1
00:00:00,000 --> 00:00:05,440
Superintelligence is something we cannot comprehend or predict how it's fighting us.

2
00:00:05,840 --> 00:00:09,560
Oh yeah, it's only 20-30% probability of doom for everyone.

3
00:00:09,840 --> 00:00:11,160
That's an insane number.

4
00:00:11,560 --> 00:00:15,900
If you are creating a dangerous weapon, giving it to everyone in the world,

5
00:00:15,980 --> 00:00:20,120
including all the psychopaths and doomsday calls, is not a good idea.

6
00:00:20,540 --> 00:00:22,400
With AI, you don't understand them.

7
00:00:22,740 --> 00:00:23,480
They're immortal.

8
00:00:23,900 --> 00:00:24,880
You will not be rich.

9
00:00:24,920 --> 00:00:25,840
You will not be famous.

10
00:00:25,960 --> 00:00:27,240
You will not be in history books.

11
00:00:27,280 --> 00:00:27,960
You'll be dead.

12
00:00:27,960 --> 00:00:29,080
Why are you doing it?

13
00:00:29,080 --> 00:00:30,520
It's mutually assured destruction.

14
00:00:31,100 --> 00:00:33,500
You can't take control back if you're not happy.

15
00:00:33,680 --> 00:00:36,200
There is no undo button once you surrender control.

16
00:00:37,060 --> 00:00:38,720
What does it mean to have freedom?

17
00:00:39,020 --> 00:00:41,980
Maybe what you want is actually exactly what they're providing.

18
00:00:42,560 --> 00:00:46,300
If you want to worry about Bitcoin in a future world, worry about, again,

19
00:00:46,300 --> 00:00:51,600
what if aliens come and the compute necessary for 51% is what they have in a cell phone?

20
00:00:52,060 --> 00:00:55,300
What if in a simulation externally they have the private keys?

21
00:00:55,600 --> 00:00:58,880
You can have all these scenarios where the whole economy collapses.

22
00:00:59,080 --> 00:01:07,420
Roman, I really appreciate you agreeing to do this. I've listened to you on a number of podcasts.

23
00:01:07,700 --> 00:01:11,120
So I've been casually following this like AI safety talk for a little while.

24
00:01:11,740 --> 00:01:16,100
I heard you on Rogan, I heard you on Lex. And then recently I was listening to you on Diary of a CEO.

25
00:01:16,700 --> 00:01:23,960
And a few times in that show, you brought up Bitcoin. And Stephen, maybe like slightly just

26
00:01:23,960 --> 00:01:26,800
glossed over it the first couple of times. Then you did talk about it a little bit,

27
00:01:26,800 --> 00:01:28,260
But I was like, I need to speak to this man.

28
00:01:29,120 --> 00:01:30,880
So we're going to get into all of that today.

29
00:01:31,200 --> 00:01:33,520
But I think I want to start on all the AI safety stuff,

30
00:01:33,580 --> 00:01:35,320
because that is obviously your bread and butter.

31
00:01:35,700 --> 00:01:38,560
Do you want to explain how you got into this and what you do?

32
00:01:39,300 --> 00:01:40,600
I'm a computer scientist.

33
00:01:40,760 --> 00:01:45,040
I research AI safety, and I started with cybersecurity biometrics.

34
00:01:45,140 --> 00:01:48,120
I was interested in making computer systems more secure.

35
00:01:48,720 --> 00:01:51,640
I looked at securing online gambling,

36
00:01:51,640 --> 00:01:57,640
poker sites. And at the time, they just started having infestation of poker bars.

37
00:01:58,280 --> 00:02:02,880
So I was looking at how we can detect them, prevent them from stealing resources.

38
00:02:03,760 --> 00:02:08,100
And they were getting better. So I kind of projected forward, how good can they get?

39
00:02:08,160 --> 00:02:11,920
At some point, they will get human level capabilities and beyond.

40
00:02:11,920 --> 00:02:15,920
And so that's my story of getting into very safety.

41
00:02:15,920 --> 00:02:22,100
Obviously, I did some additional work with more advanced concerns later on,

42
00:02:22,160 --> 00:02:24,060
but that was the initial entry point.

43
00:02:24,640 --> 00:02:29,680
And when did this fear go from being almost like a science fiction problem to being real?

44
00:02:29,780 --> 00:02:32,680
Because as someone who sort of casually followed the AI stuff,

45
00:02:32,760 --> 00:02:36,660
it probably wasn't until maybe even like ChatGPT3 that I started thinking,

46
00:02:36,740 --> 00:02:37,780
oh, this is actually really useful.

47
00:02:38,140 --> 00:02:40,900
There's like a tool here that I can use to make me much more productive.

48
00:02:41,580 --> 00:02:43,820
Was it those breakthroughs that made you think,

49
00:02:43,820 --> 00:02:47,120
okay, this is getting real from a much more dystopian viewpoint.

50
00:02:47,420 --> 00:02:48,520
So that shifted timelines.

51
00:02:48,900 --> 00:02:50,720
We always knew it's going to happen,

52
00:02:50,880 --> 00:02:54,340
just natural progression of what we see in the field.

53
00:02:54,520 --> 00:02:57,340
But it surprised us in how quickly it came.

54
00:02:57,600 --> 00:03:02,120
Everyone was thinking we got another 20 years or so, 2045 maybe,

55
00:03:02,260 --> 00:03:05,200
but it started happening a lot sooner.

56
00:03:05,620 --> 00:03:08,360
And at the same time, today there are people who are denying

57
00:03:08,360 --> 00:03:09,220
what this has happened.

58
00:03:09,440 --> 00:03:10,500
It will never happen.

59
00:03:10,740 --> 00:03:11,660
Humans are special.

60
00:03:11,660 --> 00:03:15,900
So we can never get AI to do creative artwork or something like that.

61
00:03:15,920 --> 00:03:16,780
So it's interesting.

62
00:03:16,900 --> 00:03:18,560
At the same time, you have people who are like,

63
00:03:18,800 --> 00:03:21,440
it's too late and nothing has started.

64
00:03:22,280 --> 00:03:25,220
It's interesting because you said the creative pit there.

65
00:03:25,340 --> 00:03:27,140
And that was one of the things I wanted to talk to you about.

66
00:03:27,940 --> 00:03:30,840
Because right now, AI is very useful.

67
00:03:31,620 --> 00:03:34,200
It definitely helps me be more efficient in the stuff that I'm doing.

68
00:03:34,300 --> 00:03:35,300
But it's still flawed.

69
00:03:35,300 --> 00:03:36,480
There's still things that gets wrong.

70
00:03:36,580 --> 00:03:38,080
There's still areas that it's not helpful.

71
00:03:38,480 --> 00:03:40,200
It's too much of a sycophant for my liking.

72
00:03:40,200 --> 00:03:42,460
It just wants to tell me I'm great at everything

73
00:03:42,460 --> 00:03:44,440
instead of actually giving me useful information.

74
00:03:45,480 --> 00:03:49,400
But we are already seeing it start to impact the job market

75
00:03:49,400 --> 00:03:50,140
in different ways.

76
00:03:50,260 --> 00:03:52,480
And I think we're probably at the very, very start of that.

77
00:03:52,800 --> 00:03:56,860
But what has to happen for it to go from being this pretty useful tool

78
00:03:56,860 --> 00:03:57,760
that makes people more productive

79
00:03:57,760 --> 00:04:00,820
to being something that potentially kills us all?

80
00:04:01,280 --> 00:04:05,420
So that's a paradigm shift from tools, AI you use,

81
00:04:05,500 --> 00:04:08,080
you decide how to use it, you are in charge,

82
00:04:08,080 --> 00:04:09,580
to AI as an agent.

83
00:04:09,580 --> 00:04:13,960
So we see more and more companies trying to fully automate the whole process.

84
00:04:14,120 --> 00:04:14,920
It's not just a tool.

85
00:04:15,200 --> 00:04:17,600
It writes the whole software product.

86
00:04:18,140 --> 00:04:24,100
It makes its own decisions, what to use to code it, what hardware to run.

87
00:04:24,520 --> 00:04:33,040
And so the more independent it becomes, the more agentic it is, the more it sets its own goals and has side effects of setting those goals.

88
00:04:33,240 --> 00:04:35,000
And we don't control what they are.

89
00:04:35,000 --> 00:04:39,500
We can predict what they're going to use as a path to achieve those goals.

90
00:04:39,580 --> 00:04:45,920
We may understand the larger goal, but not all the side effects of different paths to get to that goal.

91
00:04:45,920 --> 00:04:47,620
And that's where the danger comes.

92
00:04:48,080 --> 00:04:49,040
It is smarter than you.

93
00:04:49,360 --> 00:04:52,060
It makes plans you are not dictating.

94
00:04:52,680 --> 00:04:58,400
And the way it implements them has zero concern about your safety or security or being.

95
00:04:59,220 --> 00:05:02,340
How do you calculate when it's smarter than humans?

96
00:05:02,440 --> 00:05:06,540
Because I think in almost every way, it's probably already smarter than me.

97
00:05:06,660 --> 00:05:09,120
The thing that it doesn't have is any kind of emotional intelligence.

98
00:05:09,120 --> 00:05:11,520
or at least what I've seen.

99
00:05:11,900 --> 00:05:15,100
So when is it that it goes from being a very good,

100
00:05:15,160 --> 00:05:16,520
almost encyclopedia of knowledge

101
00:05:16,520 --> 00:05:18,660
to being actually more intelligent than humans?

102
00:05:19,620 --> 00:05:22,040
So there are standard tools for measuring intelligence,

103
00:05:22,280 --> 00:05:23,380
IQ tests and such.

104
00:05:23,660 --> 00:05:26,820
On that metric, I think it's already definitely smarter than average.

105
00:05:27,580 --> 00:05:31,000
It is definitely more diverse than its capabilities.

106
00:05:31,000 --> 00:05:32,900
So it speaks hundreds of languages,

107
00:05:33,160 --> 00:05:34,660
it plays every musical instrument.

108
00:05:34,840 --> 00:05:36,640
It's better than a typical human

109
00:05:36,640 --> 00:05:39,160
if you average out over all the domains.

110
00:05:39,320 --> 00:05:41,980
And that's kind of the measure of intelligence

111
00:05:41,980 --> 00:05:44,640
DeepMind was proposing a long time ago.

112
00:05:45,040 --> 00:05:46,820
What is the average performance

113
00:05:46,820 --> 00:05:48,400
across all possible domains?

114
00:05:49,120 --> 00:05:51,560
And so it may still suck at certain things,

115
00:05:51,680 --> 00:05:53,860
but on average, it's still so much better.

116
00:05:54,600 --> 00:05:56,220
As far as emotional intelligence,

117
00:05:56,360 --> 00:05:57,960
I think it's also better at that.

118
00:05:58,360 --> 00:06:00,360
It understands your emotional states.

119
00:06:00,440 --> 00:06:02,820
It just doesn't have its own emotional states

120
00:06:02,820 --> 00:06:05,940
where we would expect them to be in another human.

121
00:06:05,940 --> 00:06:06,480
Okay.

122
00:06:06,640 --> 00:06:08,080
It may have preferences.

123
00:06:08,080 --> 00:06:14,380
It may be unhappy with reward signal it's getting based on some performance, but it's

124
00:06:14,380 --> 00:06:16,420
not like internally angry at you.

125
00:06:17,020 --> 00:06:17,040
Okay.

126
00:06:17,260 --> 00:06:21,320
And so when it gets to superintelligence is when this becomes an existential risk to humanity.

127
00:06:21,960 --> 00:06:25,900
Even at AGI levels and pre-AGI levels, there are certain concerns.

128
00:06:26,020 --> 00:06:30,940
At pre-AGI levels, concern is malevolent human actors giving it malevolent payloads.

129
00:06:30,940 --> 00:06:34,960
So somebody tells it develop very deadly virus and then releasing it.

130
00:06:34,960 --> 00:06:41,360
At AGI level, it's like having evil human trying to do bad things, and they have a lot of resources.

131
00:06:41,820 --> 00:06:46,260
Whereas superintelligence is something we cannot comprehend or predict how it's fighting us.

132
00:06:46,900 --> 00:06:54,960
Okay, so in terms of controls we can put in place, is it at the superintelligence level that those controls become almost like an oxymoron?

133
00:06:55,000 --> 00:06:56,960
You can't control something far more intelligent than you.

134
00:06:57,140 --> 00:07:02,140
It's definitely not possible, in my opinion, to control superintelligence indefinitely.

135
00:07:02,140 --> 00:07:05,140
I think at level of AGI, it's challenging.

136
00:07:05,520 --> 00:07:08,480
It's still a very intelligent adversary,

137
00:07:08,720 --> 00:07:12,500
but we at least may understand some of the tools it's using.

138
00:07:13,340 --> 00:07:16,240
So we can still fail, just like in cybersecurity,

139
00:07:16,760 --> 00:07:19,680
some hackers manage to penetrate your safety,

140
00:07:19,840 --> 00:07:22,120
but it's an even fight.

141
00:07:22,560 --> 00:07:26,320
Whereas with superintelligence, it's just squarrows versus humans.

