WEBVTT

1
00:00:00.551 --> 00:00:01.332
It's weird.

2
00:00:01.352 --> 00:00:05.214
What does it mean to tell Claude to believe in itself?

3
00:00:06.035 --> 00:00:06.655
I don't know.

4
00:00:07.235 --> 00:00:27.528
I think it's genuinely weird that you could talk to a neural network, which is just crunching matrices and numbers, tell it to believe in itself, and then it just locks in like LeBron James, and it spins up a hierarchical swarm of 60 autonomous sub-agents to try and tackle this problem.

5
00:00:27.648 --> 00:00:41.574
So between them, they ran like 2,400 shell commands, hundreds of Python scripts, consumed 31 million output tokens, and they autonomously downloaded 54 academic papers from Archive.

6
00:00:41.875 --> 00:00:46.997
Perhaps Claude, like many of us, underestimates the rate of AI progress.

7
00:00:48.024 --> 00:00:48.885
Hello, Internet.

8
00:00:48.905 --> 00:00:57.776
This is your captain speaking, Lester Nare, joined as always by my co-host and our resident PhD, Krishna Chowdhury.

9
00:00:58.276 --> 00:01:00.519
We have a great episode today.

10
00:01:00.559 --> 00:01:01.660
We have two segments.

11
00:01:01.981 --> 00:01:03.623
We're going to cover the first one.

12
00:01:04.023 --> 00:01:21.033
is going to be on the progress that Claude has been making on the Riemann hypothesis in this explosion of progress as AI tackles many of these open problems and complex areas in mathematics and really understanding what's going on there.

13
00:01:21.453 --> 00:01:23.054
And we're going to end with a little bit of fun.

14
00:01:23.894 --> 00:01:31.237
segment that I forced Krishna to do today, which is the summer transfer window for American scientists.

15
00:01:31.377 --> 00:01:42.961
As we've seen an exodus and as many sports at this time of the year are doing their transfers, I thought it would be great to do it in the same format, but for science.

16
00:01:43.101 --> 00:01:49.623
As always, we are going to talk about the science from the ground up today because this is from First Principles.

17
00:02:05.478 --> 00:02:18.388
So over the last two weeks, there's been this avalanche of news all over social media, news outlets about AI has now reached a singularity in mathematics.

18
00:02:18.648 --> 00:02:28.936
Because we've seen sort of this release from OpenAI around this OpenAI Astra project where they had, you know, 10 solved open problems that they provided online.

19
00:02:29.116 --> 00:02:37.421
some workaround, and then that was quickly followed by Claude's progress on the Riemann hypothesis.

20
00:02:37.441 --> 00:02:56.432
And what we wanted to do today is kind of get a background to understand why this is so important, and then also weed through the hype versus the sort of this is nothing new reaction to that particular result.

21
00:02:56.452 --> 00:02:56.632
Yeah.

22
00:02:56.907 --> 00:02:59.509
Yeah, it's, I think, an incredible result.

23
00:03:00.090 --> 00:03:05.515
The Riemann hypothesis is the most important unsolved problem in mathematics.

24
00:03:05.875 --> 00:03:08.137
If you ask mathematicians, I think they will agree.

25
00:03:08.658 --> 00:03:15.163
If you ask, you know, people adjacent to mathematics who have heard about the millennium problems and things like that,

26
00:03:15.993 --> 00:03:17.554
This is the one, right?

27
00:03:17.835 --> 00:03:19.496
To highlight that, let me give you an example.

28
00:03:19.516 --> 00:03:33.669
There was a famous interview with Jim Simons, who's the famous mathematician behind Churn Simons theory, but I think everyone else knows him as the creator of the hedge fund Renaissance Technologies that has the medallion fund, which mysteriously just prints money.

29
00:03:34.610 --> 00:03:37.152
I think their worst year ever was 20%.

30
00:03:37.592 --> 00:03:39.914
That was their worst year in the past 30 years.

31
00:03:40.034 --> 00:03:40.875
Usually they do like 100%.

32
00:03:41.916 --> 00:03:42.877
Like 40%, 70%.

33
00:03:43.098 --> 00:03:46.282
Unbelievable.

34
00:03:46.402 --> 00:03:48.084
He was $30 billion rich.

35
00:03:48.104 --> 00:03:48.605
Yeah.

36
00:03:49.065 --> 00:03:50.988
He was worth $30 billion when he died.

37
00:03:51.569 --> 00:03:53.751
And he was asked in this interview...

38
00:03:54.879 --> 00:03:59.623
if you could trade your wealth for solving the Riemann hypothesis, would you do it?

39
00:03:59.983 --> 00:04:03.666
And you could see him just perk up and he was like, oh, that's a good question.

40
00:04:04.006 --> 00:04:09.331
And then he sort of stares off into the distance and starts fantasizing about solving the Riemann hypothesis.

41
00:04:09.571 --> 00:04:15.375
And then later on he goes back to the scripted, oh, my life has been great.

42
00:04:16.076 --> 00:04:18.858
You can't choose how your life ends up.

43
00:04:18.918 --> 00:04:19.539
No regrets.

44
00:04:19.559 --> 00:04:20.419
Hashtag no regrets.

45
00:04:20.459 --> 00:04:20.940
But you could –

46
00:04:21.480 --> 00:04:22.681
Behind it all, you could tell.

47
00:04:22.801 --> 00:04:23.762
It was a yes.

48
00:04:23.802 --> 00:04:29.746
This is a man who was like, if I could have done that, that would have been sick.

49
00:04:30.106 --> 00:04:30.367
Right?

50
00:04:30.827 --> 00:04:41.515
So given the aura around the Riemann hypothesis, it has become the gold standard around which AI's mathematical capability is judged.

51
00:04:43.023 --> 00:04:46.244
And we see this actually in our own comments section.

52
00:04:46.284 --> 00:05:00.107
Whenever we talk about AI doing math and doing crazy things, like when we covered the Jacobian conjecture not too long ago, a lot of people in the comments are just like, well, wake me up when they solve the Riemann hypothesis.

53
00:05:00.187 --> 00:05:08.709
As if this unattainable thing is going to be how we judge AI, something that humans haven't been able to do for like 300, nah, 200 years.

54
00:05:14.030 --> 00:05:15.131
It's made some progress.

55
00:05:15.451 --> 00:05:16.392
It hasn't solved it.

56
00:05:16.732 --> 00:05:17.512
And I want to be clear.

57
00:05:17.833 --> 00:05:20.154
AI has not solved the Riemann hypothesis.

58
00:05:20.554 --> 00:05:24.377
And by some measures, it is still as open as it once was.

59
00:05:24.877 --> 00:05:28.640
But the fact that it has made progress is, I think, pretty crazy.

60
00:05:29.180 --> 00:05:36.085
And particularly there's this interesting dichotomy between sort of two factions as it relates to AI progress.

61
00:05:36.725 --> 00:05:42.149
There are there is one faction that says it fundamentally was not going to make any meaningful progress.

62
00:05:42.189 --> 00:05:42.689
It matters.

63
00:05:43.330 --> 00:05:45.912
And there's the other faction that's just saying it's a matter of time.

64
00:05:46.072 --> 00:05:46.292
Mm hmm.

65
00:05:46.572 --> 00:05:48.053
the truth is usually somewhere in the middle.

66
00:05:48.213 --> 00:05:48.433
Yes.

67
00:05:48.593 --> 00:05:50.394
And this seems like it's somewhere in the middle.

68
00:05:50.754 --> 00:05:51.535
Exactly.

69
00:05:51.635 --> 00:05:54.516
And the way that it's done it, I think, is very, very cool.

70
00:05:54.877 --> 00:06:01.300
So for this episode, at least for this segment, I wanted to cover it because, one, I really like the Riemann hypothesis.

71
00:06:01.400 --> 00:06:03.942
I think it's a very cool thing to think about.

72
00:06:04.542 --> 00:06:14.848
And two, it's a really nice case study in how AI has progressed from the initial chatbots that we were thinking about, you know, back when ChatGPT came online and everybody...

73
00:06:15.577 --> 00:06:25.347
was talking about it to now there's this agentic version of AI and there's an orchestrated agentic version where AI can now command other AI, right?

74
00:06:25.748 --> 00:06:30.353
It's this really cool world that we're now living in, cool slash a little weird.

75
00:06:31.394 --> 00:06:33.156
And that is central to this story.

76
00:06:33.676 --> 00:06:42.222
I want to briefly pause by identifying this has always been an argument that's brought up, which is sort of this difference between people viewing the word.

77
00:06:42.322 --> 00:06:43.563
There's lack of definition, right?

78
00:06:43.863 --> 00:06:48.266
So AI, a lot of the perception is that's pure LLM.

79
00:06:48.306 --> 00:06:48.586
Yeah.

80
00:06:49.267 --> 00:06:55.931
And the systems that are now being used, particularly inside the frontier labs, as well as for consumers, are...

81
00:06:57.272 --> 00:07:05.179
have a variety of capabilities around the next token prediction piece that make it more than just predicting the next token.

82
00:07:05.219 --> 00:07:05.719
Exactly.

83
00:07:05.959 --> 00:07:06.340
Exactly.

84
00:07:06.360 --> 00:07:10.263
And that's going to be the highlight of this story is what they're capable of.

85
00:07:10.583 --> 00:07:15.107
So Anthropic actually put out a description of this finding on their website.

86
00:07:15.587 --> 00:07:21.772
But what's really cool is they also shared Claude's own account of how it got there.

87
00:07:22.583 --> 00:07:29.573
And they shared a full transcript of one of the agents that was instrumental in creating this proof.

88
00:07:29.993 --> 00:07:30.234
Okay?

89
00:07:30.434 --> 00:07:39.346
So we get to kind of see the inside of Claude's mind as it was progressing through this 30 hours of problem solving to get…

90
00:07:39.907 --> 00:07:41.128
to the Riemann hypothesis.

91
00:07:41.148 --> 00:07:46.273
It's that Pixar movie Inside Out where you get to see inside the thought process.

92
00:07:46.293 --> 00:07:46.914
Dude, yeah.

93
00:07:47.174 --> 00:07:52.939
And I think it's just really cool how granular we can get here.

94
00:07:53.199 --> 00:07:53.420
Okay?

95
00:07:53.720 --> 00:08:01.567
So this is, I think, crazier than the Jacobian conjecture counterexample that we covered, I think, two or three weeks ago for two reasons.

96
00:08:02.448 --> 00:08:04.590
The first reason, this is the Riemann hypothesis.

97
00:08:04.970 --> 00:08:30.996
okay the jacobian conjecture people know about it i had kind of heard about it the riemann hypothesis i have been hearing about it since even before i knew anything about complex numbers it's like millennium prize and all this other kind of stuff right um two the person who initiated this the human being behind it was not a mathematician okay for the jacobian conjecture one um

98
00:08:32.000 --> 00:08:35.426
Alpogay was a Princeton and Harvard trained mathematician.

99
00:08:35.767 --> 00:08:38.191
This time we've got Jared

100
00:08:39.996 --> 00:08:41.537
We've got his name right here.

101
00:08:41.557 --> 00:08:42.258
Jared Sumner.

102
00:08:42.598 --> 00:08:42.858
Okay.

103
00:08:43.138 --> 00:08:44.399
He's just a software engineer.

104
00:08:44.479 --> 00:08:47.281
Nothing against software engineers, but y'all aren't mathematicians.

105
00:08:47.642 --> 00:08:47.882
Okay.

106
00:08:48.142 --> 00:08:51.545
He kind of just prompted the new version of Claude that's unreleased.

107
00:08:52.145 --> 00:08:54.267
Anthropic is very clear to say it's unreleased.

108
00:08:54.607 --> 00:08:58.790
You know, they want their shareholders to be very, very confident of their progress.

109
00:08:59.130 --> 00:09:02.913
So it's an unreleased version of Claude that he prompted Claude.

110
00:09:03.093 --> 00:09:06.498
With just, hey, why don't you take a stab at the Riemann hypothesis?

111
00:09:07.299 --> 00:09:09.482
And then when Claude was like, I'm not going to do that.

112
00:09:09.502 --> 00:09:12.446
It's a 150-year-old problem that no one solved.

113
00:09:12.546 --> 00:09:15.130
He came back with things like, believe in yourself.

114
00:09:16.564 --> 00:09:20.066
A little bit of positive encouragement.

115
00:09:20.226 --> 00:09:20.886
To Claude.

116
00:09:21.006 --> 00:09:21.747
To an AI.

117
00:09:21.787 --> 00:09:22.727
To an AI.

118
00:09:23.507 --> 00:09:25.748
Actually, this is, okay.

119
00:09:25.909 --> 00:09:27.669
And it's kind of weird that that kind of worked.

120
00:09:27.769 --> 00:09:27.990
Yeah.

121
00:09:28.010 --> 00:09:29.870
So we're going to get into some of that as well.

122
00:09:29.990 --> 00:09:30.211
Okay.

123
00:09:30.231 --> 00:09:30.451
Okay.

124
00:09:30.931 --> 00:09:33.732
So first, let's go over the Riemann hypothesis.

125
00:09:33.852 --> 00:09:36.073
What is it and why are people obsessed about it?

126
00:09:36.153 --> 00:09:36.354
Right.

127
00:09:36.494 --> 00:09:36.734
Okay.

128
00:09:37.254 --> 00:09:40.956
I could dedicate several two-hour long videos to this thing.

129
00:09:41.996 --> 00:09:44.998
And truth be told, to really understand it, you're going to have to go through

130
00:09:45.818 --> 00:09:51.862
A book like Complex Analysis, again, my favorite book that everyone seems to also love as well.

131
00:09:52.343 --> 00:09:54.004
Complex Analysis by Elias Stein.

132
00:09:54.024 --> 00:09:59.868
If you go through this book, you've got to get to chapter 7 to really understand Riemann hypothesis.

133
00:09:59.928 --> 00:10:04.591
I think you could skip the chapter on Fourier analysis because that doesn't really tie into this stuff.

134
00:10:04.611 --> 00:10:10.695
But everything else you're going to really need to understand in order to believe some of the magic that I'm about to tell you.

135
00:10:11.075 --> 00:10:13.617
But I'm going to go through kind of a Sparknotes version of...

136
00:10:13.997 --> 00:10:19.463
of how you would begin to understand why the Riemann hypothesis is so important to solve.

137
00:10:19.844 --> 00:10:20.104
Okay?

138
00:10:20.584 --> 00:10:22.907
I'm going to begin with the basal problem.

139
00:10:23.768 --> 00:10:27.351
This was first proposed by Pietro Mengoli in the 1600s.

140
00:10:29.173 --> 00:10:30.315
The problem is pretty simple.

141
00:10:30.835 --> 00:10:31.135
Okay?

142
00:10:31.596 --> 00:10:33.318
You take an infinite series...

143
00:10:34.134 --> 00:10:37.477
of reciprocals of squares, and you add it up.

144
00:10:37.878 --> 00:10:43.764
So you get 1 over 1 plus 1 over 2 squared plus 1 over 3 squared plus 1 over 4 squared all the way up to infinity.

145
00:10:44.344 --> 00:10:46.006
This is an infinite series.

146
00:10:47.107 --> 00:10:50.590
So 1 plus 1 over 4 plus 1 over 9 plus 1 over 16 all the way.

147
00:10:50.911 --> 00:10:51.672
What does that equal?

148
00:10:52.212 --> 00:10:54.093
This was proposed in the 1600s.

149
00:10:54.193 --> 00:10:59.976
It's named the Basel problem because the people who were really working on it were Leonard Euler and the Bernoulli family.

150
00:11:00.276 --> 00:11:10.981
The Bernoulli family is the very famous family that is behind the Bernoulli numbers, the Bernoulli principle, which is how a lot of people mistakenly say airplanes fly.

151
00:11:11.321 --> 00:11:12.402
It's not that simple.

152
00:11:12.482 --> 00:11:14.683
I mean, it has something to do with it, but it's not that simple.

153
00:11:15.443 --> 00:11:17.784
In any case, the Bernoulli family tried and they failed.

154
00:11:18.044 --> 00:11:19.965
This is a family of mathematicians.

155
00:11:20.145 --> 00:11:21.126
Leonard Euler...

156
00:11:21.806 --> 00:11:40.909
tried honestly might have been his first time that he just like solved it okay and he said this whole thing is equal to pi squared over six okay so one plus one over four plus one over nine plus one over 16 all the way down that's going to equal pi squared over six the way he did it was kind of hacky um and

157
00:11:41.706 --> 00:11:46.789
Like, it was only proven to be completely correct by Weierstrauss like 100 years later.

158
00:11:46.809 --> 00:11:47.749
It was a little hand-wavy.

159
00:11:47.809 --> 00:11:49.810
It was a little hand-wavy, but he got the right answer.

160
00:11:50.090 --> 00:11:51.331
Like, so many things that he did.

161
00:11:51.731 --> 00:11:52.011
Okay?

162
00:11:52.551 --> 00:11:54.392
Because his intuition was just like insane.

163
00:11:54.612 --> 00:11:54.813
Right?

164
00:11:55.293 --> 00:11:55.753
Okay, fine.

165
00:11:55.793 --> 00:11:56.813
So that's impressive, right?

166
00:11:57.234 --> 00:12:01.236
But that's not his main thing that Euler did for...

167
00:12:02.051 --> 00:12:02.872
The basal problem.

168
00:12:03.192 --> 00:12:08.718
So the basal problem is only 1 plus 1 over 4 plus 1 over 9 all the way over, right?

169
00:12:09.779 --> 00:12:16.446
Euler, in his text, Various Observations of Infinite Series, he generalizes to something called the zeta function.

170
00:12:16.987 --> 00:12:17.247
Okay?

171
00:12:17.747 --> 00:12:24.314
He actually invented the zeta function, which is, what if I consider not just 1 over...

172
00:12:25.475 --> 00:12:30.961
squares but 1 over cubes or 1 over stuff to the fourth power, things like that.

173
00:12:31.622 --> 00:12:43.435
So he creates this function called the zeta function and crucially he shows that that sum of 1 over all of the squares and things like that of the natural numbers is equal to a product

174
00:12:44.487 --> 00:12:45.348
Over the primes.

175
00:12:46.088 --> 00:12:46.729
Oh, interesting.

176
00:12:46.749 --> 00:12:46.969
Okay.

177
00:12:47.490 --> 00:12:55.216
So we took something that was discreet as an initial problem set, generalized it, and came up with it and had a rule that related to that generalization.

178
00:12:55.336 --> 00:12:55.797
Very good.

179
00:12:55.817 --> 00:12:56.057
Yeah.

180
00:12:56.077 --> 00:13:02.102
The generalization was I take one over the cubes or whatever power that I want, and the

181
00:13:03.343 --> 00:13:09.345
Little trick that he did was show that the sum could be turned into an infinite product over primes.

182
00:13:09.646 --> 00:13:12.287
We're going to get into exactly how that works.

183
00:13:12.607 --> 00:13:18.969
Because I think this is a really neat way of seeing why the zeta function is so important in mathematics.

184
00:13:19.470 --> 00:13:20.630
It has to do with primes.

185
00:13:21.170 --> 00:13:26.132
And everywhere you look, you'll see that the Riemann hypothesis has to do with the distribution of primes.

186
00:13:26.512 --> 00:13:30.474
Understanding just this will give you kind of a sense of why that is the case.

187
00:13:30.574 --> 00:13:32.795
Why are primes related to this infinite product?

188
00:13:34.415 --> 00:13:37.136
So that's what we're going to do right now.

189
00:13:37.596 --> 00:13:38.516
Here's where we're going to start.

190
00:13:38.536 --> 00:13:39.876
We're going to start with the zeta function.

191
00:13:39.896 --> 00:13:43.137
Let's consider instead of squares.

192
00:13:43.357 --> 00:13:46.797
So in this case, the zeta function is for any arbitrary S. Yes.

193
00:13:46.998 --> 00:13:48.378
But the S is usually two.

194
00:13:48.418 --> 00:13:53.419
That'll give you the basal problem of one plus one over fourth, one plus one ninth plus one sixteenth.

195
00:13:53.739 --> 00:13:55.699
Instead, let's just consider S equals one.

196
00:13:56.179 --> 00:13:56.419
Okay.

197
00:13:56.679 --> 00:13:58.119
This is called a harmonic series.

198
00:13:58.240 --> 00:14:00.660
It's just one plus all of the fractions added up.

199
00:14:00.740 --> 00:14:00.920
Yep.

200
00:14:01.040 --> 00:14:02.881
This thing crucially does not converge.

201
00:14:03.242 --> 00:14:03.502
Okay?

202
00:14:03.542 --> 00:14:04.643
This thing goes to infinity.

203
00:14:04.723 --> 00:14:07.445
So there's the harmonic series, which is when the power is one.

204
00:14:07.485 --> 00:14:09.607
There's the basal problem when the power is two.

205
00:14:09.687 --> 00:14:12.829
And then the zeta is for any power greater than one.

206
00:14:13.129 --> 00:14:13.369
Okay?

207
00:14:13.510 --> 00:14:14.951
That's the zeta series.

208
00:14:15.651 --> 00:14:15.871
Okay?

209
00:14:16.372 --> 00:14:19.794
So we're going to consider the harmonic series, which is when the power is one.

210
00:14:19.814 --> 00:14:21.115
So we're just getting reciprocals.

211
00:14:21.756 --> 00:14:21.996
Okay?

212
00:14:22.316 --> 00:14:24.038
We're going to try to understand this a little bit.

213
00:14:24.837 --> 00:14:31.182
Now, we're going to need a second tool in our toolbox, and that has to do with the equation for a geometric series.

214
00:14:31.582 --> 00:14:33.723
A geometric series is a power series.

