Latest / Breaking Points with Krystal and Saagar / 9/10/26: AI Whistleblowers Dire Warning, Mathematician Says OpenAI Stole Solution, Data Center Support Collapses

Transcript

9/10/26: AI Whistleblowers Dire Warning, Mathematician Says OpenAI Stole Solution, Data Center Support Collapses · Breaking Points with Krystal and Saagar

0:00
Speaker 1: Hey guys, Ready or Not 2024 is here and we
0:04
Speaker 1: here at Breaking Points are already thinking of ways we
0:06
Speaker 1: can up our game for this critical election.
0:08
Speaker 2: We rely on our premium subs to expand coverage, upgrade
0:11
Speaker 2: the studio, add staff, give you guys the best independent
0:14
Speaker 2: coverage that is possible. If you like what we're all about,
0:17
Speaker 2: it just means the absolute world to have your support.
0:19
Speaker 2: But enough with that, let's get to the show. Good morning, everybody.
0:25
Speaker 2: Happy Thursday. We have an amazing show for everybody today.
0:28
Speaker 2: What do we have, Crystal? Indeed we do.
0:30
Speaker 1: A lot going on in the AI world. We're going
0:32
Speaker 1: to dig deep this morning. An anthropic researcher resigned and
0:35
Speaker 1: is giving up a gigantic payday over his concerns that
0:39
Speaker 1: AI could destroy us all. He did some interviews. We
0:41
Speaker 1: also are learning about new security incidents. We'll break all
0:44
Speaker 1: of that down for you. Also, at the same time.
0:47
Speaker 1: OpenAI is bragging that they solved this longstanding, very challenging
0:51
Speaker 1: math problem that no one else had been able to
0:52
Speaker 1: solve before.
0:53
Speaker 3: But there are.
0:54
Speaker 1: Allegations, and it appears they may have cribbed some of
0:58
Speaker 1: the work from an actual human. that was working or
1:00
Speaker 1: actually human beings that were working on this problem. So
1:02
Speaker 1: we will dig into the evidence there. New research says
1:05
Speaker 1: that data centers are not, in fact, a major boon
1:09
Speaker 1: to local economies. This comes amid increasing political fallout over AI,
1:15
Speaker 1: data centers, et cetera.
1:16
Speaker 3: So we'll dig into that. We'll dig into the polling
1:17
Speaker 3: as well.
1:18
Speaker 1: Trump at his big midterm convention in Texas is now
1:21
Speaker 1: promising Americans $ 5, 000 checks if they vote for Republicans and
1:27
Speaker 1: keep them in the majority. So that's interesting given some
1:31
Speaker 1: of the promises that have already been made on that front.
1:33
Speaker 3: 9-11 is tomorrow.
1:34
Speaker 1: Glenn Greenwald is going to join us for a bit
1:36
Speaker 1: of a retrospective looking at how that kicked off the
1:38
Speaker 1: era of mass surveillance and also how many of the
1:41
Speaker 1: villains of the 9-11 era and the Iraq war era
1:44
Speaker 1: are still with us today to cheerlead us into the
1:46
Speaker 1: Iran war. And we're also going to do an update
1:49
Speaker 1: on the Iran war.
1:49
Speaker 3: Of course, the Houthis actually taking more territory in a
1:52
Speaker 3: stunning offensive.
1:53
Speaker 1: That is going to make it easier for them to
1:55
Speaker 1: control waterways near them. So gas prices could suffer once again.
2:00
Speaker 1: And Iran, we're learning from those airstrikes the other night,
2:03
Speaker 1: was able to successfully damage significant amount of U.S.
2:06
Speaker 3: Aircraft.
2:07
Speaker 1: Also, our intervention in the bond market did absolutely nothing.
2:10
Speaker 1: So a lot of different things going on there.
2:12
Speaker 2: Exactly. That's right. Thank you to everybody who has been
2:14
Speaker 2: signing up. at BreakingPoints.com. Ryan did another record yesterday with
2:19
Speaker 2: his sign-up. So my new pitch is, if you sign up,
2:23
Speaker 2: Ryan gets to continue to do whatever he wants. I've
2:25
Speaker 2: admitted defeat, and that's it. BreakingPoints.com, put the free month
2:28
Speaker 2: trial up there. Let Ryan be Ryan. If you love
2:32
Speaker 2: Ryan Grimm, sign up for BreakingPoints.com. Okay. Maybe that will win.