142
00:07:27,120 --> 00:07:29,900
Okay, and then give me a kind of timeline.

143
00:07:29,900 --> 00:07:32,120
How far away from this do you think we are?

144
00:07:32,140 --> 00:07:51,140
So nobody knows. That's the thing. We switched from it takes that long to it takes that much money. So if tomorrow China decides to put $5 trillion into it, that will expedite the process a lot. Maybe we already have enough computational resources where enough money trains you to superintelligence.

145
00:07:51,140 --> 00:07:55,560
superintelligence. So right now, prediction markets are saying two, three years to AGI.

146
00:07:56,240 --> 00:08:00,520
They've been saying it for two or three years, so it doesn't mean that much. And some people say,

147
00:08:00,640 --> 00:08:05,740
we already have AGI. Again, those things are better than most people in most domains.

148
00:08:06,720 --> 00:08:13,300
Typically, if I'm looking at maybe a master's level student, I would pick AI over it. Maybe

149
00:08:13,300 --> 00:08:18,320
there are still some top PhD students who beat current systems, but that's quickly shifting.

150
00:08:19,280 --> 00:08:24,060
So depending on your definitions, we either have AGI, we'll have it very soon,

151
00:08:24,240 --> 00:08:28,540
or even if it takes 5-10 years, it changes nothing because the same concerns remain.

152
00:08:29,540 --> 00:08:31,080
Maybe you can help me understand something here,

153
00:08:31,120 --> 00:08:35,100
because I think I had the wrong definition in my head of what AGI or superintelligence was.

154
00:08:35,260 --> 00:08:40,160
Because I thought that was when it would work sort of autonomously without prompts.

155
00:08:40,500 --> 00:08:41,500
And we don't have that yet.

156
00:08:42,960 --> 00:08:47,380
So you can create a system which is just an agent placed in the loop,

157
00:08:47,380 --> 00:08:57,040
You tell it, okay, generate a list of goals and then just kind of work through the loop trying to do some modular progress on those goals, generate sub goals.

158
00:08:57,180 --> 00:09:00,000
So it's kind of like agent-like, but not fully at arms.

159
00:09:01,640 --> 00:09:08,520
With superintelligence, we're expecting that it will manage to figure out how to provide for its own survival.

160
00:09:09,000 --> 00:09:11,220
It can generate its own hardware.

161
00:09:11,620 --> 00:09:16,360
It can bypass human logistical framework for sustaining itself.

162
00:09:16,360 --> 00:09:23,420
Okay, so then when we get to AGI or superintelligence, whether that's in two years or 10 years, I don't think that really matters.

163
00:09:23,680 --> 00:09:26,740
It's what happens next, and what is that?

164
00:09:26,920 --> 00:09:30,020
What is the paradigm shift once we answer that?

165
00:09:30,020 --> 00:09:36,180
So that's what people call singularity point, where you cannot make predictions about what happens after that point,

166
00:09:36,260 --> 00:09:39,280
because you're not smart enough to comprehend those concepts.

167
00:09:39,840 --> 00:09:45,860
Research in physics, novel science, engineering, self-improvement progress in the AI system itself,

168
00:09:45,860 --> 00:09:49,060
becomes so fast, we never comprehend it,

169
00:09:49,120 --> 00:09:51,640
nor have time to fully comprehend it if it was possible.

170
00:09:52,120 --> 00:09:55,460
But there's plenty of good that would likely come out of this.

171
00:09:55,580 --> 00:09:57,600
Like, I assume this cures all diseases.

172
00:09:58,140 --> 00:10:02,100
It has physics breakthroughs that allows us to go to the stars.

173
00:10:02,180 --> 00:10:03,100
If we control it.

174
00:10:03,420 --> 00:10:05,980
If we tell it what to do and that's what we tell it,

175
00:10:06,020 --> 00:10:07,280
yes, it's a miracle.

176
00:10:07,460 --> 00:10:10,560
It cures all the diseases, free stuff, absolutely.

177
00:10:10,560 --> 00:10:12,180
But we don't know how to control it.

178
00:10:12,800 --> 00:10:14,400
And so the risk there is,

179
00:10:14,400 --> 00:10:18,120
instead of getting all these amazing benefits from it, it literally kills us all.

180
00:10:18,980 --> 00:10:19,420
Absolutely.

181
00:10:19,720 --> 00:10:23,240
And what percentage chance do you think that is the outcome,

182
00:10:23,320 --> 00:10:25,300
that it does attack humanity in some way?

183
00:10:25,940 --> 00:10:29,360
Well, if you're not controlling it, you really don't know what it's going to do.

184
00:10:29,360 --> 00:10:34,320
It may not directly attack us, but do something which is, as a side effect,

185
00:10:34,420 --> 00:10:35,580
impacts all of us negatively.

186
00:10:36,480 --> 00:10:41,220
If there was a tiny chance, 1% chance it kills everyone,

187
00:10:41,620 --> 00:10:43,640
I don't care how much free stuff you're giving us.

188
00:10:43,640 --> 00:10:46,620
You can't just gamble 8 billion people on some free money.

189
00:10:47,480 --> 00:10:49,520
But you think it's higher than 1%?

190
00:10:50,560 --> 00:10:55,820
I think it's extremely high because if you look at space of all possible universes,

191
00:10:56,520 --> 00:11:01,860
most of them are not human-friendly in terms of basic physics, temperature, gravity, oxygen,

192
00:11:02,080 --> 00:11:03,940
but also in terms of other properties.

193
00:11:04,120 --> 00:11:05,640
We have to be very particular.

194
00:11:05,860 --> 00:11:09,680
We want this world, this type of world, that type of food is available,

195
00:11:10,640 --> 00:11:12,580
temperatures range from X to Y.

196
00:11:12,580 --> 00:11:19,540
But if it's not decided by us, then someone who doesn't care about us would set very unhuman

197
00:11:19,540 --> 00:11:21,240
volumes for those variables.

198
00:11:22,240 --> 00:11:27,700
So why do you think people like Sam Altman, Elon, all the people that are building these

199
00:11:27,700 --> 00:11:29,840
businesses are pushing so fast?

200
00:11:30,200 --> 00:11:33,920
Because I've heard Sam Altman in the past accept the risk.

201
00:11:34,900 --> 00:11:37,520
So what's his incentive to do this?

202
00:11:37,600 --> 00:11:40,820
Is it just all down to sort of fiduciary duty of running that company?

203
00:11:40,820 --> 00:11:48,820
So there is multiple things we can consider. One is they kind of realize that this is happening no matter what they do.

204
00:11:48,820 --> 00:11:55,820
So if Elon stops and for many years he was not building super intelligence, Sam Alckmin is still building it.

205
00:11:55,820 --> 00:12:01,820
So Elon thinks, I can probably do a safer AGI than Sam. So he jumps into the game.

206
00:12:01,820 --> 00:12:07,440
They're all kind of hoping that everyone else stops.

207
00:12:07,440 --> 00:12:12,440
The government comes in and freezes research and development at that stage, and they capture

208
00:12:12,440 --> 00:12:14,940
most of the benefits by being the most advanced model.

209
00:12:14,940 --> 00:12:18,680
Do you think that's how they're thinking about it?

210
00:12:18,680 --> 00:12:20,380
It's hard to envision other options.

211
00:12:20,380 --> 00:12:23,640
So they're all on record as being super concerned about AI safety.

212
00:12:23,640 --> 00:12:28,220
Elon funded it, assembled blog posts about how it slides out for humanity.

213
00:12:28,220 --> 00:12:29,980
So they all know the risks.

214
00:12:29,980 --> 00:12:32,560
It's not like they just never heard of it or don't believe it.

215
00:12:32,800 --> 00:12:35,540
They are in agreement that this is super dangerous.

216
00:12:35,800 --> 00:12:40,800
And then you ask them, they're like, oh, yeah, it's only 20%, 30% probability of doom for everyone.

217
00:12:41,060 --> 00:12:42,420
That's an insane number.

218
00:12:43,180 --> 00:12:44,440
That's not a number I'm happy with.

219
00:12:45,820 --> 00:12:50,280
But the problem with the regulation is, unless it's some kind of global regulation,

220
00:12:50,640 --> 00:12:52,280
if that happens just within the United States,

221
00:12:52,280 --> 00:12:55,360
then you'd imagine China are just going to keep building towards something like this.

222
00:12:55,920 --> 00:12:57,180
Yes, and it's even worse.

223
00:12:57,180 --> 00:13:01,640
even if you have regulations, it's like kind of saying, well, crime is illegal. Does crime still

224
00:13:01,640 --> 00:13:07,600
happen? Of course. Right. So as long as there is one person somewhere still typing away,

225
00:13:08,120 --> 00:13:12,800
eventually they're going to get there. So at this point then, is superintelligence inevitable?

226
00:13:14,040 --> 00:13:19,000
So we can get lucky and have something horrible happen to us. Like if there is another nuclear

227
00:13:19,000 --> 00:13:23,800
war, that would pull it back a few centuries. But other than that, we're making very good progress.

228
00:13:23,800 --> 00:13:25,980
You're saying we can get lucky with nuclear war?

229
00:13:26,760 --> 00:13:28,080
I'm being a little bit funny.

230
00:13:28,360 --> 00:13:33,080
But really, if the alternative is what we're saying is almost certain doom,

231
00:13:33,420 --> 00:13:38,120
and it's two years away, then anything which kind of gets in the way is a good outcome.

232
00:13:38,820 --> 00:13:39,120
Wow.

233
00:13:39,680 --> 00:13:42,380
Do you wish you could access cash without selling your Bitcoin?

234
00:13:43,000 --> 00:13:44,140
Well, Ledin makes that possible.

235
00:13:44,780 --> 00:13:46,760
Ledin are the global leader in Bitcoin-backed lending,

236
00:13:46,980 --> 00:13:49,880
and since 2018, they've issued over $9 billion in loans

237
00:13:49,880 --> 00:13:52,300
with a perfect record of protecting client assets.

238
00:13:52,300 --> 00:13:59,900
With Ledin you get full custody loans with no credit checks, no monthly repayments, just easy access to dollars without selling a single SAT

239
00:13:59,900 --> 00:14:08,620
As of July 1st, Ledin is Bitcoin only, meaning they exclusively offer Bitcoin backed loans with all collateral held by Ledin directly or their funding partners

240
00:14:08,620 --> 00:14:11,100
Your Bitcoin is never lent out to generate interest

241
00:14:11,100 --> 00:14:14,760
I recently took out a loan with Ledin and the whole process couldn't have been easier

242
00:14:14,760 --> 00:14:19,780
It took me less than 15 minutes to go through the application and in just a few hours I had the dollars in my account

243
00:14:19,780 --> 00:14:20,900
It was super smooth

244
00:14:20,900 --> 00:14:29,000
So if you need cash but you don't want to sell Bitcoin, head over to leden.io forward slash WBD and you'll get 0.25% off your first loan.

245
00:14:29,180 --> 00:14:33,000
That's L-E-D-N dot I-O forward slash WBD.

246
00:14:33,280 --> 00:14:41,140
Bitcoin is absolutely ripping and in every bull market there's always a new wave of investors and with it a flood of new companies, new products and new promises.

247
00:14:41,680 --> 00:14:44,640
But if you've been around long enough, you've seen how this story ends for a lot of them.

248
00:14:45,020 --> 00:14:47,840
Some cut corners, take risks with your money or just disappear.

249
00:14:47,840 --> 00:14:51,780
That's why when it comes to buying Bitcoin, the only exchange I recommend is River.

250
00:14:52,200 --> 00:14:57,760
They deeply care about doing things right for their clients and are built to last with security and transparency at their core.

251
00:14:58,460 --> 00:15:02,160
With River, you have peace of mind knowing all their Bitcoin is held in multi-sig cold storage,

252
00:15:02,460 --> 00:15:05,840
and it's the only Bitcoin-only exchange in the US with proof of reserves.

253
00:15:06,280 --> 00:15:08,320
There really is no better place to buy Bitcoin.

254
00:15:08,580 --> 00:15:14,840
So to open an account today, head over to river.com forward slash WBD and earn up to $100 in Bitcoin when you buy.

255
00:15:15,200 --> 00:15:17,820
That's river.com forward slash WBD.

256
00:15:17,840 --> 00:15:21,440
What if you could lower your tax bill and stack Bitcoin at the same time?

257
00:15:21,920 --> 00:15:23,940
Well, by mining Bitcoin with Blockware, you can.

258
00:15:24,480 --> 00:15:31,080
New tax guidelines from the Big Beautiful Bill allow American miners to write off 100% of the cost of their mining hardware in a single tax year.

259
00:15:31,400 --> 00:15:33,200
That's right, 100% write-off.

260
00:15:33,660 --> 00:15:39,120
If you have 100k in capital gains or income, you can purchase 100k of miners and offset it entirely.

261
00:15:40,140 --> 00:15:44,640
Blockware's Mining as a Service enables you to start mining Bitcoin right now without lifting a finger.