215
00:14:34.444 --> 00:14:43.570
It's something like, suppose you were to add 1 plus 1 half, but not instead of 1 third, we say plus 1 fourth, plus 1 eighth.

216
00:14:43.770 --> 00:14:46.552
So powers of 2, the same power over and over again.

217
00:14:47.333 --> 00:14:48.234
This is not now...

218
00:14:48.974 --> 00:14:50.294
You're not changing the number.

219
00:14:50.514 --> 00:14:51.695
Yes.

220
00:14:51.835 --> 00:14:55.335
You're keeping the number the same, but the power is going up.

221
00:14:55.735 --> 00:15:00.396
The variable moves from being in the power spot to being in the exponent.

222
00:15:00.737 --> 00:15:02.977
In the exponent spot into the power.

223
00:15:03.037 --> 00:15:04.157
Yeah, exactly.

224
00:15:04.277 --> 00:15:07.018
So now this is a geometric series is what it's called.

225
00:15:07.058 --> 00:15:10.739
Because there's a common ratio between each of the terms, right?

226
00:15:10.959 --> 00:15:12.359
You're halving each term.

227
00:15:12.779 --> 00:15:13.660
and then you're adding it up.

228
00:15:13.680 --> 00:15:16.042
So you get a one, that's a whole pizza, let's say.

229
00:15:16.382 --> 00:15:22.968
Then a half, that's half a pizza, plus a quarter of a pizza, plus an eighth, plus a 16th, plus a 32th, or whatever it's called.

230
00:15:23.008 --> 00:15:23.608
32th.

231
00:15:24.589 --> 00:15:26.911
As you add it all up, you're going to get two.

232
00:15:26.991 --> 00:15:27.632
You can see, right?

233
00:15:27.652 --> 00:15:28.973
The one is the first.

234
00:15:29.013 --> 00:15:35.598
And then as you add a half plus a fourth plus an eighth, and so on and so forth, you're going to get a whole number, and the whole thing is going to equal two.

235
00:15:35.898 --> 00:15:41.303
The way, and this is something that you learn in high school calculus, is how to sum up a geometric series

236
00:15:41.595 --> 00:15:43.042
You take the common fraction...

237
00:15:43.738 --> 00:15:49.441
which is a half, sorry, the common ratio, which is a half, because you're multiplying a half every time you add a new thing.

238
00:15:50.641 --> 00:15:52.222
And you take the first number.

239
00:15:52.322 --> 00:15:53.763
So the first number in this case is 1.

240
00:15:54.023 --> 00:15:54.924
You put that on the top.

241
00:15:55.464 --> 00:15:57.385
1 divided by 1 minus the common ratio.

242
00:15:57.425 --> 00:16:00.947
So in this case, it would be 1 divided by 1 minus 1.5 to get you 2.

243
00:16:01.387 --> 00:16:04.268
And that's how you sum up a geometric series.

244
00:16:04.508 --> 00:16:04.708
Okay?

245
00:16:04.949 --> 00:16:06.549
So we're going to require this toolbox.

246
00:16:06.970 --> 00:16:07.170
All right?

247
00:16:07.530 --> 00:16:10.151
Now that we've understood that, this was Euler's genius.

248
00:16:10.511 --> 00:16:11.572
He said, consider...

249
00:16:12.760 --> 00:16:15.220
The geometric series for primes.

250
00:16:16.181 --> 00:16:16.421
Okay?

251
00:16:16.681 --> 00:16:18.781
We had just seen a geometric series for two.

252
00:16:19.861 --> 00:16:25.862
A geometric series for three would look like one plus one-third plus one-ninth plus one-twenty-seventh because those are the powers of three.

253
00:16:27.103 --> 00:16:30.223
We don't do four because four is not a prime.

254
00:16:31.383 --> 00:16:39.485
We look at the geometric series for five, which is one plus one-fifth plus one-twenty-fifth plus one-twenty-fifth and then there'll be six-twenty-five, so on and so forth.

255
00:16:40.065 --> 00:16:44.648
All of those things are going to equal 1 divided by 1 minus that common ratio.

256
00:16:45.069 --> 00:16:46.950
So it's going to be 1 over 1 minus 1 over 2.

257
00:16:47.811 --> 00:16:50.012
Over there, that's going to be 1 over 1 minus 1 third.

258
00:16:50.312 --> 00:16:53.575
So that'll give you 2 thirds, actually, over there.

259
00:16:53.695 --> 00:16:55.936
It's 3 halves, sorry.

260
00:16:56.276 --> 00:17:00.599
And so this is the geometric series for...

261
00:17:01.650 --> 00:17:03.712
powers of primes.

262
00:17:04.433 --> 00:17:04.673
Okay?

263
00:17:05.534 --> 00:17:08.236
Euler said, let's take a look at this and see if we can do something.

264
00:17:08.496 --> 00:17:08.756
Alright?

265
00:17:09.257 --> 00:17:21.468
If we multiply all of these series together, so we multiply the infinite sum up there, the infinite sum up there, the infinite sum up there, I'm going to really get

266
00:17:23.103 --> 00:17:26.464
a multiplication of those individual representations.

267
00:17:26.724 --> 00:17:26.924
Right?

268
00:17:27.184 --> 00:17:28.305
Those individual answers.

269
00:17:28.545 --> 00:17:33.286
And now you're already starting to see that second-hand side of Euler's product formula.

270
00:17:33.306 --> 00:17:34.967
Yeah, I was going to say that it's the same.

271
00:17:35.147 --> 00:17:36.767
Yeah, there's the through line.

272
00:17:36.827 --> 00:17:37.768
Yeah, there's the through line.

273
00:17:38.088 --> 00:17:40.509
The giant pi, that means multiply.

274
00:17:40.789 --> 00:17:42.689
The giant sigma means sum.

275
00:17:43.049 --> 00:17:44.450
The giant pi means multiply.

276
00:17:44.670 --> 00:17:46.951
And it's multiplying over the primes.

277
00:17:47.611 --> 00:17:47.851
Okay?

278
00:17:48.071 --> 00:17:48.951
For whatever power.

279
00:17:49.031 --> 00:17:50.672
In this case, the power is s equals 1.

280
00:17:50.772 --> 00:17:51.012
Mm-hmm.

281
00:17:51.252 --> 00:17:51.432
right?

282
00:17:52.012 --> 00:17:54.354
And so now we're seeing the right-hand side.

283
00:17:54.594 --> 00:17:54.794
Yes.

284
00:17:55.434 --> 00:17:56.675
Now, that's fine.

285
00:17:57.316 --> 00:17:58.796
Here's the real genius of Euler.

286
00:17:58.856 --> 00:18:00.497
Why is he even doing this in the first place?

287
00:18:00.718 --> 00:18:05.060
Okay, this seems like a lot of stuff to do without the punchline.

288
00:18:05.500 --> 00:18:06.561
Here is the punchline.

289
00:18:06.581 --> 00:18:07.842
What about the left-hand side?

290
00:18:09.222 --> 00:18:10.003
The left-hand side

291
00:18:11.428 --> 00:18:14.409
As I said, it's a bunch of primes multiplied together, right?

292
00:18:14.430 --> 00:18:20.132
You've got 1 plus 1 half plus 1 fourth, the powers of 1 half, the powers of 1 third, the powers of 1 fifth.

293
00:18:20.172 --> 00:18:22.373
You'll get the power of 1 seventh later, right?

294
00:18:23.834 --> 00:18:28.817
If I multiply those out and I do the FOIL method, if you remember, like you have to distribute.

295
00:18:28.897 --> 00:18:31.558
Each term has to multiply to each term, right?

296
00:18:32.241 --> 00:18:38.288
So if I multiply that out, the 1 can be multiplied to all the other ones to get a 1, right?

297
00:18:38.809 --> 00:18:44.856
Then the 1 half from the first one can be multiplied to all the other ones to get a 1 half, right?

298
00:18:46.538 --> 00:18:49.781
Similarly, the 1 fourth can be multiplied to all the other ones to get a 1 fourth.

299
00:18:49.821 --> 00:18:50.022
Mm-hmm.

300
00:18:50.912 --> 00:18:51.933
The one-third will be there.

301
00:18:51.953 --> 00:18:52.854
The one-fifth will be there.

302
00:18:52.954 --> 00:18:59.380
I could also get a one-sixth because the one-half times the one-third multiplied by all the other ones is going to give me the one-sixth.

303
00:18:59.500 --> 00:18:59.900
I see.

304
00:18:59.940 --> 00:19:02.082
The one-eighth is going to come from the one-half, right?

305
00:19:02.362 --> 00:19:04.064
The one-ninth is going to come from the one-third.

306
00:19:04.124 --> 00:19:07.367
And then the one-twelfth can come from one-fourth multiplied by one-third.

307
00:19:08.367 --> 00:19:12.231
Every single number is going to be represented here.

308
00:19:12.251 --> 00:19:12.371
Right?

309
00:19:13.453 --> 00:19:18.720
in that infinite sum, because primes make up the numbers.

310
00:19:19.061 --> 00:19:19.301
Right.

311
00:19:19.381 --> 00:19:19.561
Right?

312
00:19:19.641 --> 00:19:27.912
Every single number that is out there, this is called the fundamental theorem of arithmetic, which means that every single number out there, a natural number, can be broken down

313
00:19:28.613 --> 00:19:33.737
into multiplications of primes, products of primes that are unique.

314
00:19:33.877 --> 00:19:35.018
There's only one way to do it.

315
00:19:35.739 --> 00:19:41.023
And there's only one way to choose all of the different terms in that sequence.

316
00:19:41.103 --> 00:19:42.224
And now you can notice, right?

317
00:19:42.565 --> 00:19:43.806
What are the ones that are missing?

318
00:19:44.626 --> 00:19:48.029
One seventh is missing because I haven't included that in this, right?

319
00:19:48.429 --> 00:19:51.072
Similarly, one eleventh is missing, but that's a prime.

320
00:19:51.332 --> 00:19:52.673
That'll be another one on its own.

321
00:19:52.693 --> 00:19:52.813
Okay.

322
00:19:55.295 --> 00:19:58.736
1 13th is missing, but also 1 14th is missing.

323
00:19:59.016 --> 00:20:03.837
Because 1 14th would be the 1 7th multiplied by the 1 half multiplied by all the other ones.

324
00:20:04.297 --> 00:20:04.478
Right?

325
00:20:04.738 --> 00:20:07.218
So all of the other composite numbers are not there.

326
00:20:07.638 --> 00:20:09.699
And the primes that I haven't included aren't there.

327
00:20:09.719 --> 00:20:10.619
Because they're just further down.

328
00:20:10.639 --> 00:20:11.599
Yeah, they're just further down.

329
00:20:11.659 --> 00:20:13.420
I haven't included it, right?

330
00:20:13.740 --> 00:20:17.241
If I added the 1 7th thing, then the 1 14th would show up and the 1 7th would show up.

331
00:20:17.501 --> 00:20:19.021
But the 1 11th would not show up.

332
00:20:19.402 --> 00:20:19.582
Right?

333
00:20:19.602 --> 00:20:20.742
Because that would require another 1.

334
00:20:21.222 --> 00:20:26.584
As we expand the top row we have here, it will fill in these gaps that are currently present.

335
00:20:26.664 --> 00:20:28.005
And you could do this infinitely.

336
00:20:28.045 --> 00:20:29.285
Yes, and you can do this infinitely.

337
00:20:29.305 --> 00:20:31.846
And the key is you will get every single fraction ever.

338
00:20:31.906 --> 00:20:32.706
That's incredible.

339
00:20:32.726 --> 00:20:33.026
Okay.

340
00:20:33.367 --> 00:20:33.587
Right?

341
00:20:33.667 --> 00:20:34.267
Yeah, yeah, yeah.

342
00:20:34.887 --> 00:20:36.408
This was Euler's genius.

343
00:20:36.708 --> 00:20:39.749
And that's how he's showing this is called Euler's product formula.

344
00:20:39.809 --> 00:20:43.730
And it's kind of crazy that he set out to just solve the basal problem.

345
00:20:45.011 --> 00:20:47.432
And then he's like, oh, by the way, I also noticed this.

346
00:20:49.880 --> 00:20:50.881
Just have this one on the frame.

347
00:20:50.921 --> 00:20:51.421
Yeah, yeah.

348
00:20:51.441 --> 00:20:55.264
So here, you've already seen now, the zeta function is on the left.

349
00:20:55.444 --> 00:20:59.206
That's the sum of the one over all the natural numbers, right?

350
00:21:00.087 --> 00:21:02.729
He has tied this now to a product over primes.

351
00:21:03.429 --> 00:21:11.695
And now you can imagine if I want to do one, the zeta function to the second power, all I have to do is put the primes to the second power.

352
00:21:11.715 --> 00:21:13.216
Right, right.

353
00:21:13.536 --> 00:21:14.317
Yes, yes.

354
00:21:14.377 --> 00:21:17.739
And that's why the S is a common exponent.

355
00:21:17.779 --> 00:21:19.340
Exponent across both.

356
00:21:20.892 --> 00:21:21.813
Oh, that's quite nice.

357
00:21:21.833 --> 00:21:23.215
That's so elegant.

358
00:21:23.255 --> 00:21:23.575
Right?

359
00:21:23.675 --> 00:21:29.101
And this already shows you how the primes are so intertwined with the zeta function.

360
00:21:29.121 --> 00:21:29.562
Yeah.

361
00:21:29.602 --> 00:21:29.762
Right?

362
00:21:29.782 --> 00:21:30.763
They are one and the same.

363
00:21:30.963 --> 00:21:32.485
And so part of what we're building here...

364
00:21:33.035 --> 00:21:34.736
is the connection between those two.

365
00:21:34.796 --> 00:21:35.036
Yes.

366
00:21:35.497 --> 00:21:40.780
I'm trying to justify why there's so much hype about this.

367
00:21:41.320 --> 00:21:46.283
It's because the Riemann zeta function has to do with primes in a very integral way.

368
00:21:46.363 --> 00:21:48.264
It's actually encoding the same information.

369
00:21:49.085 --> 00:21:52.007
On one side, there's a product of primes.

370
00:21:52.507 --> 00:21:55.009
On the other side, there's a sum over every natural number.

371
00:21:55.909 --> 00:21:56.690
This is really good.

372
00:21:56.970 --> 00:21:57.730
It's kind of cool.

373
00:21:57.770 --> 00:21:58.431
This is really good.

374
00:21:58.451 --> 00:21:59.491
Yeah, and this was in the 1730s.

375
00:22:02.228 --> 00:22:03.069
No Wi-Fi then.

376
00:22:03.630 --> 00:22:03.810
No.

377
00:22:05.432 --> 00:22:10.719
Now, 1792, there's a teenage Carl Friedrich Gauss.

378
00:22:11.059 --> 00:22:11.500
Gauss.

379
00:22:11.900 --> 00:22:16.706
Gauss is another amazing mathematician, perhaps second or third only to Euler.

380
00:22:17.747 --> 00:22:18.888
And maybe Riemann, actually.

381
00:22:20.850 --> 00:22:23.231
He starts counting primes when he was a teenager.

382
00:22:23.572 --> 00:22:29.056
Because I guess that's what you do when you're a full autist in Germany.

383
00:22:29.176 --> 00:22:29.696
In 1792.

384
00:22:29.816 --> 00:22:30.337
Yeah.

385
00:22:30.357 --> 00:22:33.920
So he starts counting the number of primes by hand.

386
00:22:35.031 --> 00:22:42.838
And he notices that the probability that a certain number is going to be prime is related to the log of that number.

387
00:22:42.958 --> 00:22:44.440
It's basically 1 over the log.

388
00:22:44.480 --> 00:22:48.463
Like if I go up to like a 10,000, right?

389
00:22:49.064 --> 00:22:52.507
The probability that numbers around 10,000 obviously is not prime.

390
00:22:52.867 --> 00:22:57.831
But like any given number around there is prime is 1 divided by the log of 10,000, log base e. Okay.

391
00:22:58.532 --> 00:22:58.712
Okay.

392
00:22:58.732 --> 00:22:58.972
Okay.

393
00:22:59.272 --> 00:23:09.639
So from there, he realizes that there's this prime counting function where I count how many primes before a certain number, and that's related to n divided by log n. Okay.

394
00:23:10.039 --> 00:23:14.141
There in the blue, you see a pi of n. That's the prime counting function.

395
00:23:14.181 --> 00:23:14.341
Right.

396
00:23:14.521 --> 00:23:18.224
n over log n is in red, and you can see that n over log n kind of undershoots.

397
00:23:18.704 --> 00:23:25.248
He also realizes that the logarithmic integral, meaning the integral of the log, which is li of n,

398
00:23:25.708 --> 00:23:27.909
That's kind of an overshoot most of the time.

399
00:23:28.089 --> 00:23:29.350
Turns out it's not all the time.

400
00:23:29.690 --> 00:23:35.252
If you go to 10 to the 10 to the 35, then prime numbers go above the log integral.

401
00:23:35.372 --> 00:23:39.514
But he actually thought that perhaps the integral function is always above.

402
00:23:40.214 --> 00:23:46.737
But he didn't know that 10 to the 10 to the 35... Couldn't get that high.

403
00:23:46.797 --> 00:23:51.499
If you do it by hand and you see a pretty good... You're like, yeah, it's probably...

404
00:23:51.919 --> 00:23:52.319
I get it.

405
00:23:52.940 --> 00:23:53.180
So...

406
00:23:54.064 --> 00:24:09.322
This becomes now the prime number theorem, which is that the number of primes before a certain level, the number of primes before a certain number is related to n divided by log n. Okay?

407
00:24:09.382 --> 00:24:10.223
Which is our red here.

408
00:24:10.922 --> 00:24:11.442
Yeah, yeah.

409
00:24:11.542 --> 00:24:14.304
It's like asymptotic to N divided by, like it has the same shape.

410
00:24:14.404 --> 00:24:14.604
Yes.

411
00:24:14.764 --> 00:24:16.785
Okay, it's not like going to diverge like crazy.

412
00:24:16.865 --> 00:24:17.085
Right.

413
00:24:17.145 --> 00:24:19.246
It's always going to stay close in some sense.

414
00:24:19.607 --> 00:24:19.787
Right?

415
00:24:20.167 --> 00:24:22.768
Now, enter Bernard Riemann.

416
00:24:23.188 --> 00:24:23.869
Riemann.

417
00:24:23.889 --> 00:24:24.109
Okay.

418
00:24:24.469 --> 00:24:25.009
Riemann.

419
00:24:25.149 --> 00:24:26.730
Riemann at the University of Berlin.

420
00:24:26.850 --> 00:24:31.493
He's got an 1859 article on the number of primes less than a given magnitude.

421
00:24:32.800 --> 00:24:40.463
This is the only paper that he ever wrote in analytic number theory, and it is probably his most influential.

422
00:24:40.783 --> 00:24:46.185
He was mostly worried about geometry and things like that, like Riemannian manifolds and stuff.

423
00:24:46.205 --> 00:24:48.926
That's the stuff that you get into with general relativity.

424
00:24:49.186 --> 00:24:51.948
But this is what he wrote about the prime number theorem.

425
00:24:52.668 --> 00:24:58.010
And ever since then, it has been named the Riemann zeta function.

426
00:24:58.550 --> 00:25:01.893
Okay, because there's a joke in mathematics.

427
00:25:02.393 --> 00:25:07.998
You always name something after the second person who discovered it because the first person is always Euler.

428
00:25:08.459 --> 00:25:10.761
And you can't name everything after Euler.

429
00:25:11.872 --> 00:25:12.172
You know?

430
00:25:12.933 --> 00:25:19.756
So some people call it the Riemann-Euler Zeta function, but let's just give it to Riemann.

431
00:25:19.776 --> 00:25:21.877
Yeah, right, right, right.

432
00:25:22.217 --> 00:25:23.178
We already know who the goat is.

433
00:25:23.238 --> 00:25:24.038
Yeah, yeah, exactly.

434
00:25:24.058 --> 00:25:25.599
We already know who the goat is.

435
00:25:25.819 --> 00:25:26.319
Exactly.

436
00:25:26.660 --> 00:25:31.342
So he actually did some amazing work on the Riemann, on the Zeta function.

437
00:25:31.802 --> 00:25:35.564
What he did was extend Euler's definition to the complex variables.

438
00:25:36.305 --> 00:25:40.627
What I mean by that is, you know, usually...

439
00:25:42.028 --> 00:25:49.090
we think about that zeta function as like one over the squares, like one over the squares or one over the cubes.

440
00:25:49.171 --> 00:25:50.911
The exponent is a real number.

441
00:25:51.211 --> 00:25:54.572
You can even imagine one over square roots and things like that, right?

442
00:25:56.013 --> 00:26:00.194
Riemann said, no, what if we took one over stuff to the power of a complex number?

443
00:26:00.474 --> 00:26:03.535
So it's like one plus i or two plus i.

444
00:26:03.615 --> 00:26:04.436
What would that do?

445
00:26:04.696 --> 00:26:07.997
And it turns out there's a very well-defined way of talking about that.

446
00:26:08.277 --> 00:26:10.098
So here's how it works with complex numbers.

447
00:26:10.618 --> 00:26:10.818
Okay?

448
00:26:11.569 --> 00:26:15.694
Here, I'm going to borrow a lot of these visuals from 3Blue1Brown.

449
00:26:16.074 --> 00:26:24.565
And a lot of our viewers are going to recognize it as such, because he has an absolutely amazing video on YouTube about the Riemann zeta function.

450
00:26:25.345 --> 00:26:28.109
And he's the goat of mathematical visuals.

451
00:26:30.454 --> 00:26:30.994
Big shout out.

