877
for AI-assisted science.

1648
01:11:17,098 --> 01:11:18,361
you always know

1649
01:11:18,404 --> 01:11:21,015
Something like this is a possibility.

1650
01:11:21,059 --> 01:11:23,888
But no one really believes that will happen.

1651
01:11:23,931 --> 01:11:25,585
in his life, however.

1652
01:11:26,978 --> 01:11:29,154
-Are you already recording? -Yeah.

1653
01:11:29,197 --> 01:11:31,025
- Well, good morning everyone. - Hello.

1654
01:11:31,069 --> 01:11:32,679
Good. CASP has started.

1655
01:11:32,723 --> 01:11:36,074
It's nice to be able to sit in my pajamas all day.

1656
01:11:36,117 --> 01:11:37,597
I never thought I would live in a house.

1657
01:11:37,641 --> 01:11:39,164
where so many things were happening.

1658
01:11:39,207 --> 01:11:41,427
I'd be trying to solve protein folding in a room,

1659
01:11:41,471 --> 01:11:42,559
and my husband would be trying

1660
01:11:42,602 --> 01:11:43,908
to make robots walk on the other.

1661
01:11:46,954 --> 01:11:49,392
One of the toughest proteins we have sourced at CASP so far.

1662
01:11:49,435 --> 01:11:51,219
is the SARS-CoV-2 protein

1663
01:11:51,263 --> 01:11:52,220
called ORF8.

1664
01:11:52,264 --> 01:11:54,919
ORF8 is a coronavirus protein.

1665
01:11:54,962 --> 01:11:56,964
It's one of the main proteins, huh,

1666
01:11:57,008 --> 01:11:58,749
which weakens the immune system.

1667
01:11:58,792 --> 01:12:00,054
We try very hard

1668
01:12:00,098 --> 01:12:01,752
to improve our prediction.

1669
01:12:01,795 --> 01:12:03,493
Really, really very difficult.

1670
01:12:03,536 --> 01:12:05,582
Probably the most time we've ever spent

1671
01:12:05,625 --> 01:12:07,105
on a single objective.

1672
01:12:07,148 --> 01:12:08,933
To the point where my husband is like,

1673
01:12:08,976 --> 01:12:12,197
"It's midnight. You have to go to sleep."

1674
01:12:12,240 --> 01:12:16,419
So I think we're on day 102 since lockdown.

1675
01:12:16,462 --> 01:12:19,944
My daughter keeps a diary.

1676
01:12:19,987 --> 01:12:22,120
Now you can go out as much as you want.

1677
01:12:25,036 --> 01:12:27,212
We have received the last objective.

1678
01:12:27,255 --> 01:12:29,649
They have said they will not send any more targets.

1679
01:12:29,693 --> 01:12:31,347
in our CASP category.

1680
01:12:32,652 --> 01:12:33,653
So we're just making sure

1681
01:12:33,697 --> 01:12:35,481
We get the best possible response.

1682
01:12:40,530 --> 01:12:43,315
As soon as we start getting the results,

1683
01:12:43,359 --> 01:12:48,233
I would sit down and start observing how close someone was.

1684
01:12:48,276 --> 01:12:50,583
to achieve correct protein structures.

1685
01:13:00,245 --> 01:13:01,551
- Oh, hello. -Hello.

1686
01:13:03,814 --> 01:13:07,078
It's an incredible thing, CASP is finally over.

1687
01:13:07,121 --> 01:13:09,472
I think it's at least time to raise a glass.

1688
01:13:09,515 --> 01:13:11,212
Um, I don't know if everyone has a glass.

1689
01:13:11,256 --> 01:13:12,823
of something they can raise.

1690
01:13:12,866 --> 01:13:14,955
If not, pick up, I don't know, your laptops.

1691
01:13:14,999 --> 01:13:17,088
One...

1692
01:13:17,131 --> 01:13:18,611
I'll probably give a speech in a minute.

1693
01:13:18,655 --> 01:13:20,483
I feel like I should but I have no idea what to say.