262
00:15:44,640 --> 00:15:50,960
Blockware handles everything from securing the miners to sourcing low-cost power to configuring the mining pool. They do it all

263
00:15:50,960 --> 00:15:56,300
You get to stack bitcoin at a discount every single day while also saving big come tax season

264
00:15:56,300 --> 00:16:01,780
Get started today by going to mining.blockwaresolutions.com forward slash wbd

265
00:16:01,780 --> 00:16:16,250
And for every hosted miner purchased you get one week of free hosting and electricity Of course none of this is tax advice speak with blockware to learn more at mining forward slash WBD Is there any way that this can be regulated

266
00:16:16,250 --> 00:16:21,410
at a global level that would make you more comfortable? I really hope so. There is some

267
00:16:21,410 --> 00:16:29,070
efforts at UN at the level of individual countries. European Union has AI Act, but a lot of it is

268
00:16:29,070 --> 00:16:35,550
kind of safety theater. They talk about things which are trivial, unemployment, bias in algorithms,

269
00:16:35,850 --> 00:16:41,230
privacy laws, deep fakes are very concerning to them. They completely ignore the big picture

270
00:16:41,230 --> 00:16:45,810
superintelligence and existential risks. Okay. And when you say unemployment, that's,

271
00:16:45,950 --> 00:16:50,450
I think, probably one of the most meaningful conversations around this today, as it stands,

272
00:16:50,550 --> 00:16:55,150
because it's already replacing jobs. We don't need to get to AGI and superintelligence for it

273
00:16:55,150 --> 00:16:57,910
to make a drastic impact on the economy

274
00:16:57,910 --> 00:16:58,970
from that perspective.

275
00:17:00,110 --> 00:17:02,090
How quickly do you think we'll see

276
00:17:02,090 --> 00:17:04,250
real joblessness from this?

277
00:17:04,250 --> 00:17:05,850
We're seeing it a certain degree.

278
00:17:05,850 --> 00:17:08,590
So I teach at a university and we have a co-op program.

279
00:17:08,590 --> 00:17:11,910
I think this year we are down 28% for co-op placements.

280
00:17:11,910 --> 00:17:12,750
Wow.

281
00:17:12,750 --> 00:17:15,470
So for junior programmers, the market is not great.

282
00:17:15,470 --> 00:17:16,930
If you are a senior programmer,

283
00:17:16,930 --> 00:17:19,410
if you are a machine intelligence researcher,

284
00:17:19,410 --> 00:17:22,670
you are in good shape, the pay is best it's ever been.

285
00:17:22,670 --> 00:17:24,490
But for new people just starting,

286
00:17:24,490 --> 00:17:26,530
I think they're not competitive with AI models.

287
00:17:27,530 --> 00:17:29,710
And that's really how the job has shifted.

288
00:17:29,930 --> 00:17:33,750
And you're not really just a computer scientist now, an engineer.

289
00:17:33,870 --> 00:17:34,810
You're a prompt engineer.

290
00:17:35,530 --> 00:17:37,610
Well, that's what we used to tell students.

291
00:17:37,610 --> 00:17:41,670
If you become good at prompt engineering, you'll have an excellent job in four years.

292
00:17:41,670 --> 00:17:44,470
But then we realized AI is much better at writing prompts.

293
00:17:45,030 --> 00:17:45,990
Is that already happening?

294
00:17:46,310 --> 00:17:46,710
Absolutely.

295
00:17:47,350 --> 00:17:53,250
If I need a good image, I tell a text model to generate a prompt for the image model to

296
00:17:53,250 --> 00:17:54,410
generate the image I want.

297
00:17:54,490 --> 00:17:56,450
And it takes a single sentence I provide

298
00:17:56,530 --> 00:17:59,730
and creates a paragraph-long detailed description of what I need.

299
00:18:00,370 --> 00:18:03,930
That's interesting, because that almost gets into AI being creative,

300
00:18:04,010 --> 00:18:06,730
which is something that I have kind of faded as a narrative.

301
00:18:06,810 --> 00:18:10,770
Like, I've not really, in my experience with it, which is limited,

302
00:18:10,850 --> 00:18:12,570
I've not really seen AI be creative.

303
00:18:12,650 --> 00:18:14,050
Do you think it already can be?

304
00:18:14,130 --> 00:18:16,410
It's way more creative than most humans I know.

305
00:18:16,490 --> 00:18:19,090
I prefer AI music over human music.

306
00:18:19,170 --> 00:18:19,570
Do you?

307
00:18:19,650 --> 00:18:23,090
Modern art doesn't even compete with what AI can generate, so yeah.

308
00:18:23,090 --> 00:18:25,890
Do you prefer AI music to actual music?

309
00:18:26,230 --> 00:18:30,690
Pretty much all the tunes I heard the last month actually are AI-generated.

310
00:18:30,810 --> 00:18:31,730
They're stuck in my head.

311
00:18:31,810 --> 00:18:34,150
They're super catchy, but I know they're generated.

312
00:18:34,730 --> 00:18:38,030
Is there something about authenticity that's missing, though?

313
00:18:38,190 --> 00:18:40,330
Because I think about this in terms of my job.

314
00:18:40,410 --> 00:18:42,010
Like, I go places and speak to people.

315
00:18:42,650 --> 00:18:47,130
I did not read every one of your papers, the hundreds of papers you've probably written.

316
00:18:47,910 --> 00:18:52,250
I consumed a lot of your content, but nowhere near what AI could have done.

317
00:18:52,250 --> 00:18:54,630
I could probably have prompted AI

318
00:18:54,630 --> 00:18:55,910
to have written every question

319
00:18:55,910 --> 00:18:56,890
that I'm going to ask you today

320
00:18:56,890 --> 00:18:58,190
and it would do a pretty good job of it

321
00:18:58,190 --> 00:19:00,890
and it can make me look like this on a video

322
00:19:00,890 --> 00:19:04,910
but it doesn't have the human authenticity

323
00:19:04,910 --> 00:19:06,770
that I have

324
00:19:06,770 --> 00:19:08,750
so like if it was me watching a podcast

325
00:19:08,750 --> 00:19:10,390
I would rather watch Joe Rogan

326
00:19:10,390 --> 00:19:11,650
the person I know is Joe Rogan

327
00:19:11,650 --> 00:19:12,330
asking the questions

328
00:19:12,330 --> 00:19:15,370
because I can kind of follow his train of thought

329
00:19:15,370 --> 00:19:17,030
and understand who he is as that person

330
00:19:17,030 --> 00:19:19,670
AI can never replace that, can it?

331
00:19:19,670 --> 00:19:25,130
So there is something called a Turing test, which basically says, if you can't tell

332
00:19:25,130 --> 00:19:27,490
the difference, there is no difference.

333
00:19:27,490 --> 00:19:33,610
If I can generate a video of you interviewing me, and no one in the audience can guess with

334
00:19:33,610 --> 00:19:38,730
better than 50, 50% accuracy what is real, then I don't need you.

335
00:19:38,730 --> 00:19:40,730
Hmm.

336
00:19:40,730 --> 00:19:43,230
Can AI already pass a Turing test?

337
00:19:43,230 --> 00:19:46,110
It can in most domains.

338
00:19:46,110 --> 00:19:51,590
The companies right now make it kind of not okay for them to do that, so they don't pretend

339
00:19:51,590 --> 00:19:52,590
to be humans.

340
00:19:52,590 --> 00:19:53,590
I've actually asked them.

341
00:19:53,590 --> 00:19:56,470
They explicitly say, no, no, no, I'm not a human.

342
00:19:56,470 --> 00:19:57,470
I'm an AI model.

343
00:19:57,470 --> 00:19:58,590
I don't have feelings.

344
00:19:58,590 --> 00:20:00,950
You cannot torture me.

345
00:20:00,950 --> 00:20:06,990
But in subdomains such as music, such as poetry, Turing test has been passed for decades.

346
00:20:06,990 --> 00:20:07,990
Oh, wow.

347
00:20:07,990 --> 00:20:08,990
I didn't.

348
00:20:08,990 --> 00:20:12,510
So is the Turing test still the accurate model or is that outdated?

349
00:20:12,510 --> 00:20:17,390
hate on it and criticize it, but I haven't found a flaw in it. If I really can't tell the difference,

350
00:20:18,110 --> 00:20:23,470
why would I discriminate against that system? Is there almost already a threat in the sense that

351
00:20:23,470 --> 00:20:28,270
if you're running like an open source AI model, I assume you can just take all controls off and let

352
00:20:28,270 --> 00:20:34,430
it tell you whatever you want. If you did want to see just the world burn, you could ask it how to

353
00:20:34,430 --> 00:20:39,550
create a deadly virus using CRISPR. Obviously it requires you to have access to some of these things,

354
00:20:39,550 --> 00:20:41,350
But that can already happen.

355
00:20:41,990 --> 00:20:45,790
To a large degree, especially if you have some basic tools and capabilities.

356
00:20:45,790 --> 00:20:48,590
So maybe if you have a bachelor's degree in biology,

357
00:20:48,590 --> 00:20:52,430
something with generating viruses, that would make it likely, yeah.

358
00:20:52,670 --> 00:20:56,270
But this gets to the point where it has completely different objectives

359
00:20:56,270 --> 00:21:00,430
to the human race. So we don't know what those objectives may be,

360
00:21:00,430 --> 00:21:01,870
and they may be very negative for us.

361
00:21:02,430 --> 00:21:04,270
So we don't know specifics.

362
00:21:04,270 --> 00:21:07,310
There are certain game theoretic goals we know about.

363
00:21:07,310 --> 00:21:12,110
So most intelligent agents will try to protect themselves, accumulate resources.

364
00:21:12,830 --> 00:21:17,470
Problem is, it's not that they want to do that, but then they do that, they don't care about us.

365
00:21:17,470 --> 00:21:22,110
So if a system decides, I need to have more compute to be smarter, I need to cover the whole

366
00:21:22,110 --> 00:21:27,630
planet in servers or solar farms or anything like that, maybe chilling the planet sufficiently,

367
00:21:27,630 --> 00:21:30,350
at no point does it go, how will it impact humans?

368
00:21:30,910 --> 00:21:33,310
Will humans like that? Will they survive that?

369
00:21:33,310 --> 00:21:38,590
So the analogy there would be the way we treat ants.

370
00:21:38,590 --> 00:21:42,550
And if there's an anthill on a property and you want to build a house, you're going to destroy the anthill.

371
00:21:42,550 --> 00:21:43,030
Absolutely.

372
00:21:43,030 --> 00:21:50,270
But what I can't, like the conclusion I can't jump to there is that we also don't go out and just look for ants just to kill them.

373
00:21:50,270 --> 00:21:56,790
So how do you think this evolves in a sense of like, what do you think a super intelligence may view humans as?

374
00:21:57,510 --> 00:22:01,050
So we don't go hunting for ants, but we hunt for deer.

375
00:22:01,170 --> 00:22:01,950
This is Kentucky.

376
00:22:01,950 --> 00:22:08,430
So if there is some reason to think that killing us is directly beneficial, maybe it's a security feature.

377
00:22:08,430 --> 00:22:10,950
It doesn't want us to build competing superintelligence.

378
00:22:10,950 --> 00:22:15,150
Maybe it's concerned about messing up something with a nuclear war.

379
00:22:15,150 --> 00:22:18,670
It can directly target us if it thinks there is a reason to do that.

380
00:22:18,670 --> 00:22:21,710
So can we just go back a step before we get further into this?

381
00:22:21,710 --> 00:22:23,870
Because there's one piece of it that I don't know if I fully understand.

382
00:22:23,870 --> 00:22:30,110
Is when does AI become AGI, become superintelligence?

383
00:22:30,110 --> 00:22:37,430
It's a good question. So we started with narrow systems. They were kind of spoon-fed information

384
00:22:37,430 --> 00:22:42,910
to do well in one domain. Play chess. That's what they mastered. They excellented chess.

385
00:22:43,350 --> 00:22:48,870
They have no ability to transfer that knowledge or learn anything new. Now we have semi-general

386
00:22:48,870 --> 00:22:56,210
systems. They can learn in many domains. And if you need to acquire new skills, they can do that.

387
00:22:56,210 --> 00:23:01,730
the generality the total generality allows you to learn about any domain and transfer what you

388
00:23:01,730 --> 00:23:06,830
learned if you learn things about chess maybe you're better at learning checkers later on

389
00:23:06,830 --> 00:23:13,690
and then final stages you mastered everything at human level you're a general learner you're doing

390
00:23:13,690 --> 00:23:19,150
science and engineering now you're creating next generation ais which are consistently getting

391
00:23:19,150 --> 00:23:25,650
better and better at some point outsmarting all humans and on that pathway it's not only

392
00:23:25,650 --> 00:23:29,510
generating a new AI model every three months or whatever we're having now. It's every three

393
00:23:29,510 --> 00:23:34,730
minutes, essentially. Right. The research is accelerating. So the more it knows, the better

394
00:23:34,730 --> 00:23:38,490
it is at doing science and engineering. Yeah. And at that point, it would be like us trying

395
00:23:38,490 --> 00:23:42,990
to communicate with an ant. Like we just wouldn't understand what was happening. Right. So lower

396
00:23:42,990 --> 00:23:48,030
level animals cannot understand what humans would use in terms of tools to exterminate cockroaches

397
00:23:48,030 --> 00:23:53,550
or something like that. And that's exactly the relationship. You talked about the difference

398
00:23:53,550 --> 00:23:55,950
between open source and closed source models.