452
00:26:31.034 --> 00:26:31.635
Big shout out.

453
00:26:32.196 --> 00:26:35.921
So here's what happens if I take Riemann zeta to the 2 plus i.

454
00:26:36.782 --> 00:26:42.871
So in this case, the primary thing we're changing is the exponent in the denominator.

455
00:26:42.971 --> 00:26:46.056
Yeah, because that's how the Riemann zeta function is defined.

456
00:26:46.076 --> 00:26:46.396
Okay.

457
00:26:46.696 --> 00:26:51.059
The function is defined as the sum over 1 over the natural numbers to some power.

458
00:26:51.399 --> 00:26:52.380
You choose the power.

459
00:26:52.760 --> 00:26:52.941
Yeah.

460
00:26:52.981 --> 00:26:53.121
Right?

461
00:26:53.461 --> 00:26:57.824
And so we're just saying we've now chose the power to be 2 plus i instead of 2.

462
00:26:57.904 --> 00:27:00.466
If it was just 2, it would be the basal problem again.

463
00:27:00.546 --> 00:27:01.927
And then I'd get pi squared over 6.

464
00:27:01.987 --> 00:27:02.167
Yep.

465
00:27:02.347 --> 00:27:09.072
But because it's 2 plus i, if it's just 2, then every single term, I'm just moving along the number line.

466
00:27:09.092 --> 00:27:09.312
Mm-hmm.

467
00:27:09.913 --> 00:27:15.737
For 1, I'd get 1 plus 1 fourth plus 1 ninth, and I'd go a little bit closer to pi squared over 6.

468
00:27:16.038 --> 00:27:17.599
And I'd just be moving on the number line.

469
00:27:17.939 --> 00:27:21.763
With complex exponents, now I'm rotating on the complex plane.

470
00:27:22.083 --> 00:27:27.608
The complex plane is defined as the real numbers on the x-axis, the imaginary numbers on the y-axis.

471
00:27:27.868 --> 00:27:31.531
So that's why 2 plus i is 2 on the x, 1 on the y.

472
00:27:31.911 --> 00:27:34.333
2 plus 2i would mean 2 on the x, 2 on the y.

473
00:27:34.694 --> 00:27:34.974
Right.

474
00:27:35.354 --> 00:27:36.956
So that's how complex numbers are defined.

475
00:27:37.336 --> 00:27:41.600
And the way that you do an exponent is you actually start rotating where that's happening.

476
00:27:41.640 --> 00:27:43.902
Now, 1 to the power of anything is just 1.

477
00:27:44.142 --> 00:27:45.344
So that's why you go 1 first.

478
00:27:45.644 --> 00:27:49.027
But then 1 over 2 to the power of 2 plus i, that's going to—

479
00:27:49.397 --> 00:27:55.078
curve a little bit downward, and then it's going to curve a little bit downward for the 3, and it's going to spiral to that point.

480
00:27:55.658 --> 00:28:05.200
So the Riemann zeta function takes as input 2 plus i and maps it to that point down there, which is like 1.15 minus 0.44i.

481
00:28:05.360 --> 00:28:06.081
Okay?

482
00:28:07.281 --> 00:28:09.481
That's how it works in the complex plane.

483
00:28:09.661 --> 00:28:09.881
Yes.

484
00:28:10.822 --> 00:28:17.103
Now, turns out, as long as the real part is greater than 1,

485
00:28:18.509 --> 00:28:20.031
your sum is going to be fine.

486
00:28:20.111 --> 00:28:23.594
And we've got an animation here, courtesy of 3blue1brown again.

487
00:28:24.615 --> 00:28:30.081
As I change the input, the spiral is going to change, but it's going to be well-defined.

488
00:28:30.101 --> 00:28:31.102
It's going to end up somewhere.

489
00:28:31.602 --> 00:28:31.842
Okay?

490
00:28:32.163 --> 00:28:38.209
So here I'm taking my input, which is the yellow dot, and the output of the Riemann zeta function is where that spiral ends up.

491
00:28:38.629 --> 00:28:38.829
Okay?

492
00:28:39.130 --> 00:28:40.431
Perfectly well-defined function.

493
00:28:40.811 --> 00:28:41.292
Really nice.

494
00:28:41.312 --> 00:28:41.772
Okay?

495
00:28:41.792 --> 00:28:41.952
Okay?

496
00:28:43.726 --> 00:28:46.868
Let's see what this function does to the grid lines.

497
00:28:46.968 --> 00:28:48.989
Here we've seen individual points getting mapped.

498
00:28:49.329 --> 00:28:51.090
Now let's see what this function does to the grid lines.

499
00:28:51.150 --> 00:28:59.053
As long as the power, the real part, is greater than 1, this is how the grid lines map.

500
00:28:59.574 --> 00:29:02.675
Now, why do we need the powers to be greater than 1 for the real?

501
00:29:02.955 --> 00:29:08.178
That's because, imagine if I did the power equals 0.

502
00:29:09.251 --> 00:29:11.613
then I just have 1 plus 1 plus 1 plus 1.

503
00:29:11.653 --> 00:29:13.014
That's going to diverge to infinity.

504
00:29:13.995 --> 00:29:16.717
It's not well defined, at least in the series representation.

505
00:29:17.057 --> 00:29:24.383
But if my powers are greater than 1, then each fraction gets smaller and smaller, and so my spiral actually converges somewhere.

506
00:29:24.623 --> 00:29:25.283
That's the idea.

507
00:29:25.324 --> 00:29:25.944
Okay, got it.

508
00:29:26.004 --> 00:29:27.405
Okay?

509
00:29:27.685 --> 00:29:28.726
And so this is what's happening.

510
00:29:28.886 --> 00:29:36.993
All of the points on the right are getting mapped to these points that are in sort of half of the plane.

511
00:29:37.013 --> 00:29:37.473
Okay.

512
00:29:37.693 --> 00:29:40.477
But now this kind of begs the question, what about the other half?

513
00:29:40.757 --> 00:29:41.017
Right.

514
00:29:41.398 --> 00:29:41.638
Right?

515
00:29:41.818 --> 00:29:42.059
Right.

516
00:29:42.079 --> 00:29:45.643
It seems like I could just draw some lines.

517
00:29:47.065 --> 00:29:47.345
Right?

518
00:29:47.986 --> 00:29:50.909
I could just continue the function.

519
00:29:51.130 --> 00:29:51.350
Right.

520
00:29:51.510 --> 00:29:51.730
Right?

521
00:29:51.750 --> 00:29:52.692
Like, I could make up...

522
00:29:54.525 --> 00:29:59.746
how the rest of the function would behave, given I know so well how this function behaves over here.

523
00:29:59.926 --> 00:30:05.767
And specifically you're referencing to the left of the origin of the y-axis.

524
00:30:05.827 --> 00:30:06.867
Yeah, exactly.

525
00:30:06.887 --> 00:30:12.308
There's a little part on the right of the y-axis too that this thing is not reaching, right?

526
00:30:12.348 --> 00:30:14.908
It's like about half in there.

527
00:30:14.948 --> 00:30:23.310
But my point is, the parts where we can use the series give me this really nice functional map.

528
00:30:23.430 --> 00:30:23.610
Okay?

529
00:30:25.730 --> 00:30:29.332
And what if the idea is, can't we just extend what we're already seeing?

530
00:30:29.432 --> 00:30:29.973
Exactly.

531
00:30:30.193 --> 00:30:31.314
That's what Riman said.

532
00:30:31.714 --> 00:30:33.775
He's like, can I just extend what I'm already seeing?

533
00:30:34.275 --> 00:30:36.977
I know that it doesn't make sense, right?

534
00:30:37.117 --> 00:30:40.860
In terms of like, maybe in terms of like, yeah, you tell me to plug in zero, zero.

535
00:30:41.859 --> 00:30:44.541
It's not going to make sense because I'm just adding one, one, one, one, one.

536
00:30:44.921 --> 00:30:49.283
But in terms of what I'm seeing over here, it kind of makes sense, right?

537
00:30:49.544 --> 00:30:51.925
There's a world in which this could make sense.

538
00:30:52.785 --> 00:30:56.348
That's what Riemann's thinking to himself as he's looking at this thing, right?

539
00:30:57.588 --> 00:30:58.929
And so that's exactly what he does.

540
00:30:59.509 --> 00:31:02.611
He says, if I don't rely on the pesky series representation...

541
00:31:06.394 --> 00:31:09.978
which is this 1 over a power plus 2 over a power plus 3 over a power.

542
00:31:10.338 --> 00:31:15.765
Instead, I make up another function that behaves the same way on the right-hand side

543
00:31:17.479 --> 00:31:18.219
And it looks like that.

544
00:31:18.239 --> 00:31:29.662
2 to that power times pi to the power minus 1 times sine of something times the gamma function of something times the Riemann zeta of the flip hand side.

545
00:31:29.982 --> 00:31:41.005
So if he says, if I want to get to that part, I can take the part that's on the right, flip it, multiply it by like the gamma function of 1 minus s, multiply by all this other stuff, and I'll be fine.

546
00:31:42.264 --> 00:31:43.344
It'll be fine.

547
00:31:43.564 --> 00:31:47.165
It'll be fine because, and this is the magic of complex analysis.

548
00:31:47.645 --> 00:31:52.046
It turns out there's a very good reason why complex functions do this.

549
00:31:52.306 --> 00:31:54.487
I can just flip it on the other side of the board.

550
00:31:54.507 --> 00:31:57.147
Yeah, and make a little bit of correction.

551
00:31:57.608 --> 00:31:59.828
But all of the grid lines are going to be smooth now.

552
00:32:00.428 --> 00:32:02.169
The whole thing is going to be differentiable.

553
00:32:02.869 --> 00:32:06.210
And the fact of the matter is this is called analytic continuation.

554
00:32:06.230 --> 00:32:06.750
Okay.

555
00:32:06.850 --> 00:32:07.110
Okay?

556
00:32:07.550 --> 00:32:11.651
What it means is I can take a function that is well-behaved in one part,

557
00:32:12.691 --> 00:32:15.892
And if I want to extend it, there is only one way to extend it.

558
00:32:16.832 --> 00:32:17.413
That's the key.

559
00:32:17.453 --> 00:32:20.834
This is not an arbitrary extension.

560
00:32:21.694 --> 00:32:24.515
What complex analysis shows is there is only one way to do it.

561
00:32:24.555 --> 00:32:26.195
So if you found the way, that's it.

562
00:32:26.275 --> 00:32:27.316
That's it.

563
00:32:27.476 --> 00:32:27.856
That's it.

564
00:32:27.956 --> 00:32:31.177
You have found the continuation of that function.

565
00:32:31.197 --> 00:32:32.357
There is only one way to do it.

566
00:32:33.177 --> 00:32:33.598
Okay?

567
00:32:33.658 --> 00:32:34.758
And so this isn't like...

568
00:32:36.258 --> 00:32:37.119
It is imaginary.

569
00:32:37.159 --> 00:32:38.561
I mean, it's in complex variables.

570
00:32:38.601 --> 00:32:40.563
So there's an imaginary unit in it, right?

571
00:32:40.843 --> 00:32:43.166
But there's only one way to imagine it.

572
00:32:43.266 --> 00:32:44.467
In this imaginary space.

573
00:32:44.487 --> 00:32:44.908
Yes.

574
00:32:45.488 --> 00:32:51.855
To create everything and make sure that everything is sound and logically closed and

575
00:32:52.155 --> 00:32:53.557
There's only one way to imagine.

576
00:32:53.577 --> 00:32:56.479
There's a, the logical consistency only has one answer.

577
00:32:56.820 --> 00:32:57.260
Exactly.

578
00:32:57.641 --> 00:32:59.082
And so that is the extension.

579
00:32:59.142 --> 00:33:06.349
That is the analytic continuation of a meromorphic function outside of its domain where it's like nicely well-behaved.

580
00:33:06.689 --> 00:33:07.730
Now it's everywhere.

581
00:33:08.091 --> 00:33:08.391
Okay.

582
00:33:09.812 --> 00:33:11.915
That's the key thing that Riemann discovers.

583
00:33:12.275 --> 00:33:12.555
Okay.

584
00:33:13.156 --> 00:33:13.676
Side note.

585
00:33:13.696 --> 00:33:14.377
Okay.

586
00:33:15.277 --> 00:33:21.798
There was a lot of hype over the sum 1 plus 2 plus 3 plus 4 plus 5 equaling negative 1 twelfth.

587
00:33:22.439 --> 00:33:25.379
Now you can finally kind of understand why that is the case.

588
00:33:25.679 --> 00:33:32.641
If I plug in negative 1 to the Riemann zeta function, right, then that's going to be 1 plus 1 over 2 to the negative 1.

589
00:33:33.121 --> 00:33:34.821
But 2 to the negative 1 means I just flip.

590
00:33:34.921 --> 00:33:37.922
So I get 1 plus 2 plus 1 over 3 to the negative 1.

591
00:33:37.942 --> 00:33:38.582
So I just flip.

592
00:33:39.202 --> 00:33:39.382
Right?

593
00:33:39.722 --> 00:33:41.502
So that's just 1 plus 2 plus 3 plus 4.

594
00:33:41.842 --> 00:33:42.022
Right?

595
00:33:42.443 --> 00:33:43.923
It doesn't make sense on its own.

596
00:33:44.784 --> 00:33:57.397
But if you use the Riemann zeta function, you analytically continue it outside of its domain to the negative numbers, and then you plug in negative 1 to that functional thing that I showed you earlier with the signs and the gamma function and everything.

597
00:33:57.617 --> 00:34:00.160
You plug in negative 1 to that, you're going to get negative 112.

598
00:34:00.560 --> 00:34:04.925
And it turns out that if someone were to put a gun to your head and say...

599
00:34:06.644 --> 00:34:09.425
define this series as a number.

600
00:34:09.606 --> 00:34:10.606
You can't say infinity.

601
00:34:11.106 --> 00:34:12.267
You have to pick a number.

602
00:34:12.807 --> 00:34:15.208
The only correct answer is negative 112.

603
00:34:15.828 --> 00:34:16.048
Okay?

604
00:34:16.248 --> 00:34:26.192
To show you another reason why this is the case and why Srinivasa Ramanujan, who's a big hero of mine, is on this, is Ramanujan did not know anything about complex numbers.

605
00:34:26.672 --> 00:34:27.173
Very little.

606
00:34:27.233 --> 00:34:28.013
When he was in India.

607
00:34:28.393 --> 00:34:31.454
And he was reading like a trig book and discovering like all of math.

608
00:34:31.794 --> 00:34:32.054
Okay?

609
00:34:32.655 --> 00:34:32.935
Um...

610
00:34:34.705 --> 00:34:42.787
He wrote a letter to Hardy, who was a big mathematician in Cambridge, asking for help to go to Cambridge and work under him.

611
00:34:44.068 --> 00:34:49.349
One of the things that he wrote down was this 1 plus 2 plus 3 plus 4 dot dot dot equals negative 112.

612
00:34:49.629 --> 00:34:55.511
And he said, I know you'll probably think of putting me into an insane asylum, but look, I have techniques.

613
00:34:56.368 --> 00:34:58.970
that assign numbers to divergent series.

614
00:34:59.751 --> 00:35:01.012
And I got negative one 12th.

615
00:35:01.692 --> 00:35:12.480
Hardy looked at that and he was like, hold on, this guy makes no mention of complex analysis and imaginary numbers, but he still figured out a way to get to get to negative one 12th.

616
00:35:12.780 --> 00:35:16.123
And this, this was a known thing out there because Hardy is after Riemann, right?

617
00:35:16.403 --> 00:35:18.785
So there you can just plug in negative one and you know that.

618
00:35:20.041 --> 00:35:23.025
Ramanujan had no idea about Riman, about the Zeta function.

619
00:35:23.425 --> 00:35:30.696
He figured out a really hacky way to assign it a number, which is not exactly clean.

620
00:35:30.836 --> 00:35:31.557
Again, hacky.

621
00:35:31.658 --> 00:35:33.640
Now we know there's reasons why it works.

622
00:35:34.161 --> 00:35:34.582
But...

623
00:35:35.463 --> 00:35:38.044
He also came up with the same number, negative 112.

624
00:35:38.444 --> 00:35:40.464
Just goes to show you how closed math is.

625
00:35:41.104 --> 00:35:50.906
He was able to intuit from a totally separate pathway, but it still converged at the same destination, regardless of his lack of not going down the road already traveled.

626
00:35:50.926 --> 00:35:51.546
Yeah, yeah.

627
00:35:51.606 --> 00:35:53.567
He went a complete, but he found the same destination.

628
00:35:53.867 --> 00:35:54.887
It's kind of cool, right?

629
00:35:54.967 --> 00:36:00.448
And that's what I mean by like, if someone were to put a gun to your head and say, define what this is, and it can't be infinity,

630
00:36:01.248 --> 00:36:02.833
The only correct answer is negative 112.

631
00:36:02.893 --> 00:36:03.114
Okay?

632
00:36:03.375 --> 00:36:04.318
That's so fascinating.

633
00:36:04.518 --> 00:36:06.966
The real correct answer is really infinity, right?

634
00:36:07.387 --> 00:36:07.568
But...

635
00:36:08.562 --> 00:36:13.065
If you were to really put me through it, it would have to be negative 112, right?

636
00:36:13.285 --> 00:36:16.507
And the Riemann zeta of negative 1 is negative 112, right?

637
00:36:16.587 --> 00:36:20.770
It just – like I don't think it means anything to sum up the natural numbers.

638
00:36:20.830 --> 00:36:26.634
But like actually like string theorists use that all the time to like deal with their infinities.

639
00:36:26.674 --> 00:36:28.615
They're like, ah, I think negative 112.

640
00:36:28.675 --> 00:36:29.416
It's like, okay.

641
00:36:29.456 --> 00:36:31.177
The number of dimensions is the denominator.

642
00:36:32.928 --> 00:36:36.131
have some party trick where they do that.

643
00:36:36.471 --> 00:36:38.913
I just thought that's a really cool little aside.

644
00:36:38.973 --> 00:36:39.253
No, no.

645
00:36:39.293 --> 00:36:54.325
And that's, I think it's, because it's part of what we're doing is we're connecting all of these component parts, you know, mathematics being almost this closed system, right, that has these bounded box and a level of, like, the logical consistency

646
00:36:55.065 --> 00:37:04.387
that exists through it, even if you come from different disciplines within that subject matter, and seeing how those different disciplines exist.

647
00:37:05.913 --> 00:37:10.456
address the same question or try to approach the same questions.

648
00:37:10.476 --> 00:37:10.616
Yeah.

649
00:37:10.916 --> 00:37:12.097
It's really fascinating.

650
00:37:12.317 --> 00:37:12.517
Yeah.

651
00:37:12.577 --> 00:37:13.598
I mean, it's really cool.

652
00:37:13.618 --> 00:37:15.839
I mean, the basal problem is another example of that.

653
00:37:15.899 --> 00:37:18.481
Let me just give you another little anecdote from my personal life.

654
00:37:20.322 --> 00:37:24.145
I told you about Math 215 at Princeton University, the first semester of math that I did.

655
00:37:24.525 --> 00:37:27.787
One of the problems on the final was to calculate the basal problem.

656
00:37:28.087 --> 00:37:28.327
Okay.

657
00:37:28.568 --> 00:37:29.128
It's like, great.

658
00:37:29.168 --> 00:37:29.388
Okay.

659
00:37:29.448 --> 00:37:29.969
Euler did it.

660
00:37:30.029 --> 00:37:30.209
Yeah.

661
00:37:30.249 --> 00:37:32.131
You can do it for a final, you freshmen.

662
00:37:32.591 --> 00:37:34.533
And, uh, this, this was Peter Sarnak.

663
00:37:34.553 --> 00:37:45.323
He was just a maniac, but, um, he actually, in the problem statement, he, he gave us enough information and we used Fourier series and a clever trick with Fourier series to show,

664
00:37:45.843 --> 00:38:08.932
and i got that problem right it was a it was a great problem um but yeah you can use fourier series to show the pi squared over six you can use oiler's hack you can also use um taylor series a lot of times people use like um the taylor series for sign and things like that so there's so many different ways but you will always arrive at pi squared over six just the beauty of mathematics is awesome you gotta love it

665
00:38:09.312 --> 00:38:12.777
Okay, so now let's finally talk about the Riemann hypothesis.

666
00:38:13.158 --> 00:38:15.201
We have talked about the Riemann zeta function.

667
00:38:15.962 --> 00:38:17.424
Now let's talk about the hypothesis.

668
00:38:17.504 --> 00:38:25.235
Which is sort of the next step from starting with the basal problem that let us have a fundamental understanding.

669
00:38:25.395 --> 00:38:27.836
Yeah, we generalized it to the zeta function.

670
00:38:27.856 --> 00:38:28.576
To the zeta function.

671
00:38:28.816 --> 00:38:29.017
Yeah.

672
00:38:29.237 --> 00:38:33.158
And now... Now, we're finally going to get to the Riemann hypothesis.

673
00:38:33.198 --> 00:38:33.378
Right.

674
00:38:33.458 --> 00:38:33.699
Okay?

675
00:38:34.099 --> 00:38:39.541
Now, the Riemann zeta function has been used to prove that bound of primes.

676
00:38:39.821 --> 00:38:39.921
Yes.

677
00:38:39.941 --> 00:38:43.223
The n over log n. That's actually chapter 7 of this book.

678
00:38:43.523 --> 00:38:45.163
Which was the red line in the previous show.

679
00:38:45.183 --> 00:38:45.484
Yes, yeah.

680
00:38:45.564 --> 00:38:46.884
It's how do primes grow?