1694
01:13:21,005 --> 01:13:24,269
Well... let's see.

1695
01:13:24,312 --> 01:13:27,054
I feel like I'm reading an email...

1696
01:13:27,098 --> 01:13:28,534
It's the right thing to do.

1697
01:13:29,927 --> 01:13:31,232
When Juan said:

1698
01:13:31,276 --> 01:13:33,191
"I'm going to read an email" at a team social gathering.

1699
01:13:33,234 --> 01:13:35,498
I thought, "Wow, John, you sure know how to have fun."

1700
01:13:35,541 --> 01:13:38,370
Now let's read an email.

1701
01:13:38,414 --> 01:13:41,634
Uh, I received this around four today.

1702
01:13:42,722 --> 01:13:44,724
Umm, it's by John Moult.

1703
01:13:45,725 --> 01:13:47,031
And I'll just read it.

1704
01:13:47,074 --> 01:13:49,381
He says: "As I assume you already know,

1705
01:13:49,425 --> 01:13:53,603
"Your group has done surprisingly well in CASP 14,

1706
01:13:53,646 --> 01:13:55,387
"both in relation to other groups

1707
01:13:55,431 --> 01:13:57,911
"and with absolute model precision."

1708
01:13:59,826 --> 01:14:01,219
"Congratulations on this work.

1709
01:14:01,262 --> 01:14:03,047
"It's really extraordinary."

1710
01:14:03,090 --> 01:14:05,266
The structures were so good,

1711
01:14:05,310 --> 01:14:07,443
It was... it was just amazing.

1712
01:14:09,140 --> 01:14:10,750
After half a century,

1713
01:14:10,794 --> 01:14:12,230
Finally we have a solution

1714
01:14:12,273 --> 01:14:14,928
to the problem of protein folding.

1715
01:14:14,972 --> 01:14:17,409
When I saw this email, I read it,

1716
01:14:17,453 --> 01:14:19,585
I'm like, "Oh, shit!"

1717
01:14:19,629 --> 01:14:21,587
And my wife says, "Is everything okay?"

1718
01:14:21,631 --> 01:14:24,242
I call my parents and say, "Hi, Mom."

1719
01:14:24,285 --> 01:14:26,244
"Umm, I have something to tell you.

1720
01:14:26,287 --> 01:14:27,550
"We have done this

1721
01:14:27,593 --> 01:14:29,813
"And it could be something very important."

1722
01:14:29,856 --> 01:14:31,641
When I found out the results of CASP 14,

1723
01:14:32,642 --> 01:14:34,034
I was stunned.

1724
01:14:34,078 --> 01:14:35,819
I was just excited.

1725
01:14:35,862 --> 01:14:38,909
This is a problem I was starting to think about.

1726
01:14:38,952 --> 01:14:42,086
It wouldn't be resolved in my lifetime.

1727
01:14:42,129 --> 01:14:44,741
We now have a tool that can be used

1728
01:14:44,784 --> 01:14:46,612
practically by scientists.

1729
01:14:46,656 --> 01:14:48,440
These people ask us, you know,

1730
01:14:48,484 --> 01:14:50,224
"I have this protein involved in malaria,"

1731
01:14:50,268 --> 01:14:52,139
or you know, some infectious disease.

1732
01:14:52,183 --> 01:14:53,227
"We don't know the structure.

1733
01:14:53,271 --> 01:14:55,186
"Can we use AlphaFold to solve it?"

1734
01:14:55,229 --> 01:14:56,970
We can easily predict all known sequences

1735
01:14:57,014 --> 01:14:58,276
in a month.

1736
01:14:58,319 --> 01:14:59,973
All known sequences in one month?

1737
01:15:00,017 --> 01:15:01,279
- Yes, easily. - Mmm-hmm?

1738
01:15:01,322 --> 01:15:02,585
One trillion, two billion.

1739
01:15:02,628 --> 01:15:03,673
Um, and they are...

1740
01:15:03,716 --> 01:15:05,196
Why don't we do that? Yes.

1741
01:15:05,239 --> 01:15:07,111
- We should do that a lot. - Well, I mean...