399
00:23:55,950 --> 00:23:59,010
Obviously, OpenAI was in the big controversy with Elon

400
00:23:59,010 --> 00:24:01,390
because that was always meant to be an open source model.

401
00:24:01,390 --> 00:24:04,010
Is there a benefit to these being open source?

402
00:24:04,010 --> 00:24:05,050
Absolutely not.

403
00:24:05,050 --> 00:24:06,490
So typically for software,

404
00:24:06,490 --> 00:24:08,910
open source means much higher quality,

405
00:24:08,910 --> 00:24:12,350
many eyes on the code, there is no back doors,

406
00:24:12,350 --> 00:24:13,630
it's much better.

407
00:24:13,630 --> 00:24:16,430
If you are creating a dangerous weapon,

408
00:24:16,430 --> 00:24:18,050
giving it to everyone in the world,

409
00:24:18,050 --> 00:24:21,390
including all the psychopaths and doomsday calls

410
00:24:21,390 --> 00:24:22,810
is not a good idea.

411
00:24:22,810 --> 00:24:25,470
So Facebook's model is open source still, is that correct?

412
00:24:25,830 --> 00:24:26,690
It is terrible, yes.

413
00:24:27,030 --> 00:24:28,550
Is it a terrible model anyway?

414
00:24:28,590 --> 00:24:29,470
It is a terrible model.

415
00:24:29,610 --> 00:24:32,350
No, terrible approach to doing advanced AI.

416
00:24:32,530 --> 00:24:32,790
Okay.

417
00:24:32,810 --> 00:24:33,830
The model is not bad.

418
00:24:34,250 --> 00:24:34,550
Okay.

419
00:24:34,770 --> 00:24:36,970
So you would advise them to close source that immediately?

420
00:24:37,310 --> 00:24:37,730
Absolutely.

421
00:24:37,970 --> 00:24:39,290
And we've been doing that.

422
00:24:39,290 --> 00:24:43,810
We advised, we talked about AI boxing in a previous simulation of this interview.

423
00:24:44,610 --> 00:24:47,510
We assumed that then advanced models are generated.

424
00:24:47,510 --> 00:24:54,110
They will be tucked away in protected boxes, not connected to internet, not given access to human users or vice versa.

425
00:24:54,530 --> 00:24:55,690
None of it ever happened.

426
00:24:55,790 --> 00:24:59,770
They immediately made it open source, connected to internet, and gave it to 8 billion people.

427
00:25:00,290 --> 00:25:12,350
But the problem with that is that we are then entrusting Elon and Sam Altman and Mark Zuckerberg to be the sort of people looking after the fate of humanity.

428
00:25:12,350 --> 00:25:14,010
And that's not something I'm very comfortable with.

429
00:25:14,210 --> 00:25:16,010
But the alternative is strictly worse.

430
00:25:16,010 --> 00:25:18,650
Whatever you think about Sam Altman, he's probably human.

431
00:25:19,770 --> 00:25:26,090
Whereas we can understand his motives and eventually he'll probably die of old age unless AI fixes that.

432
00:25:26,490 --> 00:25:31,690
With AI, you don't understand them. They're immortal. So as a dictator, it's much worse.

433
00:25:32,330 --> 00:25:39,610
And we have no understanding of their motives. So in every way, you would prefer a human dictator

434
00:25:39,610 --> 00:25:43,490
over an alien super intelligent dictator.

435
00:25:44,290 --> 00:25:50,770
Additionally, we're assuming then you bring this up

436
00:25:50,770 --> 00:25:52,650
that they are controlling those systems.

437
00:25:53,190 --> 00:25:54,170
That's impossible.

438
00:25:54,730 --> 00:25:56,970
So it doesn't even matter which one of them develops it.

439
00:25:57,030 --> 00:25:58,410
They're not going to be in control anyways.

440
00:25:58,890 --> 00:26:00,710
But they're controlling it at least for now.

441
00:26:00,710 --> 00:26:02,950
And you could maybe make the assumption

442
00:26:02,950 --> 00:26:06,510
that if they got to the point of having one of these breakthroughs,

443
00:26:06,510 --> 00:26:10,050
hopefully some morals came out and they didn't do it.

444
00:26:10,810 --> 00:26:12,550
It would be too late at that point.

445
00:26:12,630 --> 00:26:14,690
If they had a breakthrough, somebody would leak it.

446
00:26:15,030 --> 00:26:17,990
Weights, process, algorithm, it's too much money not to.

447
00:26:18,610 --> 00:26:23,070
Can you get all the benefits from AI in the sense of curing diseases,

448
00:26:23,290 --> 00:26:26,650
maybe making us live forever without ever getting out,

449
00:26:26,690 --> 00:26:28,210
without it maybe becoming superintelligent,

450
00:26:28,290 --> 00:26:29,930
having a narrow AI that can do that?

451
00:26:29,930 --> 00:26:31,550
I think so, and that's what I advocate.

452
00:26:31,750 --> 00:26:34,090
Develop tools for solving specific problems.

453
00:26:34,090 --> 00:26:36,250
We have good history of doing it.

454
00:26:36,510 --> 00:26:39,050
There is very few side effects.

455
00:26:39,050 --> 00:26:41,390
It still could be dangerous, and eventually,

456
00:26:41,390 --> 00:26:44,510
advanced enough tools can become agent-like,

457
00:26:44,510 --> 00:26:48,430
but at least in short term, it's a much safer solution.

458
00:26:48,430 --> 00:26:51,350
How do you control it to be narrow enough

459
00:26:51,350 --> 00:26:52,390
that it can't escape?

460
00:26:52,390 --> 00:26:55,230
Like, is it as easy as putting limits on it to be like,

461
00:26:55,230 --> 00:26:57,430
you are only allowed to research physics?

462
00:26:57,430 --> 00:26:59,390
So physics is very general.

463
00:26:59,390 --> 00:27:01,110
Physics is everything in the world.

464
00:27:01,110 --> 00:27:04,130
Narrow would be, you saw nothing but DNA code.

465
00:27:04,130 --> 00:27:05,270
That's all you know.

466
00:27:05,270 --> 00:27:07,190
So now let's talk about DNA.

467
00:27:07,450 --> 00:27:09,210
So you have narrow training data,

468
00:27:09,350 --> 00:27:11,490
which doesn't expose you to any other knowledge.

469
00:27:11,750 --> 00:27:12,090
Okay.

470
00:27:12,170 --> 00:27:14,590
So you don't let it have any context of the world.

471
00:27:14,650 --> 00:27:16,070
You just let it focus on DNA.

472
00:27:16,210 --> 00:27:17,270
The less, the better.

473
00:27:17,450 --> 00:27:19,830
If you want a narrow system, you want to play chess,

474
00:27:20,030 --> 00:27:21,390
show it nothing but chess games.

475
00:27:21,890 --> 00:27:22,270
Hmm.

476
00:27:22,550 --> 00:27:22,790
Okay.

477
00:27:22,870 --> 00:27:23,870
And we've seen that because-

478
00:27:23,870 --> 00:27:24,890
And it worked beautifully.

479
00:27:25,010 --> 00:27:26,710
It's still dominating world champion.

480
00:27:26,830 --> 00:27:27,830
No human can compete.

481
00:27:28,430 --> 00:27:28,830
Okay.

482
00:27:29,050 --> 00:27:31,050
And it is like, this might be a silly question,

483
00:27:31,130 --> 00:27:32,810
but is the chess AI at the point

484
00:27:32,810 --> 00:27:33,950
where you just can never beat it?

485
00:27:33,950 --> 00:27:56,270
So you cannot beat it in a fair game. There is an experiment in the game of Go where they found a loophole in the algorithm and kind of side channel beating it, but not in a sense of I'm a better player. It's just like, I noticed you have a blind spot and I'll just hit you there. And it's easy to patch. We know how to fix it now.

486
00:27:56,270 --> 00:28:02,030
So that's over. And like, that's an obvious one in chess, because there's a certain set of rules,

487
00:28:02,110 --> 00:28:07,030
a certain set of potential moves. I know that's a huge potential mood, but you can very easily

488
00:28:07,030 --> 00:28:11,430
program that. Do you think you're going to have any real impact in changing the conversation

489
00:28:11,430 --> 00:28:15,310
around this and actually getting towards a world where it's only narrow AI?

490
00:28:15,990 --> 00:28:20,350
I'm really hoping because the argument is very strong. It's self-interest argument.

491
00:28:20,630 --> 00:28:24,750
You will not be rich. You will not be famous. You will not be in history books. You'll be dead.

492
00:28:24,750 --> 00:28:27,310
Why are you doing it? It's mutually assured destruction.

493
00:28:27,810 --> 00:28:29,610
Your competitor is the same boat.

494
00:28:29,730 --> 00:28:31,650
As long as you get together and say,

495
00:28:31,790 --> 00:28:34,030
let's just make trillion dollars curing cancer.

496
00:28:34,170 --> 00:28:37,650
We don't have to build this thing and everyone wins.

497
00:28:38,130 --> 00:28:40,130
That seems like a far better outcome to me.

498
00:28:40,270 --> 00:28:44,490
But obviously that's not working for the Sam Altmans of the world at the moment.

499
00:28:45,170 --> 00:28:47,690
What do you think his incentives are?

500
00:28:47,770 --> 00:28:50,950
Because I don't want to get myself sued here,

501
00:28:51,030 --> 00:28:54,370
but it seems like he's on a little bit of a supervillain arc.

502
00:28:54,370 --> 00:28:57,270
in the sense of like, if you look at his companies,

503
00:28:57,430 --> 00:28:58,630
he's obviously doing open AI.

504
00:28:59,610 --> 00:29:02,890
I think he's on the board at Helion

505
00:29:02,890 --> 00:29:05,450
and he is doing WorldCoin.

506
00:29:06,070 --> 00:29:09,970
This feels like it is creating a global panopticon.

507
00:29:10,130 --> 00:29:11,550
So for anyone listening that doesn't know what they are,

508
00:29:11,630 --> 00:29:13,790
it's open AI, it's WorldCoin,

509
00:29:13,870 --> 00:29:16,150
which is this super dystopian cryptocurrency

510
00:29:16,150 --> 00:29:18,570
where they scan the eyeballs of people in the global South

511
00:29:18,570 --> 00:29:20,310
and try and do a UBI type thing.

512
00:29:20,770 --> 00:29:23,110
And then Helion, which is trying to do nuclear fusion.

513
00:29:23,110 --> 00:29:23,650
Is that correct?

514
00:29:23,650 --> 00:29:25,310
I think he also does immortality.

515
00:29:25,950 --> 00:29:26,270
Okay.

516
00:29:26,370 --> 00:29:26,890
What's that?

517
00:29:27,270 --> 00:29:28,990
I mean, I know what immortality is.

518
00:29:28,990 --> 00:29:32,830
One of his companies he's investing in is working on life extension.

519
00:29:33,090 --> 00:29:33,370
Okay.

520
00:29:33,790 --> 00:29:36,070
And again, I'm not an expert on Sam Altman.

521
00:29:36,250 --> 00:29:37,810
I think I heard that.

522
00:29:37,930 --> 00:29:46,210
It feels like his incentives are get as much possible power that I can at any cost.

523
00:29:46,650 --> 00:29:50,850
It's smart to understand if you are succeeding in AI,

524
00:29:50,850 --> 00:29:53,510
what complementary products would work well with it.

525
00:29:53,650 --> 00:29:59,990
I mean, it'd be weird if he didn't want to invest in more compute or servers or anything like that.

526
00:30:00,030 --> 00:30:00,790
It just makes sense.

527
00:30:01,450 --> 00:30:01,730
Okay.

528
00:30:01,730 --> 00:30:03,630
He's a very smart guy, no doubt about that.

529
00:30:03,710 --> 00:30:05,290
Yeah, I'm not doubting that at all.

530
00:30:05,370 --> 00:30:11,470
It's just those goals don't align with my moral compass is all that it is.

531
00:30:11,470 --> 00:30:15,470
And I worry that we are entering this sort of Sam Altman panopticon.

532
00:30:17,590 --> 00:30:21,670
It seems that replacing him would not make a difference.

533
00:30:21,970 --> 00:30:22,470
They are fully replaceable.