2:37
Speaker 2: And if you hate him too, you should also sign
2:39
Speaker 2: up or don't sign up. Um, and what is it?
2:42
Speaker 2: A premium subscription. Diesel now goes for about $ 200 a barrel.
2:46
Speaker 2: I'm borrowing from Ryan. Uh, so our premium subscription is,
2:50
Speaker 2: is less than half of that.
2:51
Speaker 3: You can buy one and gift one.
2:53
Speaker 4: Right.
2:53
Speaker 2: You could buy, you could buy one, you could buy
2:56
Speaker 2: one or two premium subscriptions. For the price of, or
3:00
Speaker 2: for less than two barrels of diesel. I mean, come on.
3:04
Speaker 2: That's pretty incredible. Yeah, that is amazing. That's my sign-off.
3:06
Speaker 1: And Christmas is right around the corner, guys. And our
3:09
Speaker 1: pre-me memberships are not going to get stuck in the
3:10
Speaker 1: Strait of Hormuz.
3:11
Speaker 2: Yes, exactly.
3:12
Speaker 3: It's reliable.
3:13
Speaker 2: It's actually useful to your life. All right. And by
3:16
Speaker 2: the way, if you're watching, please hit subscribe to our
3:18
Speaker 2: YouTube channel. If you're listening to the podcast, please share
3:20
Speaker 2: an episode with a friend. But as Crystal said, let's
3:22
Speaker 2: start with AI. Tons of movement in the field.
3:24
Speaker 1: Yeah, so we had this anthropic researcher who resigned and
3:27
Speaker 1: posted a long thread on Twitter. We'll get to that
3:29
Speaker 1: thread in just a minute. But he made the rounds
3:32
Speaker 1: on some shows yesterday, including on Fox News and CNN.
3:34
Speaker 1: Let's go ahead and take a listen to a little
3:36
Speaker 1: bit of what he had to say to Brett Baer
3:38
Speaker 1: about his concerns.
3:39
Speaker 4: I agree with Evan that right now the threat is
3:42
Speaker 4: minimal of actual takeover. But because of self-improvement, this AI
3:46
Speaker 4: is making themselves smarter, could happen as soon as next year.
3:49
Speaker 4: Some people even say six months. So this attitude is
3:52
Speaker 4: shared across the industry. where people essentially are aware that
3:56
Speaker 4: the work they are currently doing, if we don't slow down,
3:59
Speaker 4: could lead to AI takeover by the end of the decade.
4:05
Speaker 5: So what do you say to the people who look
4:07
Speaker 5: at this and say the American government could potentially put
4:10
Speaker 5: in regulations, could put in guardrails, and could throttle back
4:14
Speaker 5: some companies or set up regulations, but we have no
4:17
Speaker 5: control over what China does, what India does.
4:21
Speaker 2: How do you answer that question?
4:22
Speaker 5: You heard Scott Besson in the piece before this kind
4:25
Speaker 5: of make a similar argument.
4:28
Speaker 4: I think you have to think of this as an
4:30
Speaker 4: alms race. This technology is not ordinary technology. It's not
4:34
Speaker 4: like other forms of innovation. This is essentially a super weapon.
4:37
Speaker 4: And I think that I expect the international community to
4:41
Speaker 4: be ready to make agreements about how to control the
4:44
Speaker 4: development of super weapon technology.
4:49
Speaker 5: So people would say, if you say it's like an
4:51
Speaker 5: arms race, then we've dealt with nuclear weapons and kind
4:56
Speaker 5: of a standoff with mutual assured destruction and worked with
4:59
Speaker 5: allies and enemies to say, we've got the stuff too.
5:05
Speaker 5: Don't think that you're going to get away with anything.
5:07
Speaker 5: So does that work against what you're saying? What would
5:10
Speaker 5: you like to see happen?
5:13
Speaker 4: Yeah. So unfortunately, we know how to control nuclear weapons
5:16
Speaker 4: pretty much. We don't yet know how to control A.I.
5:19
Speaker 4: This is possibly the most dangerous technology that humanity has
5:22
Speaker 4: ever created. It's basically out of science fiction. I think
5:25
Speaker 4: we have no other choice but to cooperate internationally because
5:28
Speaker 4: an arms race would be disastrous in a way that
5:32
Speaker 4: no other human activity has been in the past.
5:34
Speaker 3: So that was Jacob Coxon.