1742
01:15:07,154 --> 01:15:09,243
That's much better. Why don't we do that?

1743
01:15:09,287 --> 01:15:11,115
So that's one of the options.

1744
01:15:11,158 --> 01:15:12,638
-Good. - There is this...

1745
01:15:12,682 --> 01:15:15,119
We should just... Well, that's a great idea.

1746
01:15:15,162 --> 01:15:17,513
We should simply analyze all existing proteins.

1747
01:15:18,296 --> 01:15:19,471
And then release that.

1748
01:15:19,515 --> 01:15:20,994
Why didn't anyone suggest this before?

1749
01:15:21,038 --> 01:15:22,126
Of course that's what we should do.

1750
01:15:22,169 --> 01:15:23,954
Why are we thinking about doing a service?

1751
01:15:23,997 --> 01:15:25,651
And then people send their proteins?

1752
01:15:25,695 --> 01:15:26,913
We just folded everything.

1753
01:15:26,957 --> 01:15:28,654
And then give it to everyone.

1754
01:15:28,698 --> 01:15:31,483
Who knows how many discoveries will be made from this?

1755
01:15:31,527 --> 01:15:33,790
Demis called us and told us:

1756
01:15:33,833 --> 01:15:35,618
"We want this to be open.

1757
01:15:35,661 --> 01:15:37,837
"Not only do you have to make sure that the source is open,

1758
01:15:37,881 --> 01:15:39,578
"But we'll make it really easy."

1759
01:15:39,622 --> 01:15:42,668
"so that everyone has access to the predictions."

1760
01:15:45,062 --> 01:15:47,238
That's fantastic.

1761
01:15:47,281 --> 01:15:49,327
It's like pulling back the curtain.

1762
01:15:49,370 --> 01:15:52,852
and see the whole world of protein structures.

1763
01:15:55,246 --> 01:15:56,987
They freed the structures

1764
01:15:57,030 --> 01:15:59,772
of 200 million proteins.

1765
01:15:59,816 --> 01:16:01,818
These are gifts to humanity.

1766
01:16:07,650 --> 01:16:10,914
By the time AlphaFold is available to the world,

1767
01:16:10,957 --> 01:16:13,873
We will no longer be the most important people

1768
01:16:13,917 --> 01:16:15,222
in AlphaFold history.

1769
01:16:15,266 --> 01:16:16,833
I can't believe it's all come to light.

1770
01:16:16,876 --> 01:16:18,356
Oh!

1771
01:16:18,399 --> 01:16:20,314
One hundred and sixty-four users.

1772
01:16:20,358 --> 01:16:22,578
Lots of activity in Japan.

1773
01:16:22,621 --> 01:16:24,928
We currently have 655 users.

1774
01:16:24,971 --> 01:16:26,930
We currently have 100,000 simultaneous users.

1775
01:16:26,973 --> 01:16:28,192
Wow!

1776
01:16:31,108 --> 01:16:33,893
Today is a crazy day.

1777
01:16:33,937 --> 01:16:36,504
What an absolutely incredible effort!

1778
01:16:36,548 --> 01:16:37,723
From all over the world.

1779
01:16:37,767 --> 01:16:38,550
We will all remember these moments.

1780
01:16:38,594 --> 01:16:40,030
for the rest of our lives.

1781
01:16:40,073 --> 01:16:41,727
I'm excited about AlphaFold.

1782
01:16:41,771 --> 01:16:45,601
As far as my research is concerned, it is already driving a lot of progress.

1783
01:16:45,644 --> 01:16:47,385
And this is just the beginning.

1784
01:16:47,428 --> 01:16:48,908
My guess is that

1785
01:16:48,952 --> 01:16:53,043
Every biological and chemical achievement

1786
01:16:53,086 --> 01:16:55,698
It will be related to AlphaFold in some way.

1787
01:17:13,367 --> 01:17:15,413
AlphaFold is a moment index.

1788
01:17:15,456 --> 01:17:18,068
It is a moment that people will not forget.