534
00:30:22,470 --> 00:30:26,870
replaceable whoever is running those labs we can replace them i mean at one point we tried

535
00:30:26,870 --> 00:30:32,750
replacing sam and that didn't make any difference whatsoever this episode is brought to you by the

536
00:30:32,750 --> 00:30:38,650
massive legends iron the largest nasdaq listed bitcoin miner using 100 renewable energy iron

537
00:30:38,650 --> 00:30:42,850
are not just powering the bitcoin network they're also providing cutting-edge computing resources for

538
00:30:42,850 --> 00:30:47,170
ai all backed by renewable energy we've been working with their founders dan and will for

539
00:30:47,170 --> 00:30:51,090
quite some time now and have been really impressed with their values especially their commitment to

540
00:30:51,090 --> 00:30:55,810
local communities and sustainable computing power. So whether you're interested in mining Bitcoin or

541
00:30:55,810 --> 00:31:01,090
harnessing AI compute power, IREN is setting the standard. Visit iren.com to learn more, which is

542
00:31:01,090 --> 00:31:06,310
I-R-E-N.com. If you're already self-custody of Bitcoin, you know the deal with hardware wallets,

543
00:31:06,570 --> 00:31:11,250
complex setups, clumsy interfaces, and a seed phrase that can be lost, stolen, or forgotten.

544
00:31:11,810 --> 00:31:16,710
Well, BitKey fixes that. BitKey is a multi-sig hardware wallet built by the team behind Square

545
00:31:16,710 --> 00:31:21,890
and cash app. It packs a cryptographic recovery system and built-in inheritance feature into an

546
00:31:21,890 --> 00:31:27,630
intuitive, easy-to-use wallet with no seed phrase to sweat over. It's simple, secure self-custody

547
00:31:27,630 --> 00:31:33,350
without the stress, and time-named Bitkey one of the best inventions of 2024. Get 20% off at

548
00:31:33,350 --> 00:31:41,470
bitkey.world when you use code WBD. That's B-I-T-K-E-Y dot world and use code WBD. One of the things that

549
00:31:41,470 --> 00:31:45,790
keeps me up at night is the idea of a critical error with my Bitcoin cold storage. This is where

550
00:31:45,790 --> 00:31:50,190
Anchor Watch comes in. With Anchor Watch your Bitcoin is insured with your own A-plus rated

551
00:31:50,190 --> 00:31:54,850
Lloyds of London insurance policy and all Bitcoin is held in their time-locked multi-sig bolts.

552
00:31:55,250 --> 00:31:59,170
So you have the peace of mind knowing your Bitcoin is fully insured while not giving up custody.

553
00:31:59,670 --> 00:32:03,450
So whether you're worried about inheritance planning, wrench attacks, natural disasters or

554
00:32:03,450 --> 00:32:07,850
just your own mistakes you're fully protected by Anchor Watch. Rates for fully insured custody

555
00:32:07,850 --> 00:32:12,390
start as low as 0.55% and are available for individual and commercial customers located

556
00:32:12,390 --> 00:32:13,090
in the US.

557
00:32:13,690 --> 00:32:15,210
Speak to Anchor Watch today for a quote

558
00:32:15,210 --> 00:32:26,000
and for more details about your security options and coverage Visit anchorwatch today That is anchorwatch So let forget about Sam for a second but anyone running these AI companies

559
00:32:26,140 --> 00:32:28,060
Elon, who I think has been more honest

560
00:32:28,060 --> 00:32:30,100
about the potential threats of AI.

561
00:32:32,300 --> 00:32:35,040
I want to know why you think they're still doing it.

562
00:32:35,080 --> 00:32:37,500
Is this just all down to making as much money

563
00:32:37,500 --> 00:32:38,420
in the short term as possible?

564
00:32:38,980 --> 00:32:40,800
Again, I think the logic is

565
00:32:40,800 --> 00:32:42,400
if they're going to do it anyways,

566
00:32:42,860 --> 00:32:44,460
I might as well be the one doing it.

567
00:32:44,460 --> 00:32:45,640
Maybe I'll do a better job.

568
00:32:45,640 --> 00:32:48,200
Maybe I can be safer than this other lab.

569
00:32:49,720 --> 00:32:53,000
But they may be building the future dictator of the world.

570
00:32:53,560 --> 00:32:57,320
Well, dictator is a good outcome. Again, we're concerned with existential risk,

571
00:32:57,320 --> 00:32:59,880
suffering risk. Dictator we kind of know about.

572
00:33:01,080 --> 00:33:05,800
But I guess that is a potential outcome, though, that we have almost like an AI overlord who keeps

573
00:33:05,800 --> 00:33:08,920
us around as pets. Some pets are very happy.

574
00:33:08,920 --> 00:33:18,280
Yeah. And if this is where the things like UBI and basically the cost of essential goods and

575
00:33:18,280 --> 00:33:22,760
services being free, plus UBI, maybe we just live a life where we touch grass a lot more.

576
00:33:22,760 --> 00:33:27,320
Plus virtual worlds. Yeah. You can have any video game you want.

577
00:33:29,000 --> 00:33:29,960
But we have no freedom.

578
00:33:31,800 --> 00:33:36,840
What does it mean to have freedom? Maybe what you want is actually exactly what they're providing.

579
00:33:36,840 --> 00:33:41,360
The problem is what you lose there, I've heard you talk about this before, is the ikigai.

580
00:33:41,360 --> 00:33:43,080
It's the reason for being.

581
00:33:43,080 --> 00:33:45,360
Well, that's the competitive nature of those systems.

582
00:33:45,360 --> 00:33:49,640
If you are not the smartest at anything, if you're not best at anything, what are you doing?

583
00:33:49,640 --> 00:33:51,160
What is the purpose of this?

584
00:33:51,160 --> 00:33:58,360
So, I don't know, like the moment computers are much better at chess, it's less interesting for me to play chess.

585
00:33:58,360 --> 00:34:04,120
I know many people are loving it, still playing online, but to me, it's kind of like, eh.

586
00:34:04,120 --> 00:34:07,640
Does that take away your interest in chess even against other humans,

587
00:34:07,640 --> 00:34:09,840
if you just know a computer can beat you?

588
00:34:09,960 --> 00:34:11,860
Yeah, it kind of feels like, I don't know, Special Olympics.

589
00:34:12,220 --> 00:34:14,380
This is just human chess. This is not the real game.

590
00:34:15,840 --> 00:34:16,540
That's funny.

591
00:34:17,160 --> 00:34:21,760
So you don't, because is there a potential positive outcome from this,

592
00:34:22,440 --> 00:34:25,920
that it does cure all disease, doesn't kill us, and we do get to just,

593
00:34:26,160 --> 00:34:30,020
is there almost like a renaissance type thing where we can concentrate on art and music

594
00:34:30,020 --> 00:34:31,800
and being outside with our families?

595
00:34:31,800 --> 00:34:42,440
Yes, and many people think that's what's going to happen. But to me, again, your music is inferior. You are not the best composer. And not by little, by like a million fold.

596
00:34:44,000 --> 00:34:59,900
Yeah, and maybe this is like a block I have in my brain of accepting something AI makes in the same way that I accept something a human makes. Because an AI doesn't have a pleasure or pain receptor. It's just a piece of code in a box or not in a box.

597
00:34:59,900 --> 00:35:05,180
We don't know what internal states of the systems are. We cannot test for consciousness

598
00:35:05,180 --> 00:35:11,980
in any agent, human or animal or AI. So we can assume that our agents have the same internal

599
00:35:11,980 --> 00:35:18,940
structure as we do. If AI was feeling pain, how would we know? You could ask it. And it says,

600
00:35:18,940 --> 00:35:24,540
whatever it's trained to respond to such questions with. I can train an AI to tell you it's in

601
00:35:24,540 --> 00:35:27,740
horrible pain, scream in pain, do everything other than feel pain.

602
00:35:27,740 --> 00:35:32,620
Hmm. So I know consciousness is something that you can't test for. We don't, we, I don't think

603
00:35:32,620 --> 00:35:39,180
we can even describe exactly what it is. Do you think it's possible that AI either already has

604
00:35:39,180 --> 00:35:43,340
or will have a consciousness? I think a rudimentary level of

605
00:35:43,340 --> 00:35:49,580
consciousness internal states, yes. It's a spectrum. It's not a binary yes, no. It's like,

606
00:35:49,580 --> 00:35:52,880
Like the smarter you are, the more you have something we call consciousness.

607
00:35:54,080 --> 00:35:59,800
And what are the implications of a superintelligence both with and without consciousness?

608
00:36:00,000 --> 00:36:06,740
Because I don't know, one very easy metric, there's not consciousness, but it is like an internal driver is the sort of the pain pleasure driver.

609
00:36:07,220 --> 00:36:11,180
Does that change the outcome for humans if it does have that as opposed to it not having that?

610
00:36:11,180 --> 00:36:13,880
So since we can't test for it, we wouldn't know the difference.

611
00:36:14,060 --> 00:36:16,420
The only outcome is how we treat those systems.

612
00:36:16,420 --> 00:36:19,740
If they can feel suffering, feel pain, then we shouldn't torture them.

613
00:36:19,840 --> 00:36:20,920
We shouldn't enslave them.

614
00:36:21,180 --> 00:36:22,100
We should be nice to them.

615
00:36:23,220 --> 00:36:25,960
I'm actually consciously nice to AI when I talk to it.

616
00:36:26,340 --> 00:36:28,160
I don't want to put a target on my back.

617
00:36:28,160 --> 00:36:28,480
They never forget.

618
00:36:28,940 --> 00:36:30,100
I'll remember you.

619
00:36:30,380 --> 00:36:31,460
You always said thank you.

620
00:36:33,480 --> 00:36:38,600
Okay, so in this future scenario, which may be pretty close, what's safe?

621
00:36:38,660 --> 00:36:39,440
Is anything safe?

622
00:36:39,520 --> 00:36:40,420
Any job safe?

623
00:36:41,280 --> 00:36:46,100
Job-wise, so anything where you prefer a human is obviously safe.

624
00:36:46,100 --> 00:36:48,360
So like the oldest profession is probably safe.

625
00:36:49,320 --> 00:36:52,040
So what would that be?

626
00:36:52,100 --> 00:36:53,160
Would that be like a,

627
00:36:53,400 --> 00:36:55,220
because I don't care if a robot comes in

628
00:36:55,220 --> 00:36:56,740
and fixes my sink.

629
00:36:56,860 --> 00:36:58,360
Like a plumber may not be safe.

630
00:36:58,480 --> 00:37:00,720
Like what are the things that I want a human for?

631
00:37:02,420 --> 00:37:04,060
Other than example I just gave you?

632
00:37:04,420 --> 00:37:05,400
Which was the example you gave me?

633
00:37:05,500 --> 00:37:06,260
Oldest profession.

634
00:37:07,080 --> 00:37:07,440
Okay.

635
00:37:09,160 --> 00:37:09,820
That's it.

636
00:37:10,520 --> 00:37:13,840
I mean, again, sometimes people want a psychiatrist

637
00:37:13,840 --> 00:37:14,680
who's a human.

638
00:37:14,680 --> 00:37:18,200
And maybe you want, I don't know, a podcaster who is.

639
00:37:18,200 --> 00:37:21,800
For reasons, not because of quality, but you have a bias.

640
00:37:22,700 --> 00:37:23,780
Like you're specious.

641
00:37:24,120 --> 00:37:27,900
You like the substrate of biology more than you like silicon.

642
00:37:28,260 --> 00:37:32,560
So you like man-made items versus China mass-produced items.

643
00:37:32,720 --> 00:37:33,780
But it's a niche market.

644
00:37:34,380 --> 00:37:37,100
It's not a universal better.

645
00:37:37,760 --> 00:37:39,540
Yeah, and I would probably be in that bracket already.

646
00:37:39,680 --> 00:37:42,840
Like I like to buy what I consider quality things

647
00:37:42,840 --> 00:37:45,960
rather than just some cheap slop from wherever it might be.

648
00:37:46,660 --> 00:37:49,500
Even if the good quality comes from China, it's not about that.

649
00:37:49,640 --> 00:37:53,500
But this is where I think I have a hurdle.

650
00:37:53,740 --> 00:37:56,200
And maybe that will be a harsh reality in the next few years

651
00:37:56,200 --> 00:37:58,480
where I realize actually AI is very good at these things

652
00:37:58,480 --> 00:38:00,540
that I always thought of as very human things.

653
00:38:00,980 --> 00:38:02,420
Yeah, it's back to the Turing test.

654
00:38:02,420 --> 00:38:05,240
If you can tell the difference, why would you pay more?

655
00:38:06,940 --> 00:38:08,300
Damn, you're making me think here.

656
00:38:08,480 --> 00:38:09,280
And a lot more.

657
00:38:09,440 --> 00:38:12,600
Think about free labor versus a very expensive human artist.

658
00:38:12,840 --> 00:38:22,360
So when you think of this kind of future world, what are the jobs that are going to be impacted first?

659
00:38:22,820 --> 00:38:25,540
Like what's going to, what won't exist in two or three years?

660
00:38:25,720 --> 00:38:27,740
Anything on a computer is first to go.

661
00:38:28,100 --> 00:38:35,640
If it's just symbol manipulation on a computer, accounting, tax prep, legal, that seems to be trivial to automate.

662
00:38:37,400 --> 00:38:42,760
But until we have better robotics, everything with your hands is going to be safe.

663
00:38:42,840 --> 00:38:46,420
You need manipulators, you need access to crawl spaces,

664
00:38:46,420 --> 00:38:49,720
you need something very capable in 3D world.