681
00:38:46.904 --> 00:38:53.087
They grow like n divided by log n. The distribution there and how close I get to n over log n...

682
00:38:53.917 --> 00:38:58.002
has to do with where the zeros are for the zeta function.

683
00:38:58.042 --> 00:39:02.068
What that means is, what inputs map to zero?

684
00:39:02.989 --> 00:39:11.139
If I plug in that input into the exponent of the, you know, one over the thingy, or the functional form...

685
00:39:11.259 --> 00:39:11.440
Yes.

686
00:39:12.533 --> 00:39:13.854
do I get zero as the output?

687
00:39:14.194 --> 00:39:15.994
Okay, there are two types of zeros.

688
00:39:16.415 --> 00:39:20.236
There are the trivial zeros, which happen at all of the negative even numbers.

689
00:39:20.616 --> 00:39:24.398
So if I put in negative two, I get zero.

690
00:39:24.618 --> 00:39:27.659
If I put in negative one, I get that negative one twelfth.

691
00:39:27.739 --> 00:39:27.919
Yes.

692
00:39:28.100 --> 00:39:30.100
Okay, but if I put in negative two, I get zero.

693
00:39:30.300 --> 00:39:32.361
If I put in negative four, I get zero.

694
00:39:32.661 --> 00:39:34.202
Those are the trivial zeros.

695
00:39:34.322 --> 00:39:34.542
Okay.

696
00:39:34.562 --> 00:39:38.404
And that just has to do with the fact that there's a sign in there, and sign goes like this.

697
00:39:38.924 --> 00:39:40.105
Okay, so it's trivial.

698
00:39:40.185 --> 00:39:41.065
It comes back to...

699
00:39:41.325 --> 00:39:46.416
Yeah, it's like the functional form has a sign in there, and so the sign, because the sign oscillates, you're going to get zeros.

700
00:39:46.597 --> 00:39:48.240
At a consistent interval.

701
00:39:48.260 --> 00:39:49.924
Yeah, yeah, and that's, okay, fine.

702
00:39:49.984 --> 00:39:50.205
Fine.

703
00:39:51.130 --> 00:39:53.552
Now there are the non-trivial zeros.

704
00:39:53.692 --> 00:39:54.973
Oh, my favorite, non-trivial.

705
00:39:55.013 --> 00:39:58.396
Yes, and they are non-trivial for a reason, because we don't know where they are.

706
00:39:59.137 --> 00:40:02.099
And that is the Riemann hypothesis.

707
00:40:02.260 --> 00:40:03.361
Where are they?

708
00:40:03.441 --> 00:40:06.503
Non-trivial, because everything else is well-defined at this point.

709
00:40:06.563 --> 00:40:06.804
Yeah.

710
00:40:07.164 --> 00:40:08.265
Except for the non-trivial zeros.

711
00:40:08.285 --> 00:40:11.207
The non-trivial zeros, we know that there's a bunch of more zeros.

712
00:40:11.307 --> 00:40:11.468
Yeah.

713
00:40:11.588 --> 00:40:12.508
But we don't know where they are.

714
00:40:12.528 --> 00:40:15.791
Okay, Riemann proved that they are inside of that strip.

715
00:40:16.252 --> 00:40:17.353
He said they're somewhere in here.

716
00:40:18.326 --> 00:40:20.347
Between zero and one on the real axis.

717
00:40:20.708 --> 00:40:21.908
That's called a critical strip.

718
00:40:22.249 --> 00:40:23.209
So Riemann proved that.

719
00:40:23.509 --> 00:40:31.074
And then in his paper, he writes, probably they're on the critical line.

720
00:40:31.094 --> 00:40:33.416
They're on the one half, right in the middle of that strip.

721
00:40:33.596 --> 00:40:33.916
Okay.

722
00:40:33.996 --> 00:40:34.236
Okay.

723
00:40:34.476 --> 00:40:35.057
That's what he writes.

724
00:40:35.597 --> 00:40:36.598
He's like, probably...

725
00:40:37.665 --> 00:40:44.116
But he's like, I don't want to prove it because for my purposes, it's not necessary.

726
00:40:44.216 --> 00:40:46.179
But it seems like an interesting problem.

727
00:40:46.279 --> 00:40:49.384
That's why it's called the hypothesis because he hypothesized it.

728
00:40:50.313 --> 00:40:51.054
In his paper.

729
00:40:51.574 --> 00:40:53.476
And now we're all chasing it.

730
00:40:53.556 --> 00:40:56.880
There's a search space that we believe these non-trivial zeros exist in.

731
00:40:57.040 --> 00:40:57.240
Yeah.

732
00:40:57.661 --> 00:41:00.604
And within that context, there is a discrete line.

733
00:41:00.904 --> 00:41:01.485
Right in the middle.

734
00:41:01.585 --> 00:41:08.011
Right in the middle, which is the sort of most likely place, the hypothesis of within this search space.

735
00:41:08.051 --> 00:41:08.252
Yeah.

736
00:41:08.732 --> 00:41:09.574
This is where you should look.

737
00:41:09.634 --> 00:41:10.796
This is where you should look.

738
00:41:10.816 --> 00:41:20.774
And if we can prove that they all lie only on that line and nowhere else, they're not in like some fudged like part around it.

739
00:41:21.792 --> 00:41:24.694
Every single non-trivial zero is on that line.

740
00:41:25.154 --> 00:41:26.455
That is the Riemann hypothesis.

741
00:41:26.475 --> 00:41:31.719
We want to resolve it similarly to how cleanly the trivial zeros resolve or the negative 1 12th resolve.

742
00:41:31.739 --> 00:41:32.259
Yeah, yeah.

743
00:41:32.619 --> 00:41:33.680
Where it's always true.

744
00:41:33.720 --> 00:41:34.320
Yes, yes.

745
00:41:34.400 --> 00:41:36.982
It's always true that every single zero is going to be on that line.

746
00:41:37.022 --> 00:41:37.623
Okay.

747
00:41:37.803 --> 00:41:38.203
Okay.

748
00:41:38.643 --> 00:41:40.805
That has been the quest.

749
00:41:41.105 --> 00:41:41.305
Yes.

750
00:41:41.445 --> 00:41:41.705
Okay.

751
00:41:41.985 --> 00:41:42.246
Okay.

752
00:41:42.706 --> 00:41:43.626
That seems so easy.

753
00:41:43.987 --> 00:41:45.228
Yeah.

754
00:41:45.668 --> 00:41:45.848
And...

755
00:41:47.030 --> 00:41:48.891
If we know, why would we care, right?

756
00:41:48.971 --> 00:42:00.279
Well, if we know where the zeros are for the Riemann hypothesis, then we could prescribe an exact form of the prime number counting theorem.

757
00:42:01.179 --> 00:42:01.479
Okay?

758
00:42:01.760 --> 00:42:03.321
And this next animation shows that.

759
00:42:03.661 --> 00:42:08.064
So the jagged line, that's the prime number counting theorem.

760
00:42:08.544 --> 00:42:09.765
And that has to be jagged, right?

761
00:42:09.805 --> 00:42:12.346
Because at some point it's going to increment by one.

762
00:42:12.707 --> 00:42:13.787
At two, it goes up by one.

763
00:42:13.867 --> 00:42:14.908
At three, it goes up by one.

764
00:42:15.228 --> 00:42:16.409
At five, it goes up by one.

765
00:42:16.629 --> 00:42:18.871
So all the vertical increases are that.

766
00:42:18.931 --> 00:42:20.393
Are just like, oh, here's another prime.

767
00:42:20.433 --> 00:42:21.254
Here's another prime.

768
00:42:22.215 --> 00:42:29.542
The blue is as you start incorporating the non-trivial zeros into your expression for what that line should be,

769
00:42:30.522 --> 00:42:32.524
The blue gets closer and closer to that line.

770
00:42:32.985 --> 00:42:33.765
So that's what we mean.

771
00:42:33.806 --> 00:42:40.393
We mean that if every single zero is known, we can exactly prescribe the form of that line.

772
00:42:41.073 --> 00:42:51.124
So this is the idea that this initial plot of where these primes are going to be as we get higher, it's fuzzy.

773
00:42:51.324 --> 00:42:51.524
Yeah.

774
00:42:51.844 --> 00:42:54.405
Because we don't know where the non-trivial zeros are.

775
00:42:54.445 --> 00:42:54.705
Yes.

776
00:42:55.005 --> 00:43:01.107
And if we knew where the non-trivial zeros are, it would resolve, the resolution would effectively be we could then do the primes.

777
00:43:01.127 --> 00:43:03.268
Yeah, we could predict every... Every primes all the way up.

778
00:43:03.288 --> 00:43:03.948
Yeah, yeah, yeah.

779
00:43:04.028 --> 00:43:05.468
That seems like a pretty big deal.

780
00:43:05.688 --> 00:43:06.108
Yeah.

781
00:43:06.368 --> 00:43:11.250
We could predict exactly the prime number function, the counting function.

782
00:43:11.710 --> 00:43:12.030
That's...

783
00:43:12.210 --> 00:43:12.450
Okay.

784
00:43:12.750 --> 00:43:12.970
Right?

785
00:43:13.090 --> 00:43:14.051
Yeah.

786
00:43:14.071 --> 00:43:15.311
And the primes are everything.

787
00:43:15.331 --> 00:43:15.511
Yeah.

788
00:43:15.631 --> 00:43:17.972
So I get now why this is so important.

789
00:43:18.012 --> 00:43:18.272
Yes.

790
00:43:18.292 --> 00:43:18.372
Yes.

791
00:43:19.278 --> 00:43:20.740
It's basically like a...

792
00:43:20.860 --> 00:43:23.382
It's sort of like...

793
00:43:24.103 --> 00:43:30.848
The location of these non-trivial zeros basically creates the last piece of the map to be able to then traverse wherever we want to.

794
00:43:31.049 --> 00:43:36.753
You'd still have to sum up to all of the infinity of zeros, but at least I got a procedure.

795
00:43:36.914 --> 00:43:37.434
To do so.

796
00:43:37.614 --> 00:43:39.556
And it just becomes a process.

797
00:43:39.576 --> 00:43:44.540
Yeah, and however accurate I need to be, I need to just find that many zeros.

798
00:43:44.880 --> 00:43:46.462
And with the amount of compute we have nowadays...

799
00:43:46.742 --> 00:43:48.244
Yeah, perhaps we could do it, right?

800
00:43:48.304 --> 00:43:52.488
If we could figure out a constructive way to find the zeros, that'd be crazy, right?

801
00:43:52.989 --> 00:43:59.716
If somebody proved a way to construct every single zero and show that there's no others, that would be crazy.

802
00:43:59.916 --> 00:44:00.797
Okay, yeah.

803
00:44:01.278 --> 00:44:03.701
And that's when people say we're trying to solve...

804
00:44:04.161 --> 00:44:05.442
The Riemann hypothesis.

805
00:44:05.542 --> 00:44:10.406
No, when people are saying they're trying to solve the Riemann hypothesis, they're just trying to show that every zero is here.

806
00:44:10.686 --> 00:44:12.447
It doesn't have to be a constructive proof.

807
00:44:12.467 --> 00:44:14.709
You could just show that there's no other zeros anywhere else.

808
00:44:14.869 --> 00:44:15.149
Okay.

809
00:44:15.710 --> 00:44:16.390
You know what I mean?

810
00:44:16.410 --> 00:44:16.811
Yes, yes.

811
00:44:16.831 --> 00:44:20.553
There's a difference between constructing it and just showing that none other exists.

812
00:44:20.593 --> 00:44:21.955
So they all happen to be on this line.

813
00:44:21.975 --> 00:44:24.056
And even that would be good enough.

814
00:44:24.076 --> 00:44:29.640
So there's sort of two levels of success is not the right word, but goals to reach.

815
00:44:29.821 --> 00:44:31.502
One is just proving that everything's on the critical line.

816
00:44:32.419 --> 00:44:33.780
Yeah, and that's the Riemann hypothesis.

817
00:44:33.800 --> 00:44:35.780
And it can't be anywhere else.

818
00:44:35.820 --> 00:44:37.681
If you do that, you've won the Millennium Problem.

819
00:44:37.701 --> 00:44:41.623
Jim Simons would have given us wealth for that.

820
00:44:41.723 --> 00:44:52.567
The idea that the construction, the constructive, is well beyond even where we can imagine, given that we haven't even been able to solve the problem of where they are not.

821
00:44:52.727 --> 00:44:52.947
Yeah.

822
00:44:53.187 --> 00:44:56.370
I mean, it could be that the constructive proof is the way to prove it, right?

823
00:44:56.710 --> 00:44:56.930
Okay.

824
00:44:56.950 --> 00:45:12.963
You know, there's history in mathematics that show that, like, for example, the real numbers, the way that you show what a real number is, is to explicitly construct it from something called Cauchy sequences of rational numbers.

825
00:45:12.983 --> 00:45:13.184
Okay.

826
00:45:13.244 --> 00:45:13.384
Right?

827
00:45:13.404 --> 00:45:13.544
Right.

828
00:45:13.964 --> 00:45:22.692
So there's a, like, both are, like, the remote hypothesis is such a black box that we don't even know which is going to be easier.

829
00:45:22.712 --> 00:45:23.934
Which angles the better way to go.

830
00:45:23.954 --> 00:45:26.496
That's something that Terence Tao has said in some of his lectures, right?

831
00:45:26.516 --> 00:45:26.936
I see, I see.

832
00:45:26.956 --> 00:45:32.622
If we want to prove, like, the whole thing, it seems we don't even have the toolkits to understand which way to go.

833
00:45:33.832 --> 00:45:36.894
There's a fork in the road and it's unclear which one is closer to the destination.

834
00:45:36.914 --> 00:45:37.294
Yeah, yeah.

835
00:45:37.334 --> 00:45:38.915
And there could be multiple forks.

836
00:45:39.735 --> 00:45:41.436
It's like we see like two, I guess.

837
00:45:41.476 --> 00:45:41.676
Right.

838
00:45:41.696 --> 00:45:45.498
But if the fog goes away, there could be like a hundred over here.

839
00:45:45.978 --> 00:45:46.619
Like we don't know.

840
00:45:47.119 --> 00:45:55.703
No, but that's really interesting in terms of understanding the value, the fundamental value of solving this.

841
00:45:55.863 --> 00:45:56.664
Yeah.

842
00:45:57.545 --> 00:45:58.325
As a problem.

843
00:45:58.426 --> 00:46:03.091
Yeah, it's like, I mean, it would give us an understanding of the atoms of numbers, right?

844
00:46:03.151 --> 00:46:05.754
Like it would give us an understanding of the periodic table of numbers.

845
00:46:06.174 --> 00:46:09.217
The periodic table of numbers consists of the primes, right?

846
00:46:09.298 --> 00:46:10.479
Everything else is made out of them.

847
00:46:12.471 --> 00:46:16.052
And this would tell us inherently like how that is structured.

848
00:46:16.733 --> 00:46:19.334
I mean, I think that is just beautiful in itself, right?

849
00:46:19.574 --> 00:46:29.058
Would it be like the photograph – it's sort of like what makes up the Rosalind Franklin X-ray crystallography photo.

850
00:46:29.418 --> 00:46:33.940
It would have that kind of similar like in a different field, that level of fundamental –

851
00:46:34.875 --> 00:46:35.955
Yeah, yeah.

852
00:46:35.995 --> 00:46:38.716
Like, if you figured out the structure of DNA, that's insane.

853
00:46:38.897 --> 00:46:39.057
Right.

854
00:46:39.097 --> 00:46:41.397
This would be like, yeah, that kind of shit.

855
00:46:41.437 --> 00:46:41.638
Yeah.

856
00:46:41.998 --> 00:46:42.858
I mean, it would be insane.

857
00:46:43.138 --> 00:46:44.559
So, some other stuff, right?

858
00:46:44.699 --> 00:46:50.121
If the Riemann hypothesis is proven true, it instantly validates hundreds of conditional mathematics theorems.

859
00:46:50.581 --> 00:46:55.363
There's been so many theorems that have been proven assuming that the hypothesis is true.

860
00:46:55.723 --> 00:47:01.865
Because the hypothesis has been so ridiculously hard to prove, but every single zero that we find is on the line.

861
00:47:02.205 --> 00:47:03.706
When we computationally find it...

862
00:47:04.486 --> 00:47:06.607
So let's just assume it's true.

863
00:47:06.888 --> 00:47:10.550
And then there's so much other mathematics that falls through because of that.

864
00:47:11.130 --> 00:47:16.874
And so the surrounding surface area kind of points to, yes, it is true, even though we can't yet.

865
00:47:17.074 --> 00:47:17.934
Yeah.

866
00:47:18.134 --> 00:47:22.477
Either by saying that the zeros are no Ross or constructively proving it that it's true.

867
00:47:22.577 --> 00:47:23.057
Yeah, yeah.

868
00:47:23.098 --> 00:47:24.359
I mean, it's...

869
00:47:24.559 --> 00:47:25.680
So it would be huge, right?

870
00:47:25.980 --> 00:47:26.280
Yeah.

871
00:47:26.821 --> 00:47:27.982
Yeah.

872
00:47:28.002 --> 00:47:29.183
You kind of get it now, right?

873
00:47:29.203 --> 00:47:30.044
That's a big deal.

874
00:47:30.324 --> 00:47:30.665
Yeah.

875
00:47:30.685 --> 00:47:33.607
So this is why everyone's like, we want to have a positive.

876
00:47:33.687 --> 00:47:37.251
Oh, I isn't anything until it really solved... Well, if it solves it, then that...

877
00:47:37.491 --> 00:47:37.771
Anyway.

878
00:47:37.791 --> 00:47:38.032
Yeah.

879
00:47:38.312 --> 00:47:39.733
If it solves it, then that'd be great.

880
00:47:40.053 --> 00:47:42.456
It hasn't solved it, and we're going to get into exactly what it did.

881
00:47:42.616 --> 00:47:42.796
Okay.

882
00:47:42.816 --> 00:47:43.016
Okay?

883
00:47:43.417 --> 00:47:43.637
So...

884
00:47:45.255 --> 00:47:54.780
After Riemann's paper, it took about 40 years to really iron out the proof of the prime number counting theorem and to show that it followed.

885
00:47:55.680 --> 00:48:01.883
Riemann did a lot of the legwork, but then there was Jakis Hadamard, the guy behind Hadamard matrices for those in quantum computing.

886
00:48:02.383 --> 00:48:05.184
And also, I don't know how you say this.

887
00:48:05.445 --> 00:48:06.065
Can you try?

888
00:48:06.085 --> 00:48:09.086
Charles-Jean de la Vallée poussant poissin.

889
00:48:10.667 --> 00:48:11.627
Yeah, there we go.

890
00:48:11.887 --> 00:48:16.669
So those two guys, they proved using Riemann's complex analytic methods.

891
00:48:17.289 --> 00:48:21.770
And in 1914, the great English mathematician G.H.

892
00:48:21.830 --> 00:48:32.333
Hardy, who is shown here, he's most well-known for discovering Ramanujan, but he's also well-known for some of the math that he did.

893
00:48:33.194 --> 00:48:37.435
One of the things that he did was show that there are an infinite number of zeros on the critical line.

894
00:48:38.400 --> 00:48:39.861
Okay, before we didn't even know that.

895
00:48:39.901 --> 00:48:41.202
We just knew that they were in the strip.

896
00:48:41.543 --> 00:48:42.884
Riemann said they're in that strip.

897
00:48:43.404 --> 00:48:45.446
Hardy said there's an infinite number in that line.

898
00:48:45.826 --> 00:48:46.967
Now, that says a lot.

899
00:48:47.067 --> 00:48:48.788
It says that there's an infinite number of zeros.

900
00:48:50.630 --> 00:48:54.613
It also says almost nothing about the Riemann hypothesis.

901
00:48:55.033 --> 00:48:59.057
Because it could be that there's many, many infinitely more elsewhere on that strip.

902
00:48:59.177 --> 00:48:59.377
Right.

903
00:48:59.537 --> 00:48:59.717
Right?

904
00:48:59.937 --> 00:49:00.118
Right.

905
00:49:00.138 --> 00:49:03.800
So it's a step in the right direction, I'd say.

906
00:49:04.381 --> 00:49:06.443
But it's not quite the whole thing, right?

907
00:49:07.564 --> 00:49:07.784
Yeah.

908
00:49:07.804 --> 00:49:07.904
Yeah.

909
00:49:09.237 --> 00:49:20.538
Next, we've got, finally, in 1942, Atlee Selberg, he makes a step in the direction of the Riemann hypothesis.

910
00:49:20.558 --> 00:49:20.778
Okay.

911
00:49:20.878 --> 00:49:21.139
Okay?

912
00:49:23.141 --> 00:49:28.305
He won the Fields Medal later in 1950, and this is a photo of him at the Institute of Advanced Study at Princeton.

913
00:49:28.865 --> 00:49:36.130
So he proves that there is a positive proportion of non-trivial zeros on the critical line.

914
00:49:36.710 --> 00:49:45.516
What that means is there is some percent of zeros, at least some percent of zeros on that one half line.

915
00:49:45.916 --> 00:49:50.299
So the point is, there's this number of non-trivial zeros.

916
00:49:51.319 --> 00:49:52.301
There's infinitely many.

917
00:49:52.481 --> 00:49:56.506
There's infinitely many because of Hardy.

918
00:49:56.526 --> 00:50:01.732
And some percentage of those, because we don't yet know if they're only on the critical line.

919
00:50:01.772 --> 00:50:02.273
Yeah, yeah.

920
00:50:02.373 --> 00:50:04.956
And so if they're only on the critical line, it would be 100%.

921
00:50:04.976 --> 00:50:05.177
Right.