1789
01:17:18,111 --> 01:17:20,244
Because the world changed.

1790
01:17:39,655 --> 01:17:41,482
Everyone has realized now

1791
01:17:41,526 --> 01:17:43,746
What Shane and I have known for over 20 years,

1792
01:17:43,789 --> 01:17:46,618
that AI is going to be the most important thing

1793
01:17:46,662 --> 01:17:48,446
that humanity will never invent.

1794
01:17:48,489 --> 01:17:50,230
We will arrive shortly

1795
01:17:50,274 --> 01:17:52,058
at our final destination.

1796
01:18:02,068 --> 01:18:04,767
The pace of innovation and capabilities

1797
01:18:04,810 --> 01:18:06,507
is accelerating,

1798
01:18:06,551 --> 01:18:09,293
like a rock that rolls down the hill from which we have kicked

1799
01:18:09,336 --> 01:18:12,644
And now it continues to gain speed.

1800
01:18:12,688 --> 01:18:15,299
We are at a crossroads in human history.

1801
01:18:15,342 --> 01:18:16,735
AI has the potential

1802
01:18:16,779 --> 01:18:19,172
to transform our lives in all aspects.

1803
01:18:19,216 --> 01:18:23,786
It is no less important than the discovery of electricity.

1804
01:18:23,829 --> 01:18:26,484
We should be looking at the scientific method.

1805
01:18:26,527 --> 01:18:28,834
and trying to understand every step of the way

1806
01:18:28,878 --> 01:18:30,096
rigorously.

1807
01:18:30,140 --> 01:18:32,664
This is a moment of profound opportunity.

1808
01:18:32,708 --> 01:18:34,753
Take advantage of this technology

1809
01:18:34,797 --> 01:18:37,713
It could eclipse anything we've known so far.

1810
01:18:42,195 --> 01:18:43,675
Hello, Alpha.

1811
01:18:44,676 --> 01:18:45,764
Hello.

1812
01:18:47,157 --> 01:18:48,419
What is this?

1813
01:18:50,682 --> 01:18:53,729
This is a chess board.

1814
01:18:53,772 --> 01:18:56,514
If I had to play white, what move would you recommend?

1815
01:18:59,865 --> 01:19:00,953
I would recommend it

1816
01:19:00,997 --> 01:19:02,781
move your pawn from E2 to E4.

1817
01:19:05,871 --> 01:19:08,787
And now, if you were black, what would you play now?

1818
01:19:11,572 --> 01:19:13,618
I would play the Sicilian Defense.

1819
01:19:15,838 --> 01:19:16,882
That's a good choice.

1820
01:19:19,406 --> 01:19:21,452
Thank you.

1821
01:19:23,715 --> 01:19:25,891
So what do you see? What is this object?

1822
01:19:28,546 --> 01:19:30,504
This is a pencil sculpture.

1823
01:19:32,811 --> 01:19:35,031
What happens if I move one of the pencils?

1824
01:19:37,990 --> 01:19:39,470
If you move one of the pencils,

1825
01:19:39,513 --> 01:19:42,081
The sculpture will crumble.

1826
01:19:42,125 --> 01:19:44,301
I'd better leave him alone then.

1827
01:19:44,344 --> 01:19:45,868
It's probably a good idea.

1828
01:19:50,568 --> 01:19:52,744
The IAG is already on the horizon.

1829
01:19:54,833 --> 01:19:56,661
Very clearly the next generation

1830
01:19:56,704 --> 01:19:58,141
will live in a future world

1831
01:19:58,184 --> 01:20:01,057
where things will be radically different thanks to AI.

1832
01:20:02,493 --> 01:20:05,496
And if you want to manage it responsibly,

1833
01:20:05,539 --> 01:20:09,239
Every moment is vital.

1834
01:20:09,282 --> 01:20:12,677
This is the moment I've been living my whole life.

1835
01:20:19,162 --> 01:20:21,120
It's just a good thinking game.

1836
01:20:22,305 --> 01:21:22,495
Support us and become a VIP member To
remove all ads