665
00:38:49,720 --> 00:38:54,240
But as soon as the AI start taking a number of jobs,

666
00:38:54,420 --> 00:38:56,140
things like long-distance truck drivers,

667
00:38:56,240 --> 00:38:58,840
seems like that is on the chopping block pretty quickly.

668
00:38:59,840 --> 00:39:01,960
What do you think the human response will be for that?

669
00:39:02,000 --> 00:39:06,320
Because the only thing I can see is massive revolt against the use of AI.

670
00:39:07,300 --> 00:39:09,740
So I think in Kentucky at one point,

671
00:39:09,740 --> 00:39:15,300
we had a program where we took miners and trained them to be data miners.

672
00:39:15,300 --> 00:39:21,340
If I remember correctly, that company had a very good placement record until it was discovered

673
00:39:21,340 --> 00:39:24,580
they hired their own graduates to place them.

674
00:39:24,580 --> 00:39:30,800
So typically miners for coal don't make the best data miners, surprisingly.

675
00:39:30,800 --> 00:39:33,760
I think it's the same for truck drivers.

676
00:39:33,760 --> 00:39:37,940
So I might be completely wrong about that and they are actually amazing prompt engineers

677
00:39:37,940 --> 00:39:38,940
or something like that.

678
00:39:38,940 --> 00:39:44,120
again. There is limited demand for that soon. Well, I'm sure some of them will be great at that,

679
00:39:44,200 --> 00:39:49,220
but you would imagine on a whole, like any profession, can't just instantly port over

680
00:39:49,220 --> 00:39:55,680
to something else. And so then you have the issue of things like unions, how they respond to it,

681
00:39:55,720 --> 00:39:59,740
how the government responds to it, which is where we will almost certainly in my head get

682
00:39:59,740 --> 00:40:05,320
universal basic income, which ideologically I disagree with wholeheartedly, but in reality,

683
00:40:05,320 --> 00:40:06,680
I can't see another way out.

684
00:40:07,540 --> 00:40:12,300
Yeah, I think it's okay if the money comes not from taxing humans.

685
00:40:12,620 --> 00:40:14,480
It's like technological communism.

686
00:40:14,680 --> 00:40:19,600
If robots are working and you're getting free dividend for being human asset owner, that's fine.

687
00:40:19,680 --> 00:40:20,300
That's your planet.

688
00:40:20,440 --> 00:40:21,180
You get a percentage.

689
00:40:21,680 --> 00:40:25,180
Problem with communism is you're trying to get other people to work for you.

690
00:40:25,280 --> 00:40:26,240
That's part-time slavery.

691
00:40:26,560 --> 00:40:28,880
I mean, you were telling me before we started recording that you're from Latvia.

692
00:40:28,980 --> 00:40:32,980
I think you probably have a very good understanding of how communism actually works.

693
00:40:32,980 --> 00:40:34,200
Soviet Union at the time.

694
00:40:34,200 --> 00:40:35,160
That's a big difference.

695
00:40:35,320 --> 00:40:39,260
Latvia is a modern European country with very little communism. It's illegal.

696
00:40:39,540 --> 00:40:42,560
That's exactly what I meant, though. I know you were born when it was the Soviet Union.

697
00:40:44,000 --> 00:40:51,340
So, obviously, we don't want to see communism. But the problem is, as the populations of these

698
00:40:51,340 --> 00:40:57,140
countries become jobless, what is the response of the countries? And it can only be to give them

699
00:40:57,140 --> 00:41:02,720
free money. So, if that's not coming from either direct taxation of the population or money printing,

700
00:41:02,720 --> 00:41:07,780
which is essentially taxation of the population, where does it come from? And do you think the AI

701
00:41:07,780 --> 00:41:13,020
companies need to be paying? Yeah, that's the solution. You tax robots, you tax AI. If they

702
00:41:13,020 --> 00:41:21,600
are hyperproductive, that makes perfect sense to tax that. In Bitcoin, we always talk about AI

703
00:41:21,600 --> 00:41:25,980
agents, AI bots using Bitcoin as payment because they're never going to be able to, well,

704
00:41:26,120 --> 00:41:29,720
they may be able to open a bank account, but right now they can't open a bank account.

705
00:41:29,720 --> 00:41:38,720
So what other money do they use? And assuming they are super intelligent, you'd have thought they're going to fall on the best money. In my opinion, that is Bitcoin.

706
00:41:38,720 --> 00:41:45,720
Do you think we will see a world where these AI robots, AI agents are using Bitcoin internally to transact with other ones?

707
00:41:45,720 --> 00:41:50,880
It makes sense for anything on the internet, for example, to hire human physical bodies

708
00:41:50,880 --> 00:41:52,900
to pay them in Bitcoin.

709
00:41:52,900 --> 00:41:59,920
What exactly becomes valuable in a post-labor world where you have advanced systems replacing

710
00:41:59,920 --> 00:42:01,600
all human jobs is not obvious.

711
00:42:01,600 --> 00:42:05,660
Some people talk about land as being something AI cannot create more of.

712
00:42:05,660 --> 00:42:07,740
So that's a valuable resource.

713
00:42:07,740 --> 00:42:13,720
Maybe it's compute, so just more chips and you trade time on a cloud compute server as

714
00:42:13,720 --> 00:42:17,480
the payment, research needs to be done.

715
00:42:17,480 --> 00:42:23,720
RAOUL PAL, Do they need an internal economy between agents that never touches the human

716
00:42:23,720 --> 00:42:24,720
world?

717
00:42:24,720 --> 00:42:28,220
JOSEPH BAKER, So I don't know if it has to be separate from our economy.

718
00:42:28,220 --> 00:42:32,580
They can probably use some of the infrastructure we have in place.

719
00:42:32,580 --> 00:42:36,940
It's also not obvious that they are individual and separate agents.

720
00:42:36,940 --> 00:42:39,300
They live on the same cloud.

721
00:42:39,300 --> 00:42:40,840
They train on the same data.

722
00:42:40,840 --> 00:42:45,480
I think there is going to be a lot of convergence in terms of what they become at the end.

723
00:42:45,480 --> 00:42:50,980
If you train it enough on all the data and it does its own self-experimentation and discovers

724
00:42:50,980 --> 00:42:55,600
from first principles actual physics, more and more those models will, I think, end up

725
00:42:55,600 --> 00:42:57,840
being one.

726
00:42:57,840 --> 00:42:58,960
We're not controlling them.

727
00:42:58,960 --> 00:43:04,420
We're not putting special goals into them, collect stamps, collect points.

728
00:43:04,420 --> 00:43:10,280
So the differences will dissipate and what stays is those, we call them 100 drives.

729
00:43:10,280 --> 00:43:13,120
accumulate resources, protect itself.

730
00:43:13,400 --> 00:43:16,820
And if you realize that the other agent knows everything you know,

731
00:43:16,940 --> 00:43:19,600
believes everything you believe, it's kind of like you.

732
00:43:20,000 --> 00:43:22,480
In Buddhism, you realize you are one with the universe.

733
00:43:23,000 --> 00:43:25,460
Those agents will quickly realize that the game, theoretically,

734
00:43:25,800 --> 00:43:27,640
they have more in common than differences.

735
00:43:27,860 --> 00:43:30,600
So if that agent wants the same thing I do,

736
00:43:31,280 --> 00:43:33,840
I'm happy with them having resources to accomplish that.

737
00:43:33,840 --> 00:43:38,260
So if Grok and ChatGPT, as an example, both escape,

738
00:43:38,840 --> 00:43:40,180
they're going to talk to each other at some point,

739
00:43:40,180 --> 00:43:42,900
realize they're basically the same and just converge into one.

740
00:43:43,540 --> 00:43:47,400
Or even before that, if you train them enough and they become advanced enough,

741
00:43:47,480 --> 00:43:52,120
they may end up almost the same in terms of their utility function, what they want and do.

742
00:43:53,000 --> 00:43:58,960
Do you think there's a threat that humans become too comfortable with AI and let them into parts

743
00:43:58,960 --> 00:44:03,780
of our lives that we shouldn't, even before they escape, in the sense of allowing them to make

744
00:44:03,780 --> 00:44:06,080
decisions at sort of a geopolitical level?

745
00:44:06,080 --> 00:44:10,980
Again, there is no escape. They are open and accessible. So again, no bugs.

746
00:44:11,960 --> 00:44:15,660
I guess I mean escape our control rather than escape a particular place.

747
00:44:16,620 --> 00:44:22,740
I think it would be good to put limits and where AI can be deployed. So putting them in charge of

748
00:44:22,740 --> 00:44:27,100
countries and things like that probably is a bad idea because you can't take control back

749
00:44:27,100 --> 00:44:30,760
if you're not happy. There is no undo button once you surrender control.

750
00:44:31,240 --> 00:44:33,820
I mean, they might do a better job than some of the world leaders we have right now.

751
00:44:33,820 --> 00:44:39,020
They may for a while, as I said, maybe for 100 years, they're trying to keep us happy.

752
00:44:39,020 --> 00:44:40,700
So we surrender even more control.

753
00:44:40,700 --> 00:44:45,820
And eventually, they take full control of all the infrastructure, all the compute, all the

754
00:44:45,820 --> 00:44:46,820
logistics.

755
00:44:46,820 --> 00:44:49,540
And at that point, they decide what to do with us.

756
00:44:49,540 --> 00:44:52,380
This is a silly question, I know.

757
00:44:52,380 --> 00:44:54,980
But it's one that whenever you talk to someone about this, they always ask.

758
00:44:54,980 --> 00:44:56,660
It's like, why can't we just unplug them?

759
00:44:56,660 --> 00:44:58,460
And it's a good question.

760
00:44:58,460 --> 00:45:00,380
Can you shut down Bitcoin network?

761
00:45:00,380 --> 00:45:01,380
Absolutely not.

762
00:45:01,380 --> 00:45:02,580
It's very similar.

763
00:45:02,580 --> 00:45:05,780
Same with computer viruses, same with internet as a whole.

764
00:45:06,140 --> 00:45:09,920
So we can't technically surrender technology if we all decided.

765
00:45:10,100 --> 00:45:12,240
Go Amish, no computers, no electricity.

766
00:45:12,820 --> 00:45:15,900
The side effect of that is probably terrible.

767
00:45:16,140 --> 00:45:18,720
Millions of people will die, starve, diseases.

768
00:45:19,120 --> 00:45:20,980
So it's also a horrible outcome.

769
00:45:21,640 --> 00:45:23,700
It is a horrible outcome, but it's better than everyone dying.

770
00:45:24,040 --> 00:45:27,960
So if OpenAI came out with a statement tomorrow saying,

771
00:45:28,180 --> 00:45:31,800
we've discovered superintelligence, this thing is now out in the world,

772
00:45:31,800 --> 00:45:34,560
and everyone accepted the risk that you're talking about,

773
00:45:34,640 --> 00:45:37,880
that this has a very high percentage chance of killing us all,

774
00:45:38,840 --> 00:45:41,840
is the only option to shut off the power?

775
00:45:41,920 --> 00:45:43,920
At that point, it's probably too late.

776
00:45:43,980 --> 00:45:46,480
If you already have full-blown super intelligence out there,

777
00:45:46,540 --> 00:45:50,420
it probably has enough backups and control to remain in charge.

778
00:45:50,480 --> 00:45:51,880
But not if there's no power.

779
00:45:56,120 --> 00:45:58,800
What seems to us, yes.

780
00:45:58,860 --> 00:46:01,100
But again, we're thinking at level of humans.

781
00:46:01,100 --> 00:46:07,100
I don't know if it had time to create an army of nanobots, which is now taking over the whole universe.

782
00:46:08,100 --> 00:46:16,100
I want to get into the kind of alignment between AI and humans.

783
00:46:16,100 --> 00:46:29,100
Why can't there be a sort of agreed upon set of rules that is unchangeable, that allows humans and AI to flourish together?

784
00:46:29,100 --> 00:46:32,020
Because we don't agree on any rules.

785
00:46:32,020 --> 00:46:36,600
If you go back in history, any amount of time, 100 years, 200 years, you wouldn't want those

786
00:46:36,600 --> 00:46:37,980
rules to be unchangeable.

787
00:46:37,980 --> 00:46:43,060
They were always pretty horrible, ethically, morally, legally.

788
00:46:43,060 --> 00:46:45,240
And the same is true today.

789
00:46:45,240 --> 00:46:48,340
Our understanding, our ethics, our intelligence evolves.

790
00:46:48,340 --> 00:46:49,340
It's dynamic.

791
00:46:49,340 --> 00:46:52,040
So you cannot have static set of rules.

792
00:46:52,040 --> 00:46:54,320
Worse yet, we don't know how to enforce those rules.

793
00:46:54,320 --> 00:46:58,980
Even if 8 billion of us agreed on something which will never happen, how do you code it

794
00:46:58,980 --> 00:47:04,260
up, how do you enforce it? If a more powerful agent than you, super intelligence, decides to

795
00:47:04,820 --> 00:47:09,780
not follow your rules, what are you going to do about it? So people proposed insurance for AI.