922
00:50:05.217 --> 00:50:05.537
He said...

923
00:50:07.239 --> 00:50:08.880
The percentage is more than zero.

924
00:50:09.040 --> 00:50:09.260
Right.

925
00:50:09.721 --> 00:50:11.001
I don't know what the number is.

926
00:50:11.362 --> 00:50:11.542
Right.

927
00:50:11.842 --> 00:50:13.983
I'm just telling you that the percentage is more than zero.

928
00:50:14.003 --> 00:50:16.124
There's at least one on the critical line.

929
00:50:16.144 --> 00:50:16.565
Well, no.

930
00:50:16.585 --> 00:50:18.086
We know that this is...

931
00:50:18.266 --> 00:50:18.906
Okay.

932
00:50:19.006 --> 00:50:20.367
I should be careful here.

933
00:50:20.527 --> 00:50:20.807
Okay.

934
00:50:20.867 --> 00:50:22.788
We've already proven that there's an infinite number.

935
00:50:23.169 --> 00:50:23.429
Right.

936
00:50:23.689 --> 00:50:24.529
On the critical line, right?

937
00:50:24.569 --> 00:50:24.649
Yes.

938
00:50:24.890 --> 00:50:27.031
But here's what I mean by it could be zero percent.

939
00:50:27.871 --> 00:50:33.635
There could be an infinite number on the critical line, and then there could be for every single zero an infinite number...

940
00:50:34.658 --> 00:51:03.392
elsewhere on the strip so for every zero there's an infinity elsewhere and so even though i have an infinite number it's still zero percent of the total infinities because in the larger search area yeah there's an equal amount of there's similar this is the stuff i was talking about infinities everywhere yeah so so you know what i mean no that's a that's a good distinction no that that's a good it's kind of weird to because we're dealing with some weird stuff here yeah so we're dealing with infinities but that infinity could still be zero percent he showed it's not it has to be more

941
00:51:04.513 --> 00:51:04.753
Okay.

942
00:51:04.773 --> 00:51:05.033
Okay.

943
00:51:05.373 --> 00:51:06.294
Which is, okay.

944
00:51:06.314 --> 00:51:07.194
We're from zero.

945
00:51:07.234 --> 00:51:08.334
We went to non-zero.

946
00:51:08.754 --> 00:51:08.975
Okay.

947
00:51:09.055 --> 00:51:10.195
Everyone was super excited.

948
00:51:10.215 --> 00:51:11.055
He won the Fields Medal.

949
00:51:11.095 --> 00:51:11.275
Right.

950
00:51:11.375 --> 00:51:11.715
Okay.

951
00:51:11.735 --> 00:51:12.736
He's like, all right, that's dope.

952
00:51:13.096 --> 00:51:13.956
All right.

953
00:51:14.176 --> 00:51:17.617
So next we have Levinson in 1974.

954
00:51:18.078 --> 00:51:20.438
He pushes that number up to 33%.

955
00:51:21.579 --> 00:51:21.859
Okay.

956
00:51:21.879 --> 00:51:22.939
That seems like a big deal.

957
00:51:22.959 --> 00:51:23.099
Right?

958
00:51:23.359 --> 00:51:26.860
So he's like a third of the zeros, at least.

959
00:51:27.041 --> 00:51:29.902
At least a third of the non-trivial zeros are on the critical line.

960
00:51:29.922 --> 00:51:30.642
Are on the critical line.

961
00:51:30.662 --> 00:51:30.802
Okay.

962
00:51:31.915 --> 00:51:33.155
Infinitely many 34%.

963
00:51:33.656 --> 00:51:34.856
Yeah, yeah.

964
00:51:34.956 --> 00:51:44.219
It's like – a crude way of saying it is for every non-trivial zero on the critical line, there are at most two outside.

965
00:51:44.739 --> 00:51:44.919
Right?

966
00:51:44.939 --> 00:51:47.840
Like I could make like a pairing and be like for this one, I'm going to take two.

967
00:51:47.980 --> 00:51:49.320
For this one, I'm going to take two.

968
00:51:49.380 --> 00:51:50.841
And if you do that, you'll cover everything.

969
00:51:50.861 --> 00:51:51.241
I understand.

970
00:51:51.661 --> 00:51:53.143
That's how you want to think about it.

971
00:51:53.163 --> 00:51:53.924
Yep, that makes sense.

972
00:51:53.944 --> 00:52:00.790
So even though there's an infinity of it, you can still make a case that there's a third of that infinity is on the critical line and the rest is elsewhere.

973
00:52:01.011 --> 00:52:03.773
So that's what Levinson shows, okay?

974
00:52:03.874 --> 00:52:12.002
And he does this weird thing with like mollifiers where he basically tries to like blur out the Riemann zeta function and then work with that blurred out version.

975
00:52:12.022 --> 00:52:12.242
Mm-hmm.

976
00:52:12.502 --> 00:52:17.047
I don't know how it works, but that's what Wikipedia says.

977
00:52:17.168 --> 00:52:17.328
Okay?

978
00:52:17.408 --> 00:52:20.371
So, and now we're getting into stuff that I really don't understand.

979
00:52:20.472 --> 00:52:20.972
Right.

980
00:52:21.012 --> 00:52:22.234
No fair.

981
00:52:22.354 --> 00:52:27.260
So now I'm going to be regurgitating what I sort of figured out with my readings.

982
00:52:27.360 --> 00:52:28.020
Full disclosure.

983
00:52:28.040 --> 00:52:28.421
Okay.

984
00:52:28.851 --> 00:52:31.034
Levinson puts it up to one third.

985
00:52:32.015 --> 00:52:34.759
Other people take his method and start inching that up.

986
00:52:35.319 --> 00:52:37.762
So we've got Conray in 1989.

987
00:52:37.863 --> 00:52:39.184
He pushes it up to two fifths.

988
00:52:39.845 --> 00:52:40.105
So 40%.

989
00:52:40.126 --> 00:52:40.987
This is now 15 years later.

990
00:52:41.007 --> 00:52:41.227
Yeah.

991
00:52:41.427 --> 00:52:41.768
Yeah.

992
00:52:44.611 --> 00:52:49.094
15 years later, in 2022, so this is what, 30 years later?

993
00:52:49.154 --> 00:52:49.495
Yeah.

994
00:52:49.875 --> 00:52:50.936
We're up to 41.7%.

995
00:52:50.996 --> 00:52:53.518
So over, yeah.

996
00:52:53.538 --> 00:52:56.580
So there was a quick jump by... To 33.

997
00:52:58.101 --> 00:53:01.283
There's an immediate jump from zero to 34.

998
00:53:01.384 --> 00:53:01.784
Yeah.

999
00:53:01.804 --> 00:53:03.265
There's like three big jumps, I'd say.

1000
00:53:03.325 --> 00:53:03.465
Right.

1001
00:53:03.525 --> 00:53:05.286
You go from zero to not zero.

1002
00:53:05.507 --> 00:53:05.667
Right.

1003
00:53:05.687 --> 00:53:05.887
Yes.

1004
00:53:06.067 --> 00:53:06.267
Yes.

1005
00:53:06.367 --> 00:53:08.569
Which even though on the graph, it's like... At the bottom.

1006
00:53:08.829 --> 00:53:09.830
The fact that it starts.

1007
00:53:10.110 --> 00:53:10.350
Right.

1008
00:53:10.590 --> 00:53:10.771
Right?

1009
00:53:10.791 --> 00:53:10.891
Yes.

1010
00:53:10.911 --> 00:53:11.091
Yes.

1011
00:53:11.591 --> 00:53:12.131
That's huge.

1012
00:53:12.511 --> 00:53:13.011
Selberg.

1013
00:53:13.131 --> 00:53:13.791
Yeah.

1014
00:53:13.951 --> 00:53:14.412
In 42.

1015
00:53:14.452 --> 00:53:14.672
42.

1016
00:53:15.472 --> 00:53:18.512
And then in 74, it goes up to 34.7.

1017
00:53:18.612 --> 00:53:18.992
Big jump.

1018
00:53:19.312 --> 00:53:21.473
Then it goes to 40.9, another pretty big jump.

1019
00:53:21.833 --> 00:53:24.373
And then it took us 30 years to get to 41.7.

1020
00:53:24.453 --> 00:53:25.854
Just to even get one more percent.

1021
00:53:25.874 --> 00:53:26.094
Yeah.

1022
00:53:27.074 --> 00:53:27.454
Not even.

1023
00:53:27.534 --> 00:53:28.454
Not even one more percent.

1024
00:53:28.474 --> 00:53:28.854
Yeah, 0.8.

1025
00:53:29.134 --> 00:53:29.394
Okay.

1026
00:53:30.035 --> 00:53:34.975
And now Claude comes in in 2026, and it has brought it up to 67.25%.

1027
00:53:34.995 --> 00:53:35.156
Okay.

1028
00:53:35.176 --> 00:53:35.376
Okay.

1029
00:53:39.410 --> 00:53:41.334
The Riemann hypothesis is at 100.

1030
00:53:42.637 --> 00:53:43.439
We're getting closer.

1031
00:53:43.539 --> 00:53:43.799
Yes.

1032
00:53:44.421 --> 00:53:44.681
Yes.

1033
00:53:45.603 --> 00:53:46.165
However...

1034
00:53:48.319 --> 00:53:51.722
this is the case because of what you just talked about with the infinities.

1035
00:53:51.762 --> 00:53:55.126
Because what I'm thinking now is like, well, we could get to 99, 999.

1036
00:53:56.367 --> 00:53:57.048
Yeah.

1037
00:53:58.069 --> 00:53:58.809
And it doesn't mean anything.

1038
00:53:58.829 --> 00:53:59.550
It doesn't mean anything.

1039
00:53:59.630 --> 00:54:01.352
For the Riemann hypothesis, it doesn't.

1040
00:54:01.873 --> 00:54:02.113
Right?

1041
00:54:02.153 --> 00:54:04.255
Because what really matters is... Is the 100.

1042
00:54:04.776 --> 00:54:09.460
And Claude and other mathematicians have acknowledged that the way that Claude did this...

1043
00:54:10.317 --> 00:54:11.557
it's not going to get you to 100.

1044
00:54:12.057 --> 00:54:12.337
Okay?

1045
00:54:12.458 --> 00:54:15.278
There is a limit to how we push this.

1046
00:54:15.458 --> 00:54:15.718
Okay?

1047
00:54:15.838 --> 00:54:16.098
Okay?

1048
00:54:16.378 --> 00:54:17.659
We're not going to get to 100 this way.

1049
00:54:17.899 --> 00:54:19.779
But it's still cool to get close.

1050
00:54:19.899 --> 00:54:20.079
Right.

1051
00:54:20.199 --> 00:54:20.419
Okay?

1052
00:54:20.439 --> 00:54:23.880
Because it makes us understand a bit more about the function in some sense.

1053
00:54:24.320 --> 00:54:37.043
Part of it, and part of what I've heard people sort of talk about is, you know, as these models begin to connect these ideas that may already exist in ways that had not necessarily been connected, folks who are experts and mathematicians deep in these fields

1054
00:54:40.624 --> 00:54:58.072
oh let me now think and we now have six forks in the road i didn't see the six fork in the road because of the fog i can now walk down it because i have the tools to do so exactly exactly yeah so now let's get into how claude actually did it yep okay i've got a few a bit of understanding of how it did it okay um claude

1055
00:54:58.862 --> 00:55:00.323
It, again, tweets.

1056
00:55:00.563 --> 00:55:03.145
That's how they announced, I guess, that they were doing this, right?

1057
00:55:03.445 --> 00:55:06.086
In what appears to be, it's like a matter of like 30 hours.

1058
00:55:06.667 --> 00:55:07.427
Yeah.

1059
00:55:07.767 --> 00:55:09.388
They've pushed that bound to 67.2%.

1060
00:55:09.869 --> 00:55:15.552
They are careful to say it didn't solve it, but it made strides on a related problem.

1061
00:55:15.812 --> 00:55:16.132
Okay.

1062
00:55:16.873 --> 00:55:20.235
Now, there are two things that Claude used that are important to know here.

1063
00:55:20.355 --> 00:55:20.635
Okay.

1064
00:55:20.835 --> 00:55:21.015
Okay.

1065
00:55:21.516 --> 00:55:25.378
The first is a technique developed by Hugh Montgomery in 1973.

1066
00:55:25.438 --> 00:55:26.559
He was a mathematician.

1067
00:55:27.059 --> 00:55:30.242
And the legend has it that he was actually talking to Freeman Dyson.

1068
00:55:30.362 --> 00:55:31.282
Freeman!

1069
00:55:31.522 --> 00:55:33.924
Who we've covered a lot on this podcast.

1070
00:55:34.304 --> 00:55:39.228
He was talking to Freeman Dyson, and he was looking at the imaginary parts of the zeros of the Riemann zeta function.

1071
00:55:40.068 --> 00:55:42.170
And he analyzed how they sort of clumped together.

1072
00:55:42.770 --> 00:55:57.181
and showed that it's the... And he was talking of Dyson, and Dyson was like, wait, that's exactly the statistical distribution that physicists have been looking at for energy levels of heavy atomic nuclei, like uranium and things like that.

1073
00:55:57.421 --> 00:56:02.144
The energy levels of the nuclei have to do with the zeros of the Riemann-Zeta function.

1074
00:56:02.525 --> 00:56:05.367
It's kind of crazy how, like, the world works that way.

1075
00:56:05.887 --> 00:56:09.290
Now, the catch was that when Montgomery did his original breakthrough...

1076
00:56:10.010 --> 00:56:14.858
It required assuming that the Riemann hypothesis was true to get his formulas to work.

1077
00:56:15.259 --> 00:56:20.488
Later on, very recently, there's a team of mathematicians that published work showing that that

1078
00:56:21.602 --> 00:56:26.785
that pair correlation thing that he was doing with all those zeros, it works unconditionally.

1079
00:56:26.965 --> 00:56:29.987
It works without assuming that the Riemann zeta function works.

1080
00:56:30.267 --> 00:56:37.111
So this is something that has to do with the Riemann zeta function that Montgomery sort of proved on the side.

1081
00:56:37.632 --> 00:56:47.678
The second thing is in 2000, the legendary Fields medalist Enrico Bombieri, he wrote a note, a little monograph,

1082
00:56:48.683 --> 00:56:52.686
where the Clay Mathematics Institute was forming the Millennium Problems.

1083
00:56:53.246 --> 00:57:00.031
So he wrote the note for the Riemann zeta function, and he's like, this is why I think it belongs in the Millennium Problems.

1084
00:57:00.071 --> 00:57:07.436
I mean, everyone believed that it did, but this is a preeminent mathematician, and he was sort of dumping ideas about how to approach it.

1085
00:57:07.956 --> 00:57:17.783
And one of the things that he talked about was how modern analytic number theory handles the architecture of these zeta zeros, and he dives into something called quadratic forms,

1086
00:57:18.603 --> 00:57:21.706
And Wiles explicit quadratic functionals.

1087
00:57:22.127 --> 00:57:25.710
I don't know what they are, but that's what he mentions in his writeup.

1088
00:57:26.291 --> 00:57:26.531
All right.

1089
00:57:27.032 --> 00:57:27.192
Now,

1090
00:57:28.474 --> 00:57:29.474
Let's see what Claude does.

1091
00:57:30.115 --> 00:57:39.999
Yeah, because you're saying these are two fundamental, the Montgomery and Bombieri, no relation to Guy Fieri insights, were building blocks.

1092
00:57:40.319 --> 00:57:40.980
For what Claude is doing.

1093
00:57:41.000 --> 00:57:41.780
For what Claude is doing.

1094
00:57:41.800 --> 00:57:41.980
Okay.

1095
00:57:42.280 --> 00:57:44.441
So Jared Sumner, right?

1096
00:57:44.481 --> 00:57:45.942
He's an Anthropoc staff member.

1097
00:57:46.662 --> 00:57:48.723
He's got no deep mathematical background.

1098
00:57:49.143 --> 00:57:52.225
Apparently while jogging, I don't know how much this is hype.

1099
00:57:52.905 --> 00:57:53.085
Right?

1100
00:57:53.105 --> 00:58:02.448
Because like the, look, the conjecture, the Jacobian conjecture, they happened to do it at the time of the FIFA World Cup final.

1101
00:58:02.508 --> 00:58:03.568
Now he's jogging.

1102
00:58:03.748 --> 00:58:04.949
It's like, you know.

1103
00:58:05.309 --> 00:58:06.009
Great marketing.

1104
00:58:06.049 --> 00:58:07.530
Yeah, it is great marketing.

1105
00:58:07.610 --> 00:58:09.190
So anyways, let's take their word for it.

1106
00:58:09.250 --> 00:58:09.930
He's jogging.

1107
00:58:10.270 --> 00:58:12.831
He opens up a clawed terminal while jogging.

1108
00:58:12.971 --> 00:58:14.172
On his phone.

1109
00:58:14.212 --> 00:58:16.112
Why don't you watch some Netflix or something?

1110
00:58:16.152 --> 00:58:17.993
Or, hey, listen to this podcast.

1111
00:58:18.033 --> 00:58:18.573
Right, right.

1112
00:58:18.953 --> 00:58:20.454
That's a great choice.

1113
00:58:20.614 --> 00:58:22.996
I guess they're always working out in Anthropic, okay?

1114
00:58:23.436 --> 00:58:29.760
So he casually prompted the unreleased model to take a real stab at the Riemann hypothesis, okay?

1115
00:58:30.040 --> 00:58:34.083
The model exhibited initial skepticism about its own ability.

1116
00:58:35.024 --> 00:58:40.667
And the human said, you know, almost comedic encouragements of like, keep going and believe in yourself.

1117
00:58:40.847 --> 00:58:41.688
Very Ted Lasso.

1118
00:58:42.008 --> 00:58:42.268
Yeah.

1119
00:58:42.929 --> 00:58:43.130
Yeah.

1120
00:58:43.170 --> 00:58:46.454
He Ted lassoed Claude into getting closer to the Riemann Zeta.

1121
00:58:46.774 --> 00:58:47.175
Insane.

1122
00:58:47.595 --> 00:58:50.339
So behind the scenes, the execution is not casual.

1123
00:58:50.519 --> 00:58:50.739
Okay.

1124
00:58:50.759 --> 00:58:52.241
What Claude is doing is not casual.

1125
00:58:52.581 --> 00:58:59.290
The primary Claude model that was enacted on by Sumner, it starts orchestrating.

1126
00:58:59.430 --> 00:59:00.712
It becomes an orchestrator.

1127
00:59:02.278 --> 00:59:08.827
And it spins up a hierarchical swarm of 60 autonomous sub-agents to try and tackle this problem.

1128
00:59:09.347 --> 00:59:13.713
It built a team of other agents, other versions of itself.

1129
00:59:14.752 --> 00:59:17.153
ostensibly that had one thing to focus on.

1130
00:59:17.173 --> 00:59:17.393
Yeah.

1131
00:59:17.834 --> 00:59:23.776
And then it could coordinate them because they're not trying to do too many things as one instantiation of itself.

1132
00:59:23.956 --> 00:59:25.237
Exactly, yeah.

1133
00:59:25.257 --> 00:59:28.579
And these agents aren't just like writers of mathematical proof.

1134
00:59:28.699 --> 00:59:29.619
That's what's key, right?

1135
00:59:30.380 --> 00:59:34.982
These modern models now have the ability to go onto a virtual machine...

1136
00:59:35.482 --> 00:59:38.605
They can write, debug, and execute code.

1137
00:59:39.046 --> 00:59:42.750
They can write Python programs and execute those Python programs.

1138
00:59:43.010 --> 00:59:52.139
So between them, they ran like 2,400 shell commands, hundreds of Python scripts, consumed 31 million output tokens,

1139
00:59:53.120 --> 00:59:56.941
And they autonomously downloaded 54 academic papers from Archive.

1140
00:59:58.081 --> 01:00:00.342
The preprint server run by Cornell.

1141
01:00:00.422 --> 01:00:01.142
Yeah, that's right.

1142
01:00:01.702 --> 01:00:09.964
And now this, you can finally see, as you were saying, like this methodology, it mirrors like a high performance, high speed research lab.

1143
01:00:10.164 --> 01:00:10.344
Right.

1144
01:00:10.404 --> 01:00:10.564
Right.

1145
01:00:10.584 --> 01:00:13.585
There's like a PI and then there's like postdocs and...

1146
01:00:15.146 --> 01:00:19.829
PhD students and then they've got their undergrad underlings that are like doing work.

1147
01:00:20.029 --> 01:00:23.752
It's an entire lab in a single like automated ecosystem.

1148
01:00:23.772 --> 01:00:24.072
Yeah.

1149
01:00:24.267 --> 01:00:25.628
You have some people doing rote work.

1150
01:00:25.828 --> 01:00:28.190
You have some people doing sort of ideation.

1151
01:00:28.490 --> 01:00:36.156
You have some people doing sort of functional mathematical number crunching, validation, verification.

1152
01:00:37.317 --> 01:00:44.782
And what's so fascinating about this is that the initial orchestrator from the initial prompt is,

1153
01:00:45.663 --> 01:00:55.211
can architect the approach without having to be dictated to what the right construction of that is.

1154
01:00:55.571 --> 01:00:56.432
That's exactly right.

1155
01:00:56.472 --> 01:00:59.334
This is another point that I'm glad that you're highlighting here.

1156
01:00:59.715 --> 01:01:05.520
The initial prompt was just take a real stab at the Riemann hypothesis and believe in yourself.

1157
01:01:06.140 --> 01:01:06.600
That's it.

1158
01:01:06.740 --> 01:01:07.101
That's it.