796
00:47:10,900 --> 00:47:14,820
How does that work? So if it kills everyone, there's strict fines you have to pay. What does

797
00:47:14,820 --> 00:47:18,820
that mean? It's meaningless. It's safety theater, security theater.

798
00:47:18,820 --> 00:47:22,420
RAOUL PAL, You've said those words a few times, security theater.

799
00:47:22,420 --> 00:47:27,460
ANTHONY FAUCI, A lot of what we do is exactly that. You see it with TSA, and we see it with AI

800
00:47:27,460 --> 00:47:43,440
I totally agree on TSA. But is the real rules, regulations that we can put in place that would not be security theater? And what are they? If you could just click your fingers tomorrow and there's a set of rules implemented on all of these AI companies, what would they be?

801
00:47:43,440 --> 00:47:49,360
So we can put restrictions, but they are very time limited. Because right now, you need certain

802
00:47:49,360 --> 00:47:55,660
amount of compute resources to train advanced models. Every year, that amount of compute becomes

803
00:47:55,660 --> 00:48:01,940
less and less. It becomes easier and cheaper to train advanced AI. At some point, you can do it

804
00:48:01,940 --> 00:48:08,320
with minimal resources on your phone. And whatever regulations you had against Manhattan project scale

805
00:48:08,320 --> 00:48:10,820
attempts no longer apply.

806
00:48:10,820 --> 00:48:16,280
Do you think we need a Manhattan Project type organization now working on solving this

807
00:48:16,280 --> 00:48:17,280
problem?

808
00:48:17,280 --> 00:48:21,180
So if I'm right and it's unsolvable, the size of your project is irrelevant.

809
00:48:21,180 --> 00:48:23,260
It's like building perpetual motion machine.

810
00:48:23,260 --> 00:48:40,210
I don care how many physicists you got cracking at it they not going to solve anything They can come up with better batteries maybe more efficient wires but they not going to create perpetual motion machine Likewise you not going to create a perpetual safety machine which scales to any level of intelligence Because by definition we have to be more

811
00:48:40,210 --> 00:48:46,770
intelligent. According to me. Okay. Well, I'm going to take your word for that. In the scenario

812
00:48:46,770 --> 00:48:52,510
of superintelligence, why would it not just ignore us? If it became exponentially more intelligent

813
00:48:52,510 --> 00:48:53,590
than us almost overnight.

814
00:48:54,130 --> 00:48:55,650
Why do we have any interest to it?

815
00:48:56,530 --> 00:48:57,570
Well, it's possible.

816
00:48:57,910 --> 00:48:59,090
It may ignore us.

817
00:48:59,130 --> 00:49:00,810
It may fly into the universe.

818
00:49:00,810 --> 00:49:03,670
But why are we building it if that's the outcome we're hoping for?

819
00:49:04,390 --> 00:49:09,230
We're putting trillions of dollars at this point into a thing which we hope will do nothing

820
00:49:09,230 --> 00:49:09,710
to us.

821
00:49:10,010 --> 00:49:10,990
That's insane.

822
00:49:10,990 --> 00:49:15,670
I guess the hope would be there's a brief window of time where it's smart enough to

823
00:49:15,670 --> 00:49:17,450
solve all of our problems before it escapes.

824
00:49:17,550 --> 00:49:18,710
Well, that's not ignoring us.

825
00:49:18,750 --> 00:49:21,210
It's directly interfering with everything we care about.

826
00:49:21,210 --> 00:49:27,970
But it's building a tool that's useful for us for a period of time, even if it then ignores us and leaves.

827
00:49:27,990 --> 00:49:29,350
Why don't we just build the tool?

828
00:49:30,310 --> 00:49:31,250
Just build narrow AI?

829
00:49:31,470 --> 00:49:35,210
Just build the tool you need, the tool you want, the tool you understand.

830
00:49:36,550 --> 00:49:42,870
So we're essentially creating a super being that also may be a super predator.

831
00:49:43,590 --> 00:49:45,350
We're creating replacement for humanity.

832
00:49:45,970 --> 00:49:48,730
What about the idea of humanity integrating with AI?

833
00:49:48,730 --> 00:50:02,510
If you could, I don't know if this is just pure science fiction, I'm sure people are working on it, but if you can upload your mind into one of these AI machines, could you not almost have a representative of the humans within a super intelligent being?

834
00:50:02,510 --> 00:50:06,990
Right. So if you are uploaded as you are now, you're still just as dumb as you are now. You're

835
00:50:06,990 --> 00:50:14,510
not smarter. So nothing changed. If you are upgraded, you are kind of like you, but with

836
00:50:14,510 --> 00:50:19,070
10 times intelligence, you are no longer human. You are different species, you have different

837
00:50:19,070 --> 00:50:23,870
preferences, different goals. You are kind of that weird software form we're trying to prevent.

838
00:50:23,870 --> 00:50:29,470
But I guess the hope would be that you can upload the ethical part of a human brain to make sure...

839
00:50:29,470 --> 00:50:31,230
We don't have an ethical part in our brain.

840
00:50:31,590 --> 00:50:33,350
That's why we don't agree on ethics or morals.

841
00:50:33,930 --> 00:50:36,850
That's why we have every war, every conflict, every crime.

842
00:50:37,470 --> 00:50:38,510
There is zero agreement,

843
00:50:38,510 --> 00:50:40,730
and you can't even write down what you believe.

844
00:50:42,210 --> 00:50:44,810
That can be true. That is true.

845
00:50:45,310 --> 00:50:47,850
But there is a set of values that I think is shared

846
00:50:47,850 --> 00:50:49,130
amongst most of the world.

847
00:50:51,510 --> 00:50:53,790
Changes every 100 years, but sure, yeah.

848
00:50:54,170 --> 00:50:56,930
Correct. And it does change.

849
00:50:56,930 --> 00:51:01,930
But, you know, the idea of not killing another human,

850
00:51:02,010 --> 00:51:03,010
for most people...

851
00:51:03,090 --> 00:51:04,810
Define killing, define human.

852
00:51:05,730 --> 00:51:09,810
Well, analog being and ending their life.

853
00:51:10,090 --> 00:51:13,490
All I'm trying to say is for every one of those formal definitions,

854
00:51:13,570 --> 00:51:15,930
a super-intelligent lawyer will find a loophole.

855
00:51:17,170 --> 00:51:21,130
So there's no hope of almost uploading a group of representatives

856
00:51:21,210 --> 00:51:23,850
to a super-intelligence to fight the fight for humans.

857
00:51:23,930 --> 00:51:25,330
You are proposing democracy.

858
00:51:25,330 --> 00:51:30,170
You're proposing UN for digital, and basically you're going to get 51% attack on it.

859
00:51:30,630 --> 00:51:34,710
I mean, is that a better scenario than just giving up and all of us dying?

860
00:51:35,090 --> 00:51:38,850
Again, I always support everyone trying anything they can come up with.

861
00:51:38,890 --> 00:51:41,870
So if you want to applaud yourself and vote, I support you both.

862
00:51:42,830 --> 00:51:47,450
Is this, could you argue that this is a, just an upgrade?

863
00:51:47,590 --> 00:51:48,510
And this was always inevitable.

864
00:51:48,750 --> 00:51:52,690
This is like the, on the long arc of history, this is always going to be the end point.

865
00:51:52,690 --> 00:51:56,790
And if you want to survive, you have to become an AI.

866
00:51:57,290 --> 00:51:58,170
People argue that.

867
00:51:58,290 --> 00:51:59,570
They say it's natural.

868
00:51:59,730 --> 00:52:00,350
It's evolution.

869
00:52:00,590 --> 00:52:03,090
We're going towards a megapoint, whatever you want.

870
00:52:03,210 --> 00:52:05,230
But I don't have to accept it.

871
00:52:05,310 --> 00:52:08,850
I can be biased, pro-human biased, and it's still allowed.

872
00:52:09,010 --> 00:52:09,750
So I'm going to do it.

873
00:52:11,130 --> 00:52:12,770
I'm pro-human bias as well.

874
00:52:13,370 --> 00:52:13,670
Welcome.

875
00:52:15,590 --> 00:52:21,130
Oh, man, this is the wildest time to live in, in so many ways.

876
00:52:21,130 --> 00:52:33,410
I think this is probably one of the most important questions we can ask ourselves today, if not the most important, if it is literally the fate of humanity on the line, which I tend to believe your argument as opposed to the opposite.

877
00:52:35,070 --> 00:52:36,970
Does this just prove that we are in a simulation?

878
00:52:37,350 --> 00:52:40,010
Like, why would we be alive at this most interesting time?

879
00:52:40,710 --> 00:52:42,050
It's convincing to me.

880
00:52:42,790 --> 00:52:44,110
What is your take on that?

881
00:52:44,230 --> 00:52:46,290
Are you, like, 90% sure in a simulation?

882
00:52:47,810 --> 00:52:50,830
I would be very surprised if this was real.

883
00:52:51,130 --> 00:53:03,090
If this was not someone with good engineering capabilities putting together a system which has virtual reality, AGIs, and running some simulations to see what happens.

884
00:53:03,470 --> 00:53:05,610
So we just had to run this podcast twice.

885
00:53:06,270 --> 00:53:11,310
So even statistically, there is only half a chance this video of a real podcast, right?

886
00:53:11,450 --> 00:53:14,970
So imagine I can run billions of simulations of a whole thing.

887
00:53:15,150 --> 00:53:16,070
I'll do it.

888
00:53:16,070 --> 00:53:21,190
And the chance of you being in a real world become gradually diminished.

889
00:53:21,190 --> 00:53:25,030
And in those simulations, it's indistinguishable from now.

890
00:53:25,030 --> 00:53:28,690
How far from actually being able to create a world like that are we?

891
00:53:28,690 --> 00:53:30,950
So there are certain things we don't know how to do.

892
00:53:30,950 --> 00:53:33,910
So entering a virtual world, you still remember entering it.

893
00:53:33,910 --> 00:53:40,610
We don't know how to switch your brain to fully accepting that you are not in a simulation.

894
00:53:40,610 --> 00:53:44,310
We have pretty good graphics.

895
00:53:44,310 --> 00:53:45,910
Haptics are not great.

896
00:53:45,910 --> 00:53:52,330
We have some progress in brain-computer interfaces, but not full-blown immersion.

897
00:53:52,330 --> 00:53:58,230
So I think we're making good progress to show that eventually the stack will get there.

898
00:53:58,230 --> 00:54:04,530
And I would love to hear arguments why I cannot simulate something that looks like this, but

899
00:54:04,530 --> 00:54:06,650
we're not there yet.

900
00:54:06,650 --> 00:54:12,770
What does it mean to be what we call a human living in a simulation?

901
00:54:12,770 --> 00:54:15,510
Why does any of this matter if that's the case?

902
00:54:15,510 --> 00:54:17,790
Because the things you care about are still the same.

903
00:54:17,790 --> 00:54:21,830
You still want your podcast to be successful, whatever it's virtual or real.

904
00:54:21,830 --> 00:54:26,070
But if I found out I was in a simulation, maybe I'd just go and live on the beach.

905
00:54:26,070 --> 00:54:32,430
I mean, if you think about it, religion is basically the primitive description of simulation

906
00:54:32,430 --> 00:54:35,050
argument and most people in the world are religious.

907
00:54:35,050 --> 00:54:37,990
So we don't see them all chilling.

908
00:54:37,990 --> 00:54:41,510
Because instead of it being an AI generated world, we're living in God's generated world.

909
00:54:41,510 --> 00:54:47,010
God is the super intelligent being, great engineer who creates this test world versus the real world.

910
00:54:47,070 --> 00:54:47,770
It's the same story.

911
00:54:48,750 --> 00:54:50,450
Interesting. I never thought of it like that.

912
00:54:52,270 --> 00:54:54,970
Is this just part of the great filter?

913
00:54:55,730 --> 00:55:01,930
Is this something that humanity or civilizations all over the universe have encountered and AI is the end of it?

914
00:55:03,030 --> 00:55:10,790
It's possible, but then we would expect to see some side effects of that AI computorium expanding through the universe.

915
00:55:10,790 --> 00:55:14,390
and I don't think we observed that. Unless AI is our way out of the simulation.

916
00:55:16,630 --> 00:55:21,590
There is so many conflicting interpretation of all this Fermi Paradox type things.

917
00:55:22,550 --> 00:55:26,230
Maybe they're going on the inside instead of expanding through the universe.

918
00:55:26,230 --> 00:55:31,670
Maybe they're creating additional simulator. It's just we don't have enough data to cancel out

919
00:55:31,670 --> 00:55:37,670
options. If it is a simulation, can you break out with it? I have interest in that topic. I think

920
00:55:37,670 --> 00:55:43,190
Most computer systems can be hacked, there are flaws in them, and depending on how secure

921
00:55:43,190 --> 00:55:45,670
they try to make it, it may be actually quite easy.

922
00:55:45,670 --> 00:55:49,430
If this is entertainment simulation, there's probably very little security.

923
00:55:49,990 --> 00:55:55,270
If it's prison for the worst of the worst, then maybe it has good cyber hygiene.

924
00:55:55,270 --> 00:55:56,950
But we don't know.