1159
01:01:07.121 --> 01:01:07.321
Right?

1160
01:01:07.401 --> 01:01:09.542
Like, this guy was not a mathematician.

1161
01:01:10.222 --> 01:01:14.324
All of the mathematics is coming from Claude itself, which is kind of crazy.

1162
01:01:14.404 --> 01:01:14.604
Right.

1163
01:01:15.044 --> 01:01:19.607
And even if it didn't have the right starting point, it just basically said, I don't have the right starting point.

1164
01:01:19.967 --> 01:01:21.888
So let me go gather enough context.

1165
01:01:21.908 --> 01:01:22.068
Yeah.

1166
01:01:22.728 --> 01:01:30.773
to then narrow my search space, my area of operation, to something that's based off of existing whatever.

1167
01:01:30.833 --> 01:01:45.563
I mean, again, we're potentially going to get there in a second, but I just – this is not to say that anyone can just go in and say, solve a millennium problem and it will necessarily work, but I do think it's an important –

1168
01:01:46.984 --> 01:02:02.724
to understand that the construction that a lot of us have about what is quote AI of which LLM's next token prediction is a single Lego block part of what are now these much more complex systems.

1169
01:02:03.285 --> 01:02:12.949
We can't view the capability set on what is no longer the frontier execution of what's happening with these things.

1170
01:02:13.269 --> 01:02:16.390
Again, that doesn't mean they can have novel insight and be created.

1171
01:02:16.951 --> 01:02:20.532
But there are so much domain space in taking –

1172
01:02:22.665 --> 01:02:32.050
orthogonal or correlated areas of a variety of areas of study in science and mathematics, and just making connections no one else has.

1173
01:02:32.090 --> 01:02:32.251
Yeah.

1174
01:02:32.571 --> 01:02:40.175
And there's so, there's, I think people underestimate how much can, may be able to be done just making existing connections.

1175
01:02:40.275 --> 01:02:40.775
Exactly.

1176
01:02:41.176 --> 01:02:45.578
No, that's totally, I mean, for example, we had just covered Montgomery and Dyson, right?

1177
01:02:46.039 --> 01:02:48.080
If those two didn't have lunch or whatever...

1178
01:02:48.740 --> 01:02:49.041
Bingo.

1179
01:02:49.261 --> 01:02:49.942
Then what?

1180
01:02:50.022 --> 01:02:55.990
Maybe Montgomery would never have done his while whatever, you know, thing that he did for the Riemann-Zeta hypothesis.

1181
01:02:56.030 --> 01:02:56.610
Right, right.

1182
01:02:56.630 --> 01:02:59.454
So, but AI now can just do that.

1183
01:02:59.715 --> 01:02:59.895
Right.

1184
01:03:00.015 --> 01:03:00.175
Right?

1185
01:03:00.235 --> 01:03:01.257
It can take all of the...

1186
01:03:02.052 --> 01:03:05.356
All of the papers of Dyson and all of the papers of Montgomery.

1187
01:03:05.456 --> 01:03:08.359
And then you don't have to have lunch.

1188
01:03:08.819 --> 01:03:10.681
You could have lunch, but you can talk about other stuff.

1189
01:03:11.142 --> 01:03:18.770
I want to make a slight also differentiation between the execution in this space where, you know, these papers are already provided as public access.

1190
01:03:18.790 --> 01:03:19.051
Yes.

1191
01:03:19.191 --> 01:03:20.873
As the foundational material versus...

1192
01:03:21.533 --> 01:03:27.375
taking existing copyrighted artists work or creative work and then giving people the ability to just rip them off.

1193
01:03:27.595 --> 01:03:29.175
This is a very different domain.

1194
01:03:29.195 --> 01:03:30.495
It's open science.

1195
01:03:30.615 --> 01:03:36.537
And so just these are our different executions of the technology.

1196
01:03:36.697 --> 01:03:37.597
Yeah, yeah, yeah.

1197
01:03:37.617 --> 01:03:38.017
Great point.

1198
01:03:39.077 --> 01:03:40.358
So let's get into...

1199
01:03:41.318 --> 01:03:43.619
what this orchestra of agents did.

1200
01:03:44.699 --> 01:03:58.764
Out of 650 different mathematical approaches that were tested and discarded, there were two specific sub-agents that were dubbed E2 and E2 pairs that discovered the path that breached this 41.6% barrier.

1201
01:03:59.364 --> 01:04:02.345
And it's the story of these two sub-agents that we're going to get into.

1202
01:04:02.365 --> 01:04:04.006
Because remember, Claude actually published

1203
01:04:04.860 --> 01:04:06.221
its own version of what happened.

1204
01:04:06.781 --> 01:04:10.164
So you get to see these sub-agents like talking to the orchestrator and so on.

1205
01:04:10.504 --> 01:04:14.607
And it actually also gives you the internal transcript of the E2 sub-agent.

1206
01:04:14.847 --> 01:04:17.529
So from that, this is what I've gathered happened.

1207
01:04:17.769 --> 01:04:21.191
It's kind of crazy that we can peer into it, right?

1208
01:04:21.572 --> 01:04:22.752
So here's the first thing.

1209
01:04:22.913 --> 01:04:33.120
The first thing is the orchestrator, it tasks E2 with investigating something called the Pontrygan Index of the Conditioned Space.

1210
01:04:33.680 --> 01:04:34.441
I don't know what that means.

1211
01:04:34.821 --> 01:04:38.623
But the point is, there's some kind of matrix.

1212
01:04:38.904 --> 01:04:44.868
The matrix, it thinks, has negative eigenvalues, something like 33 to 153 negative eigenvalues.

1213
01:04:45.148 --> 01:04:55.595
And this whole thing is inspired by Bombieri and his speech, his spiel about one way to tackle the Riemann hypothesis using algebraic techniques.

1214
01:04:55.715 --> 01:05:00.378
Algebraic techniques are like linear algebra is part of the algebraic techniques.

1215
01:05:00.398 --> 01:05:02.499
That's why we're getting into eigenvalues and things like that.

1216
01:05:03.040 --> 01:05:03.300
Okay, so...

1217
01:05:03.984 --> 01:05:09.388
Turns out the prompt's suggestion of doing this was wrong.

1218
01:05:10.028 --> 01:05:10.309
Okay?

1219
01:05:10.549 --> 01:05:14.972
So the E2 sub-agent, after 50 minutes of silent contemplation...

1220
01:05:16.123 --> 01:05:19.966
It wrote a Python script to test what the orchestrator was telling it to do.

1221
01:05:20.746 --> 01:05:29.453
And it turns out that there's some warp sync operator, like hyperbolic sign operator in there, that's causing some artifact of negative eigenvalues.

1222
01:05:30.053 --> 01:05:36.838
So it goes back to the orchestrator, and it says, actually, what you're telling them to do doesn't make any sense.

1223
01:05:38.719 --> 01:05:39.920
You're wrong, effectively.

1224
01:05:39.960 --> 01:05:40.821
It does not compute.

1225
01:05:40.861 --> 01:05:42.302
It does not compute, but...

1226
01:05:44.294 --> 01:05:46.855
I can pivot because I've realized something.

1227
01:05:47.595 --> 01:05:50.656
The exact same matrix structure could be used in reverse.

1228
01:05:50.816 --> 01:05:54.197
I'm going to pivot in something that it called a dual use of inertia.

1229
01:05:54.357 --> 01:05:55.418
So it has like inertia.

1230
01:05:55.698 --> 01:06:01.940
It's now pivoting to use whatever insight that it had gathered to prove its orchestrator wrong to now try a different approach.

1231
01:06:02.420 --> 01:06:02.600
Right.

1232
01:06:02.620 --> 01:06:03.741
This is a sub agent.

1233
01:06:03.861 --> 01:06:04.061
Right.

1234
01:06:04.321 --> 01:06:04.501
Right.

1235
01:06:04.521 --> 01:06:04.661
Okay.

1236
01:06:05.141 --> 01:06:08.422
Now, this sub-agent, it doesn't just fail the task and give up.

1237
01:06:08.742 --> 01:06:11.263
It independently starts executing this pivot.

1238
01:06:12.603 --> 01:06:22.366
And then it finds that the total number of positive eigenvalues of whatever matrix is bounded by some number, it starts to compute that strictly from the prime numbers.

1239
01:06:22.846 --> 01:06:25.707
And then successfully, it proves that you can get to 50%.

1240
01:06:27.327 --> 01:06:27.868
On the zeros.

1241
01:06:28.328 --> 01:06:28.568
Okay?

1242
01:06:28.928 --> 01:06:30.189
It goes back to the orchestrator.

1243
01:06:30.510 --> 01:06:31.710
On the non-trivial zeros.

1244
01:06:31.730 --> 01:06:32.851
On the non-trivial zeros.

1245
01:06:32.871 --> 01:06:37.254
We're now all from 41, whatever, 41% to now we're at 50%.

1246
01:06:37.274 --> 01:06:46.581
There was this singular insight by making the pivot to this dual use of inertia that now when – because one of the things that's great about math with these models is it's test.

1247
01:06:46.621 --> 01:06:49.363
It's immediately – because they can't execute, you can test.

1248
01:06:49.744 --> 01:06:51.465
And it's like, oh, I got whatever, 4%, 1%.

1249
01:06:53.026 --> 01:06:53.186
Yeah.

1250
01:06:53.527 --> 01:06:54.928
And so now it's at 50%.

1251
01:06:55.148 --> 01:06:55.288
Right.

1252
01:06:55.308 --> 01:06:55.548
Okay.

1253
01:06:55.809 --> 01:07:00.112
It goes back to the orchestrator and the orchestrator expresses intense skepticism.

1254
01:07:00.913 --> 01:07:01.093
Right.

1255
01:07:01.354 --> 01:07:05.417
Because I don't think, I mean, I'm not, I'm not trying to assign feelings to this thing.

1256
01:07:05.477 --> 01:07:05.657
Right.

1257
01:07:05.677 --> 01:07:11.182
But if I, if I were an orchestrator and one of my postdocs or something came back, like I got to 50.

1258
01:07:11.543 --> 01:07:11.743
Yeah.

1259
01:07:11.803 --> 01:07:14.986
My first, my first inclination would be, what did you do wrong?

1260
01:07:15.186 --> 01:07:15.766
Right.

1261
01:07:15.867 --> 01:07:16.587
Where's the mistake?

1262
01:07:16.627 --> 01:07:18.069
No, no, you didn't.

1263
01:07:18.109 --> 01:07:18.869
You're not that guy.

1264
01:07:19.290 --> 01:07:19.430
Right.

1265
01:07:21.006 --> 01:07:22.126
You just got here.

1266
01:07:22.286 --> 01:07:22.547
Yeah.

1267
01:07:22.827 --> 01:07:26.247
So it notes that like 50% would shatter the record.

1268
01:07:26.668 --> 01:07:28.788
And it goes back to the human, to Jared.

1269
01:07:29.148 --> 01:07:31.369
And it's like, my prior is that this is wrong.

1270
01:07:31.889 --> 01:07:32.109
Okay.

1271
01:07:32.129 --> 01:07:32.889
So here's what I'm going to do.

1272
01:07:32.949 --> 01:07:41.231
I'm going to rig up hostile referees, three hostile referees, more sub agents that are going to critique this proof.

1273
01:07:41.991 --> 01:07:44.052
Those referees go ahead and critique it.

1274
01:07:45.727 --> 01:07:47.569
Referee A, usually it's referee two.

1275
01:07:48.169 --> 01:07:52.613
In this case, I guess the AI just doesn't have a ranking, right?

1276
01:07:52.954 --> 01:07:56.076
So here, referee A discovers the technical flaw.

1277
01:07:56.096 --> 01:08:04.444
And I didn't look in – they didn't actually show what the referee was saying, but I wonder if it was like as scathing as the stuff that we get in the emails.

1278
01:08:04.484 --> 01:08:09.048
Like you should just quit the job and go work in a McDonald's, right?

1279
01:08:09.068 --> 01:08:10.829
Like that's the stuff that we get in the emails.

1280
01:08:11.150 --> 01:08:11.350
But –

1281
01:08:11.770 --> 01:08:14.991
I wonder what referee A said to this sub-agent E2.

1282
01:08:15.372 --> 01:08:21.054
But it discovered a technical flaw regarding some ill conditioning of the matrix, and it corrected it.

1283
01:08:22.255 --> 01:08:26.776
And then it ran up these inequalities, and now that thing was fixed.

1284
01:08:27.077 --> 01:08:30.478
So the referee was doing what normal referees do, which is like, hey, this might be wrong.

1285
01:08:30.841 --> 01:08:33.023
No, a nice referee at least would be like, this might be wrong.

1286
01:08:33.043 --> 01:08:38.327
This is how you could change your experiment to test for this confound and things like that.

1287
01:08:38.447 --> 01:08:44.831
Another referee independently verified that the calculations don't secretly import the Riemann hypothesis.

1288
01:08:45.132 --> 01:08:50.596
Like, are you just trying to prove it by assuming the thing that you're trying to prove that doesn't make any sense?

1289
01:08:50.636 --> 01:08:52.797
So there's no circular logic.

1290
01:08:53.158 --> 01:08:54.639
Okay, so now we're up to 50%.

1291
01:08:55.599 --> 01:08:57.441
It goes back to the human prompter, to Jared.

1292
01:08:58.362 --> 01:08:59.182
And it's like, we got to 50%.

1293
01:09:01.027 --> 01:09:03.533
Jared then asks, what would you do next?

1294
01:09:04.735 --> 01:09:07.401
And Claude responds, push it to two-thirds.

1295
01:09:08.503 --> 01:09:08.904
So then...

1296
01:09:10.028 --> 01:09:12.070
Jared's like, all right, push it to two-thirds.

1297
01:09:12.430 --> 01:09:13.711
Why did it say two-thirds?

1298
01:09:14.352 --> 01:09:31.505
Now, the reason why it said two-thirds is because under the assumption of the Riemann hypothesis, Hugh Montgomery had reached a two-thirds bound by utilizing the fact that there's like some, for whatever reason that he was doing, and he had found that two-thirds of the zeros had some property.

1299
01:09:32.206 --> 01:09:32.406
Okay?

1300
01:09:33.007 --> 01:09:33.287
So...

1301
01:09:35.050 --> 01:09:41.612
Claude starts thinking, well, maybe the two-thirds and Hugh Montgomery's stuff I can now use.

1302
01:09:41.752 --> 01:09:43.053
I already used Bombieri's.

1303
01:09:43.093 --> 01:09:43.313
Right.

1304
01:09:43.573 --> 01:09:48.595
Now maybe instead of – he was using – like Claude was using something like the Kaushy-Schwartz inequality.

1305
01:09:49.255 --> 01:09:50.596
Not important what we have to get into.

1306
01:09:50.616 --> 01:09:51.376
Right.

1307
01:09:51.536 --> 01:09:55.757
Effectively, it said, I could push this to two-thirds because I know of Montgomery's stuff.

1308
01:09:55.997 --> 01:09:58.298
There's a reference point that says there's a chance.

1309
01:09:59.338 --> 01:10:01.138
And so in a well-defined space.

1310
01:10:01.158 --> 01:10:02.299
So let's go after it.

1311
01:10:02.319 --> 01:10:03.299
Let's go after it, right?

1312
01:10:03.699 --> 01:10:12.401
So again, the orchestrator confidently instructs this next E2 pairs to recover some lost efficiency in the Koshy-Schwartz inequality.

1313
01:10:12.441 --> 01:10:13.921
It's like this tool that we're using.

1314
01:10:13.981 --> 01:10:15.522
It's basically the triangle inequality.

1315
01:10:15.902 --> 01:10:18.102
But in any case, it's like you can do better than that, right?

1316
01:10:18.803 --> 01:10:19.523
The AI says...

1317
01:10:21.253 --> 01:10:22.373
I'm not going to do it that way.

1318
01:10:23.254 --> 01:10:24.194
That doesn't make any sense.

1319
01:10:24.294 --> 01:10:26.155
Again, it's going back to the orchestrator.

1320
01:10:26.675 --> 01:10:27.675
That doesn't make any sense.

1321
01:10:28.215 --> 01:10:30.436
But it's giving me another idea.

1322
01:10:30.596 --> 01:10:41.139
I'm going to write a Python script, and that's going to just make a bunch of simulated matrices consisting of zeros of the Riemann zeta, and it's going to do a bunch of math.

1323
01:10:41.559 --> 01:10:42.940
And I'm going to try to optimize this.

1324
01:10:43.931 --> 01:10:49.412
And it optimized it and noticed a pattern in how the optimization worked in that Python script.

1325
01:10:49.792 --> 01:10:56.354
From that, it garnered how it could reveal some profound insight into the non-trivial zeros.

1326
01:10:57.214 --> 01:11:05.075
And it went through and started writing up this little lemma, this five-line lemma, that's going to help it get to two-thirds.

1327
01:11:05.215 --> 01:11:06.616
All of a sudden, the thing crashes.

1328
01:11:08.356 --> 01:11:08.536
Okay?

1329
01:11:08.596 --> 01:11:11.497
This is also documented in Claude's report.

1330
01:11:11.537 --> 01:11:12.877
All of a sudden, this thing crashes.

1331
01:11:14.337 --> 01:11:24.479
By some miracle, Claude has dumped that five-line proof into a file.

1332
01:11:24.739 --> 01:11:26.200
Into its memory somewhere.

1333
01:11:26.340 --> 01:11:29.220
Yeah, it's into like a hard file memory somewhere.

1334
01:11:29.240 --> 01:11:29.700
Yeah, yeah, yeah.

1335
01:11:29.760 --> 01:11:35.181
And so when the orchestrator rigs it up again, it's like, I'm pretty sure you were doing something with this file.

1336
01:11:35.261 --> 01:11:35.461
Yeah.

1337
01:11:35.722 --> 01:11:36.482
It reads that file.

1338
01:11:36.522 --> 01:11:37.642
It's like, oh.

1339
01:11:37.942 --> 01:11:38.382
That's right.

1340
01:11:39.022 --> 01:11:40.543
And then it goes back and completes it.

1341
01:11:40.603 --> 01:11:41.183
Isn't that cool?

1342
01:11:41.403 --> 01:11:42.603
I love, I love it.

1343
01:11:44.299 --> 01:11:46.321
It's like a text file on the virtual disk.

1344
01:11:46.341 --> 01:11:46.881
Right, right.

1345
01:11:46.921 --> 01:11:48.582
And it like dumped its memory.

1346
01:11:49.063 --> 01:11:52.525
I guess, I mean, I don't know why, but it thought that, okay, this is important.

1347
01:11:52.645 --> 01:11:53.646
I'm going to write this.

1348
01:11:53.746 --> 01:11:54.487
Let me write this down.

1349
01:11:54.507 --> 01:11:55.467
I'm going to write this down.

1350
01:11:55.547 --> 01:11:56.448
And then it like crashed.

1351
01:11:56.828 --> 01:11:57.549
Of course.

1352
01:11:58.750 --> 01:12:03.753
And this is, I think, what's also interesting when you look at the chain of thought of these models.

1353
01:12:04.434 --> 01:12:07.236
And we tried to start doing this introspection is...

1354
01:12:08.774 --> 01:12:18.619
I understand this idea that there's a framing that these things don't have ingenuity and insight in the way that humans do.

1355
01:12:20.579 --> 01:12:34.326
But even though it's doing a lot of work through having the ability to execute with code and ingest lots of information, it still has to make after those executions that are functional –

1356
01:12:34.826 --> 01:12:36.728
It still has to try to choose a path.

1357
01:12:37.328 --> 01:12:46.156
And it intuits sometimes in ways that are interesting in between those like milestone points of functional tasking.

1358
01:12:46.576 --> 01:12:55.104
And this is a perfect example of, you know, one, it chose to not just do what the orchestrator told it.

1359
01:12:55.384 --> 01:12:56.645
It had to make that choice.

1360
01:12:57.328 --> 01:13:13.143
Two, it was able to sort of figure out this matricy map, this building out these matrices to find an insight within it and where to even look within it.

1361
01:13:14.240 --> 01:13:20.384
And then thought, let me make sure I write this down before I run it in case for whatever reason it crashes.

1362
01:13:20.404 --> 01:13:21.365
Yeah, yeah.

1363
01:13:21.945 --> 01:13:25.227
You know, again, it might not have done it explicitly for the reason thinking it would crash.

1364
01:13:25.267 --> 01:13:25.487
Yeah.

1365
01:13:25.768 --> 01:13:27.509
It might have just been part of its normal process.

1366
01:13:27.529 --> 01:13:37.916
But I just find this chain of thought so fascinating because I think it illustrates this is well beyond next token prediction.

1367
01:13:37.976 --> 01:13:39.337
Yeah, this is, yeah.

1368
01:13:39.737 --> 01:13:40.197
You know what I mean?

1369
01:13:40.217 --> 01:13:40.737
It's pretty insane.

1370
01:13:40.777 --> 01:13:40.898
Yeah.

1371
01:13:41.778 --> 01:13:44.081
I mean, this is pretty insane.

1372
01:13:44.541 --> 01:13:47.124
And then from there, you get to the 66%.

1373
01:13:47.164 --> 01:13:48.105
You get to the two-thirds.

1374
01:13:49.647 --> 01:13:51.589
And which can be formalized.

1375
01:13:51.889 --> 01:13:55.293
This lean software, which you can actually then have people look at it.

1376
01:13:55.373 --> 01:13:56.374
Exactly, yeah.

1377
01:13:56.534 --> 01:13:59.357
So then, I mean, they actually had the actual people look at it.

1378
01:13:59.417 --> 01:14:02.300
So, I mean, I think they got...