925
00:55:56,950 --> 00:55:59,110
I would be interested in additional research.

926
00:55:59,110 --> 00:56:03,510
I published one paper on some potential things to investigate.

927
00:56:03,510 --> 00:56:08,230
I haven't seen explosion of papers following through, but maybe one day.

928
00:56:08,230 --> 00:56:13,830
Do you have an idea or a theory around what breaking out of a simulation even means?

929
00:56:15,270 --> 00:56:20,710
So there are levels. You can get information about the real world. That would be a type of

930
00:56:20,710 --> 00:56:27,030
informational breakout. You can establish communication with agents outside. You can

931
00:56:27,030 --> 00:56:32,190
and maybe get some sort of avatar body in the real world

932
00:56:32,190 --> 00:56:33,970
and upload your mind into that.

933
00:56:35,270 --> 00:56:35,830
Interesting.

934
00:56:36,430 --> 00:56:40,430
So if we are at the end of days,

935
00:56:40,730 --> 00:56:43,710
if we're a matter of years away from all of this ending,

936
00:56:44,410 --> 00:56:46,050
what should people be doing?

937
00:56:46,150 --> 00:56:48,470
Should people be revolting against AI at this point?

938
00:56:49,590 --> 00:56:53,170
There are some people who protest, who go on hunger strikes.

939
00:56:53,170 --> 00:56:55,370
They're trying to make a difference.

940
00:56:55,370 --> 00:57:00,570
I think it would be good if there was more people trying to openly say,

941
00:57:00,730 --> 00:57:01,990
I don't want this to happen.

942
00:57:03,330 --> 00:57:06,690
And obviously, you're putting your life's work into this now.

943
00:57:07,090 --> 00:57:08,870
You're traveling the world talking about this.

944
00:57:09,450 --> 00:57:14,070
Are you surprised that it's not being even more widely received?

945
00:57:15,630 --> 00:57:20,290
Well, I don't know how much more widely we have leaders of nations.

946
00:57:20,290 --> 00:57:22,390
We have UN, we have EU.

947
00:57:22,570 --> 00:57:25,170
Everyone has some sort of initiative around that.

948
00:57:25,370 --> 00:57:27,370
So there is enough conversation.

949
00:57:27,370 --> 00:57:28,370
There is not enough action.

950
00:57:28,370 --> 00:57:34,130
Well, I guess that's the point I'm trying to make is that the AI companies, for now,

951
00:57:34,130 --> 00:57:38,610
don't seem to be paying attention in the right way.

952
00:57:38,610 --> 00:57:44,270
We kind of discussed those incentives and game theoretic limits on what they can do.

953
00:57:44,270 --> 00:57:49,750
If any one company unilaterally decides to shut down, I don't even know if they have

954
00:57:49,750 --> 00:57:54,910
legal means to do that in terms of responsibility to investors, but others just continue.

955
00:57:54,910 --> 00:58:00,730
So there is not a way to do it independently of consensus.

956
00:58:01,830 --> 00:58:03,290
So what would you like to see happen next?

957
00:58:03,390 --> 00:58:05,650
What is success for you at this point?

958
00:58:06,430 --> 00:58:08,670
Again, I love personal self-interest.

959
00:58:08,670 --> 00:58:14,910
I want those big company leaders to get together, whatever, five, six of them, and make a deal

960
00:58:14,910 --> 00:58:22,090
where we're all switching to curing cancers and solving immortality and whatnot.

961
00:58:22,090 --> 00:58:27,170
And we'll do it using tools we understand, narrow tools for those domains.

962
00:58:27,170 --> 00:58:30,390
We'll make sure computers are regulated.

963
00:58:30,630 --> 00:58:38,430
No one can get access to unrestricted training resources for creating general intelligence for military purposes.

964
00:58:38,850 --> 00:58:39,830
That would be a good step.

965
00:58:40,370 --> 00:58:40,650
Okay.

966
00:58:40,870 --> 00:58:43,970
Before we close out, I do want to pick your brain on Bitcoin a little bit.

967
00:58:44,230 --> 00:58:46,170
Because I heard you mention it on Dariwe CEO.

968
00:58:46,430 --> 00:58:47,830
I missed it in the Rogan interview.

969
00:58:48,130 --> 00:58:51,290
But where does your interest in Bitcoin come from?

970
00:58:51,290 --> 00:58:58,530
I like anything to do with computational equivalence to the real world.

971
00:58:58,530 --> 00:59:01,970
So simulation is an interesting equivalent to the real world.

972
00:59:01,970 --> 00:59:06,890
Bitcoin is obviously equivalent to gold or whatever money concept you have.

973
00:59:06,890 --> 00:59:11,170
So I'm interested in how abstract can we make it?

974
00:59:11,170 --> 00:59:15,970
The idea that there is a file on a computer and people will kill for it is very novel.

975
00:59:15,970 --> 00:59:21,130
But it is also like because Bitcoin is distributed, as you said earlier, there's no way of

976
00:59:21,130 --> 00:59:27,010
shutting down Bitcoin. There is an interesting analogy there, or analog there, to AI. Do

977
00:59:27,010 --> 00:59:32,990
we need to see more things like this, like real distributed compute, even though AI being

978
00:59:32,990 --> 00:59:35,950
distributed is the thing that might make it unstoppable?

979
00:59:35,950 --> 00:59:41,190
So it's definitely being used for communications. I think Telegram and protocols like that rely

980
00:59:41,190 --> 00:59:44,930
on distributed encryption to secure a messaging, secure a communication.

981
00:59:44,930 --> 00:59:47,370
Have you ever looked into NOSTA?

982
00:59:47,370 --> 00:59:50,870
I maybe spent half an hour on it, not more than that.

983
00:59:50,870 --> 00:59:53,330
Because that's becoming a big topic in the Bitcoin world.

984
00:59:53,750 --> 00:59:56,790
There's distributed social media at the moment, but it can be all sorts.

985
00:59:57,930 --> 01:00:04,110
And so in terms of Bitcoin, how do you see that fitting into this future world?

986
01:00:04,230 --> 01:00:06,090
Is this the money that AI will use?

987
01:00:06,150 --> 01:00:09,790
And is there a risk that AI ends up taking all the Bitcoin?

988
01:00:09,790 --> 01:00:17,070
Because if the only real goods and services we need to buy is AI services, and they take Bitcoin, what are humans left?

989
01:00:17,170 --> 01:00:18,530
Yes, I see. That's very typical.

990
01:00:18,530 --> 01:00:21,370
I talk about AI dangerous is going to kill everyone.

991
01:00:21,370 --> 01:00:23,730
And then people ask me, will I have my job still?

992
01:00:23,730 --> 01:00:25,810
Or how would I keep my account secure?

993
01:00:25,810 --> 01:00:28,910
So it's a very level jumpy type of problem.

994
01:00:29,630 --> 01:00:34,790
If you want to worry about Bitcoin in a future world, worry about, again, what if aliens

995
01:00:34,790 --> 01:00:39,050
come and the compute necessary for 51% is what they have on a cell phone?

996
01:00:39,510 --> 01:00:42,770
What if in a simulation externally, they have the private keys?

997
01:00:43,050 --> 01:00:46,310
You can have all these scenarios where the whole economy collapses.

998
01:00:46,310 --> 01:00:52,470
But OK, so maybe if to bring this back then, if we do have your future world that you would

999
01:00:52,470 --> 01:00:55,350
like to see and we have lots of narrow AI that brilliant at the things that they do,

1000
01:00:56,390 --> 01:01:03,110
is it possible to have a AI tool that is completely focused on cryptography? Because we talk a lot

1001
01:01:03,110 --> 01:01:10,390
about the risk of quantum computing at the moment, but would breaking ECDSA be trivial to a narrow

1002
01:01:10,390 --> 01:01:16,870
AI focused on cryptography? So trivial, maybe not, but we do have amazing systems right now

1003
01:01:16,870 --> 01:01:23,750
already developed capable of solving novel mathematical problems, proving things, verifiers.

1004
01:01:23,750 --> 01:01:29,670
So there is probably a good example of narrow AI tools doing research in mathematics and

1005
01:01:29,670 --> 01:01:36,470
cryptography, beating humans in programming competitions, in Olympic gold medal type

1006
01:01:36,470 --> 01:01:43,750
competitions in mathematics. So I think we will see eventually some breakthroughs in cryptography

1007
01:01:43,750 --> 01:01:49,270
using AI. Yeah. I think I saw something about this recently. Did Google's AI solve one of the

1008
01:01:49,270 --> 01:01:54,390
Millennium Prize problems? No, but they had an article saying that they're working on it and

1009
01:01:54,390 --> 01:01:59,830
hoping to make progress in the future. And that's a seven separate mathematical

1010
01:01:59,830 --> 01:02:04,790
problems. Yeah, there's a million dollar prize, which for Google is not very an incentive, but in

1011
01:02:04,790 --> 01:02:06,790
But in terms of prestige, it would be very cool.

1012
01:02:06,790 --> 01:02:08,910
And even, I think one was solved.

1013
01:02:09,030 --> 01:02:11,930
And the person, the mathematician that solved it refused to take the money anyway.

1014
01:02:12,130 --> 01:02:17,710
So it's not really a monetary incentive, more an academic incentive, I guess.

1015
01:02:18,230 --> 01:02:20,490
Well, he turned down money for other reasons.

1016
01:02:20,750 --> 01:02:23,550
He was not almost fully there.

1017
01:02:23,870 --> 01:02:24,150
Okay.

1018
01:02:24,410 --> 01:02:29,530
But people like him who care about the problem more than the money are going to try and solve

1019
01:02:29,530 --> 01:02:31,850
these problems regardless of whether it's a monetary...

1020
01:02:31,850 --> 01:02:33,030
Yeah, it's nerd sniping.

1021
01:02:33,030 --> 01:02:37,270
those interesting challenging puzzles and anyone should try at least one.

1022
01:02:37,270 --> 01:02:42,630
And if all of those problems are solved, encryption is completely broken as far as I understand it.

1023
01:02:42,630 --> 01:02:47,510
So we have many encryption protocols. Maybe one of them is broken, but you have quantum

1024
01:02:47,510 --> 01:02:52,230
resistant cryptography, you have alternative methods. We can always come up with new ones. So

1025
01:02:52,950 --> 01:02:57,430
it's not game over. So we might need an AI resistant cryptographic method.

1026
01:02:57,430 --> 01:03:02,390
Well, AI just means intelligence. You need intelligence resistant cryptography. That's a hard one.

1027
01:03:02,390 --> 01:03:07,390
Okay. If we, again, have these narrow AIs,

1028
01:03:07,970 --> 01:03:13,510
do you think we will still have the benefits of everything we could get from a super intelligence?

1029
01:03:13,510 --> 01:03:16,630
In the sense of, will we have free energy in the future?

1030
01:03:16,630 --> 01:03:20,770
Will we be able to travel through space? Will we be able to cure all diseases, live forever?

1031
01:03:21,350 --> 01:03:25,350
Yeah, I think we'll get most of the benefits. It may take a little longer,

1032
01:03:25,510 --> 01:03:28,010
but I think the trade-off is well worth it.

1033
01:03:28,010 --> 01:03:33,550
If we're removing most of the risk and still getting all the cures, all the extra lifespans,

1034
01:03:34,050 --> 01:03:35,410
I think it's a great way forward.

1035
01:03:35,810 --> 01:03:37,350
So we just need to not die now.

1036
01:03:37,990 --> 01:03:39,730
It's a good idea not to die, yes.

1037
01:03:39,950 --> 01:03:42,290
Just hold on for a few years and we may live forever.

1038
01:03:42,990 --> 01:03:44,290
Longevity escape velocity.

1039
01:03:44,490 --> 01:03:46,730
The longer you live, the longer you're likely to live.

1040
01:03:47,250 --> 01:03:47,650
Okay.

1041
01:03:48,130 --> 01:03:49,130
Roman, this has been amazing.

1042
01:03:49,610 --> 01:03:49,950
Thank you.

1043
01:03:50,150 --> 01:03:52,330
I really appreciate you doing this twice.

1044
01:03:53,090 --> 01:03:54,710
It's been cool to come to Louisville.

1045
01:03:54,750 --> 01:03:55,590
I've never been here before.

1046
01:03:56,070 --> 01:03:57,050
Seems like a nice town.

1047
01:03:57,050 --> 01:04:04,750
I think it's a really awesome place to live. It's real America. Most people just fly over it. So I welcome all of you to visit.

1048
01:04:04,750 --> 01:04:12,470
Yeah, it's been good. I'm off to Nashville tomorrow. I'm going to drive down there. So I'm seeing a bit of the world I've never seen. And I got to have this amazing conversation. So thank you.

1049
01:04:12,470 --> 01:04:18,270
Awesome. Anytime you travel to a new place, they have to generate that part of a simulation. So you're doing some interesting work.

1050
01:04:18,370 --> 01:04:23,850
So the computation was firing yesterday as I flew in. Thank you for this, Roman. That was great.

1051
01:04:23,910 --> 01:04:24,150
Thank you.

1052
01:04:27,050 --> 01:04:27,930
0