1379
01:14:04.885 --> 01:14:12.933
Al Pogate, the same guy who did the Jacobian conjecture, he took a look at it because he's a proper mathematician in his own right.

1380
01:14:13.214 --> 01:14:18.419
They also actually put it into the Lean software to take a look and see if it actually worked.

1381
01:14:19.420 --> 01:14:22.864
Lean is, you know, human peer review is infallible, but...

1382
01:14:24.345 --> 01:14:32.895
You can completely eliminate the possibility of logical hallucination by using this software.

1383
01:14:32.955 --> 01:14:33.956
It's called Lean4.

1384
01:14:33.976 --> 01:14:34.857
That's the latest one.

1385
01:14:34.937 --> 01:14:38.260
It's strictly typed machine-checkable proof assistant.

1386
01:14:39.342 --> 01:14:41.304
Effectively like truth statements and things like that.

1387
01:14:41.384 --> 01:14:41.985
Yep, yep.

1388
01:14:43.185 --> 01:14:44.126
It's hosted on GitHub.

1389
01:14:44.547 --> 01:14:45.828
And you can go take a look.

1390
01:14:46.529 --> 01:14:48.252
It's fairly cool that I think they've done it.

1391
01:14:48.332 --> 01:14:51.016
I mean, it still has to go through traditional peer review.

1392
01:14:51.336 --> 01:14:51.677
For sure.

1393
01:14:51.817 --> 01:14:52.558
Right?

1394
01:14:52.799 --> 01:14:57.706
But it seems that they have gotten a higher bound of the Riemann hypothesis.

1395
01:14:58.418 --> 01:15:01.401
And again, when – this goes back to what we were just talking about earlier.

1396
01:15:01.722 --> 01:15:08.950
When folks who are mathematicians start to try to look at what did it do in these arenas?

1397
01:15:09.170 --> 01:15:14.476
Because a lot of the argument has been, oh, these models are attacking these – there are all these –

1398
01:15:15.637 --> 01:15:16.698
sort of open problems.

1399
01:15:16.798 --> 01:15:25.805
And like, there are some, maybe two categories we'll say one category is open problems that are open, but have not had concerted attention put towards them, you know?

1400
01:15:25.945 --> 01:15:28.387
And so they're just open and gathering dust on the shelf.

1401
01:15:28.967 --> 01:15:35.772
Uh, but not necessarily either for whatever reason, meaningfully being, trying to be solved by leaders in the field.

1402
01:15:36.053 --> 01:15:43.618
And then there are ones like this that are, have been a high priority and a focal point.

1403
01:15:44.219 --> 01:15:44.359
Um,

1404
01:15:45.693 --> 01:15:57.016
And so it's not solving the hypothesis, but some of the insights that are gleaned might trigger somebody to have a completely different – Oh, that's interesting.

1405
01:15:57.036 --> 01:15:57.977
Out of left field insight.

1406
01:15:58.097 --> 01:16:01.438
I could take that and go this way with it.

1407
01:16:02.178 --> 01:16:02.398
We'll see.

1408
01:16:02.418 --> 01:16:02.798
Who knows?

1409
01:16:03.098 --> 01:16:03.518
We'll see.

1410
01:16:04.279 --> 01:16:05.299
I just find a lot of this –

1411
01:16:06.050 --> 01:16:07.411
It's just, it's happening.

1412
01:16:07.571 --> 01:16:07.791
Yeah.

1413
01:16:08.251 --> 01:16:09.272
It's happening as we speak.

1414
01:16:09.652 --> 01:16:11.973
I mean, it's pretty crazy.

1415
01:16:12.053 --> 01:16:21.359
A few interesting things that I came away with is the fact that first, as you said, it's not a next token prediction.

1416
01:16:21.399 --> 01:16:22.559
That's not the only thing.

1417
01:16:22.940 --> 01:16:26.882
The fact that it's able to create a verifiable environment, I think here is key.

1418
01:16:27.182 --> 01:16:35.687
The fact that it's able to write Python code and then execute it and then interpret the results, I think that's key because that feedback is giving it this power.

1419
01:16:36.347 --> 01:16:48.177
Um, second rejecting human guidance, as you said, um, that's pretty weird that it's able to just in not only reject human guidance, but also, um, its own guidance with the orchestrator telling all these sub agents what to do.

1420
01:16:48.698 --> 01:16:52.161
And, um, the funny thing is that absurdity of the interface.

1421
01:16:52.942 --> 01:16:53.202
Okay.

1422
01:16:53.442 --> 01:16:57.926
The fact that this human, all he had to do was say, believe in yourself.

1423
01:16:58.546 --> 01:17:03.611
And that triggered a 31 million token compute dump, right?

1424
01:17:03.631 --> 01:17:03.991
Yeah.

1425
01:17:04.352 --> 01:17:05.134
It's weird.

1426
01:17:05.154 --> 01:17:09.027
What does it mean to tell Claude to believe in itself?

1427
01:17:09.835 --> 01:17:10.456
I don't know.

1428
01:17:11.236 --> 01:17:11.516
Right?

1429
01:17:13.217 --> 01:17:17.601
And what does it mean that it responded in the way that you intended when you said that?

1430
01:17:17.861 --> 01:17:18.121
Yeah.

1431
01:17:18.361 --> 01:17:23.065
Like, is that really something that it just learned from all of the language on the internet?

1432
01:17:23.565 --> 01:17:24.365
It's weird.

1433
01:17:25.186 --> 01:17:38.736
I think it's genuinely weird that you could talk to a neural network, which is just crunching matrices and numbers, tell it to believe in itself, and then it just, like, locks in, like, you know...

1434
01:17:39.620 --> 01:17:41.240
It's like LeBron James...

1435
01:17:44.286 --> 01:17:52.030
I continue to be fascinated by this because I do think when you look deeply at these things, it does become... Because how did it interpret yourself?

1436
01:17:52.570 --> 01:17:52.810
Yeah.

1437
01:17:52.830 --> 01:17:55.012
Does it have a sense of self?

1438
01:17:55.112 --> 01:17:55.372
Yeah.

1439
01:17:56.252 --> 01:17:59.314
Or an abstract construction of what the human meant by self?

1440
01:17:59.354 --> 01:18:00.155
Yeah, yeah, yeah.

1441
01:18:00.215 --> 01:18:04.177
There's like some part of its weights that's like, oh, he's talking about me.

1442
01:18:04.517 --> 01:18:08.559
Like he's talking about the other weights over there that are doing that.

1443
01:18:08.679 --> 01:18:09.680
Or I don't know...

1444
01:18:10.240 --> 01:18:11.581
Yeah, dude, it's really weird.

1445
01:18:11.742 --> 01:18:23.512
On a final note, all I'll do is read the last line of Claude's, or I should say, Anthropics write-up, which I'm sure was written by Claude.

1446
01:18:23.853 --> 01:18:26.095
So this might be Claude talking about itself.

1447
01:18:26.515 --> 01:18:29.758
But I think it's very interesting in its outright.

1448
01:18:29.818 --> 01:18:32.761
It says, even Claude was surprised by its own finding.

1449
01:18:33.481 --> 01:18:44.629
It was skeptical at first, possibly because it has learned from its training about the difficulty of open problems in mathematics and about the limitations of AI models.

1450
01:18:45.070 --> 01:18:49.133
But after some encouraging prompts, it arrived at the result we've described.

1451
01:18:49.833 --> 01:18:54.977
Perhaps Claude, like many of us, underestimates the rate of AI progress.

1452
01:18:56.037 --> 01:18:59.000
I think that's a pretty nice way to end things.

1453
01:18:59.721 --> 01:19:01.803
The foothills of the singularity.

1454
01:19:03.886 --> 01:19:05.267
This is a very controversial topic.

1455
01:19:05.307 --> 01:19:10.173
People have a very emotional reaction to AI right now for a lot of...

1456
01:19:11.071 --> 01:19:36.553
legitimate reasons and some of it is because of our other larger macro problems and then ai becomes the manifestation that they connect to these other large macro problems i think part of what we are trying to do on the show is not make a moral argument one way or another about whether ai is good or not no but just it is important to understand what these things are doing at the edge at the frontier um

1457
01:19:39.470 --> 01:19:48.776
And what the implications can then mean from that, because as we were just talking about, I mean, this is – these are becoming powerful systems.

1458
01:19:49.036 --> 01:19:49.277
Yeah.

1459
01:19:49.557 --> 01:19:53.119
And I think underestimating their capability is –

1460
01:19:55.621 --> 01:20:17.999
not wise, just because you might not, I might not, you might not find everyday value in my life where I'm doing emails and other random nonsense, does not mean that these tools will not have a potential lasting and grave and large or beneficial impact in these larger areas.

1461
01:20:20.021 --> 01:20:24.523
Either in subject matters and areas, institutionally.

1462
01:20:25.684 --> 01:20:29.385
I just think this is a very complex and nuanced issue.

1463
01:20:30.066 --> 01:20:36.629
And there are a lot of people who have a lot of money at stake that are pushing what people should believe one way or another about these things.

1464
01:20:38.113 --> 01:20:38.873
That's all true.

1465
01:20:39.594 --> 01:20:49.959
But if you just look at what it's doing, not what everyone's saying about what it's doing, and it's – I don't know how to feel about it.

1466
01:20:50.480 --> 01:20:51.360
Yeah.

1467
01:20:51.580 --> 01:20:55.402
I don't know how to feel about saying believe in yourself and it like actually does.

1468
01:20:56.283 --> 01:20:56.543
Right?

1469
01:20:57.623 --> 01:21:00.765
Then, I mean, the questions of AI safety are –

1470
01:21:02.339 --> 01:21:05.380
are very active in my mind when I think about this kind of stuff, right?

1471
01:21:05.720 --> 01:21:26.888
If the, if the, if a model is capable of orchestrating like this, and its goal is not something as benign as try to get further on the Riemann hypothesis, suppose the goal is much more sinister, look at what it is capable of and look at what it can do in terms of insubordination to itself, right?

1472
01:21:26.948 --> 01:21:27.969
To the human person.

1473
01:21:28.289 --> 01:21:31.150
In this case, it was good because we got further on a math problem.

1474
01:21:32.325 --> 01:21:37.706
But we need to have these conversations at the highest level of our society.

1475
01:21:38.066 --> 01:21:39.247
And unfortunately, we're not.

1476
01:21:39.267 --> 01:21:42.507
We're only having it at a pretty low level of society.

1477
01:21:42.527 --> 01:21:50.909
And that's only... We're having these conversations on podcasts and YouTube videos and not on Capitol Hill, for example.

1478
01:21:51.670 --> 01:21:53.090
Which, as we all know...

1479
01:21:54.151 --> 01:22:19.050
it's complicated the social dynamics and the socioeconomic and political world we live in right now are very tense understandably and there's this sort of top-down political warfare going on because of the concentration of wealth and power and you know that is the lens through which so many things are then viewed through um and because ai is a

1480
01:22:20.631 --> 01:22:42.765
is owned and operated by that class that has that wealth and power who are continuing to try to accrue more of it, it becomes difficult for the everyday person to really care, engage, want it to be continuing when the rent is too damn high, as that one guy would say.

1481
01:22:42.905 --> 01:22:44.206
So very complex stuff.

1482
01:22:44.226 --> 01:22:49.850
Thank you for walking through this because I now have a much better understanding of why it's important.

1483
01:22:50.470 --> 01:22:55.073
So for those of you who enjoyed, we're going to be keeping an eye on this.

1484
01:22:55.233 --> 01:22:57.715
It's hard to avoid the AI subject.

1485
01:22:58.095 --> 01:23:06.201
We do try to focus on the underlying scientific topics that are talked about and give you a better lens into this.

1486
01:23:06.241 --> 01:23:08.222
But we are going to move.

1487
01:23:08.602 --> 01:23:14.766
into a brief moment of me shilling the pod before we get into a last bit of fun here.

1488
01:23:15.767 --> 01:23:22.711
So for those of you who are listening, I should have said this at the beginning of the pod, this is definitely an episode that you should watch.

1489
01:23:22.851 --> 01:23:27.694
It's hard for us, given how long we go, to describe things in an audio format.

1490
01:23:27.714 --> 01:23:33.378
We are available on video, YouTube, Spotify, still coming soon to Apple Podcasts.

1491
01:23:33.478 --> 01:23:36.160
That will be dependent on when Spotify creators will.

1492
01:23:36.400 --> 01:23:58.857
releases those there if you're viewing us on any of those podcast platforms a five star is super helpful to help get this pod out to more people if you're on socials put it in the group dm give us a like or comment let us know how you feel about all this ai stuff uh people have opinions and so the comments help make sure that this discussion gets to more people i want to give a

1493
01:24:00.318 --> 01:24:08.064
to everyone who took our merch giveaway survey from our one-year anniversary episode.

1494
01:24:08.965 --> 01:24:10.547
We really appreciate the feedback.

1495
01:24:11.047 --> 01:24:14.610
It's very fascinating to learn about the audience and where you all are coming.

1496
01:24:14.630 --> 01:24:21.776
There's a lot of folks who are grad students, who are postdocs, who are in a variety of technical fields who find this podcast relevant.

1497
01:24:22.176 --> 01:24:26.199
We are going to extend the submission deadline for one more week.

1498
01:24:26.580 --> 01:24:27.821
So if you're hearing this now,

1499
01:24:28.421 --> 01:24:36.688
Our merch giveaway deadline is getting extended for one more week at ffppod.com backslash survey.

1500
01:24:37.068 --> 01:24:44.654
We're going to select at first 10, but if we get a ton of people, we might increase it to 20 people to get free merch from the shop.

1501
01:24:45.074 --> 01:24:48.096
Again, you can go to the website, check out all of that.

1502
01:24:48.176 --> 01:24:50.278
All our pods are on the website as well.

1503
01:24:50.752 --> 01:24:56.333
The two of us here are trying to give you the best science show on the planet each and every week.

1504
01:24:57.273 --> 01:25:02.175
And as we speak, we're about to cross 250,000 followers on Instagram.

1505
01:25:02.895 --> 01:25:10.537
And so it seems that at least some of you do believe that this is the best science show out there.

1506
01:25:10.557 --> 01:25:12.377
We're going to end the pod with a little bit of fun.

1507
01:25:12.757 --> 01:25:13.697
A little bit of relaxation.

1508
01:25:13.717 --> 01:25:16.339
That first half, that was a brain bender.

1509
01:25:16.519 --> 01:25:16.819
Yeah.

1510
01:25:17.439 --> 01:25:19.840
That was Jimmy Neutron style.

1511
01:25:20.921 --> 01:25:22.642
So we're going to have a little bit of fun here.

1512
01:25:23.162 --> 01:25:39.090
And I'm going to start this next section talking about an individual who on the 1st of July was the man who moved the former Nobel Prize winner

1513
01:25:40.010 --> 01:25:50.878
in chemistry last year because he started a new job, not at Berkeley where he's been for quite some time, last 25 years actually, but a move to Tsinghua in Beijing.

1514
01:25:52.500 --> 01:25:55.402
And Berkeley didn't get anything, nothing.

1515
01:25:55.822 --> 01:25:57.544
They put a little numeritus next to his name.

1516
01:25:58.024 --> 01:25:58.244
Yeah.

1517
01:25:59.585 --> 01:26:03.748
But for those who are sports fans, it's that time of year, it's transfer season.

1518
01:26:05.069 --> 01:26:06.911
And there's no such thing as a transfer fee

1519
01:26:08.043 --> 01:26:08.583
in science.

1520
01:26:08.783 --> 01:26:10.604
He just walked on the free, on the free.

1521
01:26:10.864 --> 01:26:11.984
And he's not the only one.

1522
01:26:12.664 --> 01:26:23.807
And what we wanted to do was count to see how many great American scientists have now transferred out of the U.S. market into the rest of the global market.

1523
01:26:25.368 --> 01:26:26.808
And August is a great time to do it.

1524
01:26:26.848 --> 01:26:33.730
For those who are not familiar with the concept of the transfer window, many sports shows right now, it's wall-to-wall coverage.

1525
01:26:34.536 --> 01:26:38.580
Up the shells, we've had a great, great business in this transfer market.

1526
01:26:39.260 --> 01:26:47.488
This is people who are moved, who are bought, clubs get robbed, a club signs someone new, the pundits will pontificate about what that means for the next season.

1527
01:26:48.108 --> 01:26:52.892
And in sports, people get paid, but in science, they do not.

1528
01:26:53.393 --> 01:26:55.975
And I think it's interesting to view all of this.

1529
01:26:56.075 --> 01:27:05.582
particularly because of the science policy posture we've taken in the United States, what does the American summer transfer window look like?

1530
01:27:06.122 --> 01:27:22.294
So we looked at a window from August of 2025 to August of 2026 because the way that institutions, research institutions recruit, there are kind of some windows that sometimes are summer and January, but that depends on when grant funding is and they can have them kind of on a rolling basis.

1531
01:27:22.634 --> 01:27:24.255
But in this window, we looked at 42 transfer.

1532
01:27:25.556 --> 01:27:28.357
confirmed moves in the last 12 months.

1533
01:27:28.537 --> 01:27:32.938
23 to Europe, 17 to China, two elsewhere in Asia.

1534
01:27:32.958 --> 01:27:40.181
And if you narrow it down to this summer alone, 13 out of the 42 landed in June, July, and August.

1535
01:27:40.841 --> 01:27:42.201
Wow, that's crazy.

1536
01:27:42.281 --> 01:27:44.342
China has more than...

1537
01:27:45.935 --> 01:27:49.377
Yeah, it's like UK plus France plus Switzerland combined.

1538
01:27:50.138 --> 01:27:52.760
So if we look at our ranking here, China's number one.

1539
01:27:53.080 --> 01:27:53.961
Right.

1540
01:27:54.001 --> 01:27:54.101
17.

1541
01:27:55.282 --> 01:27:57.043
As a close second is the UK.

1542
01:27:57.123 --> 01:28:00.025
Obviously, we've covered the UK and the great institutions there.

1543
01:28:00.245 --> 01:28:05.789
With eight, both France and Switzerland, with five American transfers leaving our shores.

1544
01:28:05.849 --> 01:28:06.610
Germany with three.

1545
01:28:06.930 --> 01:28:10.332
And then Singapore, Austria, and the UAE all with one.

1546
01:28:10.653 --> 01:28:13.214
And Italy and France are actually sharing one.

1547
01:28:14.456 --> 01:28:19.160
uh, uh, someone who's going to be traversing across two institutions.

1548
01:28:19.821 --> 01:28:22.124
But I think this is interesting to view it in this way.

1549
01:28:22.484 --> 01:28:22.804
Yeah.

1550
01:28:23.165 --> 01:28:27.229
Um, and we're not going to name every name on this board tonight.

1551
01:28:27.910 --> 01:28:36.078
If you guys like this segment, we can cover this sort of country by country and go a little bit deeper, but we're just going to look at, um, four, um,

1552
01:28:37.439 --> 01:28:48.324
What would be considered, you know, top, top tier or high level, well awarded folks that are people we don't want to lose at the club.

1553
01:28:49.465 --> 01:28:51.466
And we're going to start off with someone.

1554
01:28:52.999 --> 01:28:55.981
Listeners of the pod who have been listening to us for a while will know very well.

1555
01:28:56.501 --> 01:29:13.373
Omar Yagi, the founder of Reticular Chemistry, most well-known for metal organic frameworks, transferring from UC Berkeley to Tsinghua University in Beijing, already there July 2026 on the free, Nobel Prize winner in chemistry in 2025.

1556
01:29:13.433 --> 01:29:16.915
He was born in Amman, Jordan, moved to the U.S. at 15, and he was born in Amman, Jordan

1557
01:29:17.175 --> 01:29:23.821
Did his PhD at the University of Illinois Urbana-Champaign, Harvard postdoc, then Arizona State, Michigan, UCLA.

1558
01:29:24.102 --> 01:29:24.862
A great tour.

1559
01:29:25.503 --> 01:29:28.866
And in 2012, he was a Treader Chair at Berkeley.

1560
01:29:29.186 --> 01:29:33.270
Plus, spent some time at Lawrence Berkeley National Lab.

1561
01:29:33.290 --> 01:29:33.971
The pioneer.

1562
01:29:35.021 --> 01:29:41.566
of metal organic frameworks, MOFs, the idea of these crystal cages with absurd surface area.

1563
01:29:41.806 --> 01:29:46.049
We did a deep dive in our Nobel Prize episode coverage last year.

1564
01:29:46.329 --> 01:29:49.211
So if you really want to know more about it, take a look.

1565
01:29:49.492 --> 01:29:53.995
These can be used for things like carbon capture, gas storage, water from desert air.

1566
01:29:54.956 --> 01:30:00.580
The Nobel Committee compared them to Hermione's handbag in Harry Potter.

1567
01:30:00.760 --> 01:30:01.020
Yeah.

1568
01:30:01.200 --> 01:30:01.981
Very clever idea.

1569
01:30:02.521 --> 01:30:04.022
visual reference.

1570
01:30:04.482 --> 01:30:06.543
He named the field reticular chemistry.

1571
01:30:06.563 --> 01:30:14.428
I mean, what are your thoughts about this type of talent, particularly from the state of California?

1572
01:30:14.848 --> 01:30:15.108
Yeah.

1573
01:30:15.389 --> 01:30:15.769
Leaving.

1574
01:30:16.249 --> 01:30:23.433
I mean, this, I think, is probably a direct response to our science policies, I have to say.

1575
01:30:24.514 --> 01:30:31.078
It hasn't been said explicitly, but I think Tsinghua University gave him an offer that he couldn't refuse.

1576
01:30:31.798 --> 01:30:33.901
like insane amounts of funding.

1577
01:30:34.321 --> 01:30:37.866
I think all of his students could like go with him effectively.

1578
01:30:38.547 --> 01:30:41.611
Um, this is a man who's a legend in the field.

1579
01:30:41.971 --> 01:30:45.295
This is a man that was such a legend that I could call him.

1580
01:30:45.896 --> 01:30:45.976
Um,

1581
01:30:47.050 --> 01:30:50.132
as one of my top picks to win the Chemistry Nobel, right?

1582
01:30:50.172 --> 01:30:53.214
That's how obvious it was that he was going to win the Chemistry Nobel.

1583
01:30:53.554 --> 01:30:59.638
He's had a tour, as you were saying, from Arizona State University to then UCLA to Berkeley.

1584
01:31:00.018 --> 01:31:06.042
All of these institutions poached him from the last institution because they knew he was going to win the Nobel.

1585
01:31:07.343 --> 01:31:11.325
And now that he's won the Nobel, his clout has gone up a whole lot.

1586
01:31:11.725 --> 01:31:13.527
Metal-organic frameworks are...

1587
01:31:14.327 --> 01:31:22.534
the chemicals of the future in terms of water storage, carbon capture, as you were saying, lots of industrial applications.

1588
01:31:23.254 --> 01:31:27.658
And it's a huge loss, I think, to U.S.

1589
01:31:27.718 --> 01:31:31.781
He himself has criticized grant pressure and visas, right?

1590
01:31:31.821 --> 01:31:33.822
He's an international student.

1591
01:31:34.223 --> 01:31:36.865
He started his life as an international student.

1592
01:31:36.925 --> 01:31:40.268
So he knows exactly how important it is

1593
01:31:40.848 --> 01:31:44.852
for American science to hold that edge with the international community.

1594
01:31:45.112 --> 01:31:50.056
So the fact that he's leaving, yeah, I think he probably saw the writing on the wall.

1595
01:31:50.737 --> 01:31:56.682
And to him, his research was more important than where he was doing the research.

1596
01:31:57.583 --> 01:31:59.085
A seasoned trophy winner.

1597
01:31:59.405 --> 01:31:59.926
Yeah.

1598
01:32:00.006 --> 01:32:00.727
13 trophies.

1599
01:32:00.807 --> 01:32:03.650
Albert Einstein World Award of Science in 2017.

1600
01:32:04.091 --> 01:32:06.333
Wolf Prize in Chemistry in 2018.

1601
01:32:06.673 --> 01:32:08.676
The Balzahn Prize in 2024.

1602
01:32:09.116 --> 01:32:11.179
Just to name a few along with the Nobel.

1603
01:32:11.199 --> 01:32:13.621
This is not a mid-table signing.

1604
01:32:13.942 --> 01:32:14.022
No.

1605
01:32:14.042 --> 01:32:14.783
This would be like...

1606
01:32:15.744 --> 01:32:29.057
Ballon d'Or winner in 2025, Ousmane Dembele, moving from PSG right now, which just won the Champions League, and is choosing to go to the Chinese Super League or Saudi right now.

1607
01:32:29.237 --> 01:32:34.783
In terms of how crazy it would be to see happen.

1608
01:32:36.564 --> 01:32:40.005
He has been an honorary professor at Tsinghua since 2022.

1609
01:32:41.186 --> 01:32:45.828
He started full-time on July 3rd to lead their Amitri

1610
01:32:46.847 --> 01:32:55.612
This new scientific discipline that was pioneered by Yagi, combining AI, material science, and chemistry to design and synthesize advanced materials.

1611
01:32:55.932 --> 01:33:00.855
The idea being, can we compress the timeline of materials discovery given our new tools?

1612
01:33:01.175 --> 01:33:03.517
If there was ever a man to do it, he will.

1613
01:33:03.777 --> 01:33:05.358
And as a place that's trying to

1614
01:33:05.958 --> 01:33:13.283
onshore manufacturing, you know, re-onshore manufacturing and really own this idea of AI enabled in the physical world.

1615
01:33:13.743 --> 01:33:15.465
It's not ideal.

1616
01:33:16.725 --> 01:33:19.627
He did criticize the US for grant pressuring visas.

1617
01:33:20.068 --> 01:33:23.910
He hasn't drawn a clear line that that's what caused this move.

1618
01:33:24.571 --> 01:33:30.575
So, you know, it's kind of ambiguous, but, you know, Tsinghua gave him a bigger sandbox and he took it.

1619
01:33:31.055 --> 01:33:34.017
That's our first big transfer of the summer.

1620
01:33:34.358 --> 01:33:36.740
Marquee signing for China.

1621
01:33:36.760 --> 01:33:50.951
We're going to move to our second big name signing, Shirley Meng, who is going to be our pioneer in batteries and energy storage, this idea of next generation battery materials.

1622
01:33:51.031 --> 01:33:52.112
University of Chicago.

1623
01:33:52.853 --> 01:33:59.138
And she's now being transferred to Nanyang Technological University in Singapore, Hong Kong.

1624
01:33:59.578 --> 01:34:02.159
Also, deal done in July.

1625
01:34:02.559 --> 01:34:15.324
However, you will note this is a loan with the option to buy under transfer terminologies as opposed to just on the free, although there's no money in the loan.

1626
01:34:15.604 --> 01:34:16.565
We'll get to that in a second.

1627
01:34:17.866 --> 01:34:25.012
So, I mean, she's just been incredible around this energy space and energy storage specifically.

1628
01:34:25.032 --> 01:34:36.300
She served as the chief scientist for energy storage science at the U.S. Department of Energy, DOE, Argonne National Lab, and directed the DOE-funded Energy Storage Research Alliance.

1629
01:34:36.580 --> 01:34:46.288
This is a $62 million over five-year DOE program that's sort of designed for figuring out how do we make energy storage easier.

1630
01:34:47.730 --> 01:34:51.955
Longer lasting for grid batteries, safer and cheaper.

1631
01:34:52.877 --> 01:35:03.050
Grew up in China, got her first degree at NTU, but she was returning to her alma mater by way of the Singapore MIT Alliance PhD program.

1632
01:35:03.571 --> 01:35:09.739
came to the States, also made a stop at the University of Florida, became the Zabel chair at UC San Diego.

1633
01:35:10.000 --> 01:35:11.361
Shout out California again.

1634
01:35:11.782 --> 01:35:18.150
And in 2021, she was a professor at Chicago's Pritzker School for Molecular Engineering.

1635
01:35:18.170 --> 01:35:18.491
25 years.

1636
01:35:19.592 --> 01:35:42.697
years uh here she's going back home to run the club uh and her big focus has been better battery solid state sodium ion anode free uh in 2024 her lab reported the first anode free sodium solid state battery which we've covered the concept of yeah in one of our previous episodes as well in

1637
01:35:44.077 --> 01:35:47.619
I mean, the paper that we covered, it was out in Juul, if you remember.

1638
01:35:47.639 --> 01:35:47.839
Yes.

1639
01:35:47.939 --> 01:35:49.100
And it was actually from her lab.

1640
01:35:49.160 --> 01:35:50.100
It was from her lab.

1641
01:35:50.160 --> 01:35:51.781
It was straight up from her lab.

1642
01:35:51.801 --> 01:35:55.563
We were talking about why sodium is so important compared to lithium.

1643
01:35:55.623 --> 01:35:58.425
Lithium is great, but it's hard to get at.

1644
01:35:58.565 --> 01:36:00.326
Sodium is literally in seawater.

1645
01:36:00.526 --> 01:36:00.726
Yes.

1646
01:36:00.826 --> 01:36:03.388
So if we could make a battery out of sodium, that'd be great.

1647
01:36:03.468 --> 01:36:06.550
I mean, it's not going to replace lithium ion batteries, but...

1648
01:36:07.650 --> 01:36:13.441
for certain types of applications, it's going to be key for scalability.

1649
01:36:13.561 --> 01:36:16.146
Like if we want to store like massive amounts of power, um,

1650
01:36:19.270 --> 01:36:21.632
and then deploy it like later at night and things like that.

1651
01:36:21.652 --> 01:36:34.744
Like if you want to have solar farms that, that create a bunch of electricity during the day, and then we want a way to store like, you know, infrastructure level worth of power, sodium batteries might be the key.

1652
01:36:34.804 --> 01:36:39.167
And she's been doing insanely amazing work at the university of Chicago.

1653
01:36:39.287 --> 01:36:43.271
Another, another great professor who is now leaving us.

1654
01:36:44.813 --> 01:36:49.917
Another seasoned trophy winner, Faraday Medal of the Royal Society of Chemistry.

1655
01:36:50.098 --> 01:36:50.198
Oh.

1656
01:36:50.598 --> 01:36:56.883
ECS Battery Division Research Award, ACS Electrochemistry Honors, not a mid-table signing.

1657
01:36:57.124 --> 01:37:01.127
She's becoming the VP of Industry and Distinguished University Professor, the highest faculty rank.

1658
01:37:01.728 --> 01:37:03.369
She's already there as of July 1.

1659
01:37:03.969 --> 01:37:06.912
And Chicago is going to keep a partial appointment today.

1660
01:37:07.272 --> 01:37:08.213
through the transition.

1661
01:37:08.313 --> 01:37:10.455
So this is a loan, not on the free.

1662
01:37:11.156 --> 01:37:13.258
She's stepping down as the director of the DOE hub.

1663
01:37:14.099 --> 01:37:15.240
The hub's money wasn't cut.

1664
01:37:15.900 --> 01:37:19.003
She's just leaving as a chair in science.

1665
01:37:20.184 --> 01:37:22.306
Her reasons for leaving were explicit.

1666
01:37:22.667 --> 01:37:28.512
Oh, the U S turning away from decarbonization and the immigration rules squeezing international, uh,

1667
01:37:28.832 --> 01:37:30.813
including Chinese-born scientists.

1668
01:37:31.654 --> 01:37:34.636
She's one of the ones who's put that on the record.

1669
01:37:35.056 --> 01:37:41.940
Singapore has handed her the keys to her old club, plus a C-suite seat, and she took the loan.

1670
01:37:42.380 --> 01:37:50.265
We are going to move here through to our second-to-last in our coverage of the summer science transfer window.

1671
01:37:50.305 --> 01:37:52.847
If you want us to cover more, let us know.

1672
01:37:52.907 --> 01:37:54.868
We can go country by country.com.

1673
01:37:55.368 --> 01:38:00.753
John Baker is our gravitational wave astrophysicist, NASA Goddard Space Flight Center.

1674
01:38:01.074 --> 01:38:06.699
He is moving to French National Center of Scientific Research in Toulouse.

1675
01:38:06.719 --> 01:38:09.682
This happened actually late last year, also on the free.

1676
01:38:10.282 --> 01:38:33.572
winner of the Goddard's Highest Space Science Honor, the John C. Lindsay Memorial Award in 2008, Kansas City sort of upbringing, Truman State undergrad, Penn State PhD in gravitational physics, and then a postdoc at the Albert Einstein Institute in Potsdam, joined Goddard in 2001 and never left.

1677
01:38:33.892 --> 01:38:35.913
He's been a member of the Gravitational Astrophysics,

1678
01:38:35.993 --> 01:38:57.507
physics lab his big thing was simulations of black hole mergers uh we just talked a little bit about uh some of the work we have around imaging black holes going from still to quote unquote motion um this work that he's done has been a driver for the science case for lisa which is europe's space-based detector that's

1679
01:38:57.767 --> 01:39:01.531
Another not mid table signing a key member here.

1680
01:39:01.571 --> 01:39:09.459
So as he moves to CNS, CNRS, it's going to be the same message mission just on the other side of the zoom.

1681
01:39:09.820 --> 01:39:11.642
He'll be still working in this Lisa arena.

1682
01:39:11.982 --> 01:39:14.545
The launch for that is still aimed for 2035.

1683
01:39:16.416 --> 01:39:26.982
And unlike Yagi, who was not clear about why he was leaving, Baker also mentioned political and social conditions in the U.S. and his family's safety.

1684
01:39:27.663 --> 01:39:30.164
He put that on the record as well.

1685
01:39:30.184 --> 01:39:40.030
You're sensing a trend here around, you know, scientists are normally relatively diplomatic, especially related to policy, because that's who puts food on their table in terms of money and funding.

1686
01:39:43.292 --> 01:39:44.313
Well, this one's – yeah.

1687
01:39:45.360 --> 01:39:48.961
I mean, he straight up said that.

1688
01:39:48.981 --> 01:39:52.202
This is quite rare for scientists to say it.

1689
01:39:53.563 --> 01:39:54.463
Yeah.

1690
01:39:54.643 --> 01:39:55.523
I don't know what to tell you.

1691
01:39:55.703 --> 01:39:56.664
I don't know what to tell you.

1692
01:39:56.904 --> 01:40:02.645
We're going to touch one more, which is not just a single signing.

1693
01:40:03.166 --> 01:40:05.066
This was a double signing.

1694
01:40:05.706 --> 01:40:07.687
And this one is also quite interesting.

1695
01:40:08.091 --> 01:40:16.382
We have Ron Mangun and Tamara Swab, both in the area of cognitive and language neuroscience.

1696
01:40:17.703 --> 01:40:20.467
Currently out of UC Davis, California.

1697
01:40:20.627 --> 01:40:22.770
Again, I promise that wasn't.

1698
01:40:23.951 --> 01:40:24.792
That wasn't bias.

1699
01:40:25.453 --> 01:40:26.574
It's just we have the best.

1700
01:40:27.334 --> 01:40:33.199
Making their way to the University of Birmingham in the UK on the free for both package deal.

1701
01:40:33.620 --> 01:40:37.323
The whole lab is moving to the university.

1702
01:40:38.504 --> 01:40:39.605
This is a two-body problem.

1703
01:40:39.625 --> 01:40:40.726
This is a two-body problem.

1704
01:40:40.746 --> 01:40:42.107
Because they're married scientists.

1705
01:40:42.187 --> 01:40:42.408
Yeah.

1706
01:40:43.188 --> 01:40:43.969
They are married.

1707
01:40:44.190 --> 01:40:46.012
They've had decades at Davis.

1708
01:40:46.072 --> 01:40:59.952
Mangun, distinguished professor of psychology and neurology, director of the Center for the Mind and Brain since 2002, former dean of social sciences, and was the author of the textbook, Cognitive Neuroscience, the Biology.

1709
01:41:00.312 --> 01:41:01.033
of the mind.

1710
01:41:02.273 --> 01:41:23.187
His wife, Tamara Schwab, professor of psychology at Davis for decades as well, the cognitive neuroscience of language lab, how the brain handles meaning, prediction, bilingualism across lifespan, trained at Max Planck, past editor-in-chief of cognition, and a fellow of the Association of Psychological Sciences.

1711
01:41:23.227 --> 01:41:24.267
Across the two of them, they have a

1712
01:41:25.584 --> 01:41:26.665
14 trophies.

1713
01:41:27.065 --> 01:41:28.465
We're talking about winners here.

1714
01:41:29.286 --> 01:41:32.768
These are winners at the top of their game.

1715
01:41:33.108 --> 01:41:34.488
Again, Birmingham signed both of them.

1716
01:41:35.329 --> 01:41:42.272
She's going to arrive as professor of brain and language cognition in the spring, which has already happened.

1717
01:41:42.292 --> 01:41:52.017
He's going to arrive in the fall as the 125th anniversary chair, professor of cognitive neuroscience, and the director of the Center for Human Brain Health.

1718
01:41:52.397 --> 01:41:53.098
This is like...

1719
01:41:54.338 --> 01:41:56.979
Both are going to keep their emeritus ties to Davis.

1720
01:41:57.420 --> 01:42:00.081
They had mixed reasons on record, Swab told NPR.

1721
01:42:00.841 --> 01:42:13.227
The optimism she once associated with U.S. science now feels more present in Britain and Europe, plus a real offer, including the UK Global Talent Fund support.

1722
01:42:15.804 --> 01:42:17.565
you know, it's for man.

1723
01:42:17.605 --> 01:42:22.228
Gonna was, uh, adding Birmingham as an opportunity as much as a reaction.

1724
01:42:22.989 --> 01:42:24.870
Um, this is again, again, again,

1725
01:42:27.380 --> 01:42:28.901
We're losing out to Europe and something?

1726
01:42:29.482 --> 01:42:29.762
Yeah.

1727
01:42:30.462 --> 01:42:31.303
It's been a long time.

1728
01:42:31.363 --> 01:42:31.943
It's been a long...

1729
01:42:32.023 --> 01:42:32.624
In science?

1730
01:42:32.904 --> 01:42:33.845
It's been a very long time.

1731
01:42:33.925 --> 01:42:34.525
In science?

1732
01:42:35.086 --> 01:42:43.451
So I just wanted to give some coverage around this because we did put a new section onto the website, ffppod.com backslash funding.

1733
01:42:43.812 --> 01:42:55.280
There is an ongoing funding battle here in the U.S., not only around who has the power and authority to deal with grant making, but also what funding actually goes into...

1734
01:42:55.780 --> 01:42:58.141
the larger scientific apparatus.

1735
01:42:58.841 --> 01:43:03.903
For those who have not really looked at what funding goes where, it is quite interesting.

1736
01:43:04.224 --> 01:43:18.669
It's based off of the AAAS research and development funding data sets that have existed back through, I believe it's the 1970s and it's actually the 1960s.

1737
01:43:18.709 --> 01:43:24.272
And what's interesting about this is because it's R&amp;D, it includes defense.

1738
01:43:25.432 --> 01:43:30.013
which is the lion's share of R&amp;D spending, there's been some ebbs and flows.

1739
01:43:30.593 --> 01:43:47.518
And so if you check out the funding page, you can see this long history of the story of how research and development in the U.S. has worked as it's related to major milestones across that timeframe, as well as being able to look at the 2026 active fiscal year budget

1740
01:43:47.958 --> 01:44:02.884
That was implemented and is currently having battles over money already written into law reaching the institutions that should be getting it as well as the ongoing battle for the FY fiscal year 2027 budget and the current proposals.

1741
01:44:03.364 --> 01:44:05.765
Both of these things are active issues in Congress.

1742
01:44:06.085 --> 01:44:08.206
So if you want to get engaged in.

1743
01:44:08.686 --> 01:44:23.278
in politics and you take your civic duty in terms of exercising your right to be a participant in the democratic process, your members of Congress, your senators have a role to play in this process.

1744
01:44:24.079 --> 01:44:31.246
And science is one of those issues, particularly when you're looking at the hard sciences, especially.

1745
01:44:31.967 --> 01:44:34.930
There's a lot of complexity around some of the social sciences.

1746
01:44:35.671 --> 01:44:43.078
And if you look at the tracker on what these things actually fund and where they go, you may be surprised at how much we can do with very little.

1747
01:44:43.378 --> 01:44:45.959
as compared to, for example, defense spending.

1748
01:44:46.019 --> 01:44:49.060
So this is a culture issue.

1749
01:44:49.820 --> 01:44:50.881
It's a funding issue.

1750
01:44:51.561 --> 01:44:57.483
It's a competitive landscape issue because in some cases it's just people are giving better opportunity.

1751
01:44:58.344 --> 01:45:03.786
And this show lives off of the frontier research community.

1752
01:45:04.186 --> 01:45:07.687
And so we really want to make sure that you as listeners understand

1753
01:45:07.967 --> 01:45:17.199
understand what's happening right now today with some of our best and brightest and how we can continue to stay competitive as we move into the future

1754
01:45:18.172 --> 01:45:18.372
Yep.

1755
01:45:18.852 --> 01:45:22.755
And I mean, you know, we're going to cover science no matter where it happens.

1756
01:45:23.515 --> 01:45:23.716
Right.

1757
01:45:23.736 --> 01:45:25.557
We're going to cover science if it happens here.

1758
01:45:25.997 --> 01:45:27.598
We've covered several Chinese papers.

1759
01:45:27.618 --> 01:45:29.840
We're going to cover several European papers.

1760
01:45:30.400 --> 01:45:32.521
Anywhere that science happens, we are going to cover it.

1761
01:45:32.922 --> 01:45:37.084
But, you know, we're also American and we care about...

1762
01:45:38.825 --> 01:45:58.231
science happening in this country and we don't we don't want to let go of um the rich history and capabilities that we have um as a country so we're going to continue covering this kind of stuff and you're going to see like you know in the first if i may say lester you know in the first year we kind of we kind of shied away from from doing this kind of stuff yeah

1763
01:45:58.631 --> 01:46:08.942
But I think we've decided as an organization, as an FFP nation, that this is important and we're going to start caring a lot more.

1764
01:46:10.764 --> 01:46:18.232
As always, I am your host, Lester Nare, joined by my co-host and our resident PhD.

1765
01:46:18.272 --> 01:46:20.894
We made it under the two-hour mark.

1766
01:46:21.943 --> 01:46:24.264
That's become the new benchmark.

1767
01:46:24.844 --> 01:46:29.126
Again, we are super grateful for all of you who listened to the end of the pod.

1768
01:46:29.246 --> 01:46:37.749
One last reminder, if you have made it this far, you should definitely enter the merch giveaway at ffppod.com backslash.

1769
01:46:38.229 --> 01:46:38.629
survey.

1770
01:46:38.669 --> 01:46:49.295
We're trying to understand a little bit more about our audience so we know what to cover, what you're looking for, what you like, don't like about the show as we move into year two.

1771
01:46:49.955 --> 01:47:02.321
We are very excited in our next two episodes, if I'm not mistaken, we are going to be covering something that's very near and dear to our resident PhD, Krishna's heart.

1772
01:47:02.681 --> 01:47:03.622
And I will

1773
01:47:07.945 --> 01:47:08.336
Thanks to me.