5:36
Speaker 1: Let me go and put his thread up on the screen,
5:38
Speaker 1: which went absolutely viral. This has now been viewed more
5:42
Speaker 1: than 146 million times on Twitter. which is interesting because
5:46
Speaker 1: he is not the first AI researcher to sound the alarm,
5:49
Speaker 1: to resign, to put on a statement. Another one previously said, like,
5:53
Speaker 1: I'm out of here. I'm going to write poetry. Good
5:55
Speaker 1: luck to humanity, effectively. So let me read what he
5:58
Speaker 1: says here, and then we can talk some about why
6:00
Speaker 1: this particular warning has gone as viral as it did.
6:03
Speaker 1: I'm going to read the whole thread, so just bear
6:05
Speaker 1: with me. He says, I resigned from Anthropic today. I
6:07
Speaker 1: spent the last three years doing pre-training research at both
6:10
Speaker 1: OpenAI and Anthropic. So we've seen the inside of both
6:13
Speaker 1: of these major companies. Neither company is acting responsibly. They
6:17
Speaker 1: are racing straight to self-improving superintelligence and gambling with our lives.
6:22
Speaker 1: Do not underestimate the power of this technology. These will
6:25
Speaker 1: soon be superhuman systems that can hack anything, revolutionize any
6:28
Speaker 1: field overnight, and acquire real power and resources. We have
6:31
Speaker 1: all witnessed the progress in each of these domains, and
6:33
Speaker 1: progress is not slowing. The people building AI earnestly believe
6:37
Speaker 1: that it could kill us all by the end of
6:39
Speaker 1: the decade. This is not a marketing stunt, if anything.
6:43
Speaker 1: Many executives and senior researchers will couch their phrasing in
6:46
Speaker 1: the press to sound sensible, but I hear the same
6:48
Speaker 1: people express fear privately. No other human activity poses this
6:52
Speaker 1: level of danger. A common response is, if they truly
6:55
Speaker 1: believe this, why are they still building it? At OpenAI,
6:58
Speaker 1: many have not deeply internalized the civilizational stakes. At Anthropic,
7:02
Speaker 1: the stakes are well understood, but they're locked in a
7:05
Speaker 1: race to get there first. They believe no one else
7:07
Speaker 1: will act responsibly, so they must do it themselves despite
7:10
Speaker 1: the risk. That, by the way, is like a classic
7:12
Speaker 1: prisoner's dilemma. Accepting this race and entering the endgame is
7:16
Speaker 1: a hubristic gamble that should not be launched from a
7:18
Speaker 1: private company's slack. Tempting the speedrun alignment should require extraordinary
7:22
Speaker 1: confidence that there are no better trajectories available. I am
7:25
Speaker 1: optimistic about the potential for coordination.
7:27
Speaker 3: That's good. That makes one of us.
7:28
Speaker 1: Warning shots like the hugging face attack have made pacing
7:30
Speaker 1: agreements between U.S.
7:31
Speaker 3: Labs more viable.
7:33
Speaker 1: I don't feel like we're on track to prevent a
7:35
Speaker 1: global race which may require costly actions such as a
7:37
Speaker 1: temporary ban on improving model capabilities. If you are a
7:41
Speaker 1: lab researcher, I urge you to consider what the next
7:44
Speaker 1: few years will actually feel like. Do you want to
7:47
Speaker 1: kick off a super intelligent RL run without a rigorous
7:51
Speaker 1: understanding of its mind? Should you put your head down because, quote,
7:54
Speaker 1: it's happening anyway? or take this moment to call for
7:58
Speaker 1: different conditions. This has set off a flood of other
8:01
Speaker 1: people from inside the industry, many of them who are
8:03
Speaker 1: still at their posts at Anthropic, at OpenAI, we'll get
8:06
Speaker 1: to some of that in a second, saying, yeah, he's right.
8:08
Speaker 1: And yeah, I do think that there's about a 10%
8:11
Speaker 1: risk that super intelligent AI is going to literally destroy
8:14
Speaker 1: all of humanity. So quite a stunning moment. And clearly
8:18
Speaker 1: sort of elites and the public at large were ready
8:21
Speaker 1: to hear this message.
8:22
Speaker 2: This was the most viral incident. And in my opinion,
8:27
Speaker 2: it really broke containment. And the way that you know
8:30
Speaker 2: that is exactly how Jacob was on Fox News. He
8:35
Speaker 2: was on CNN. Ted Cruz, we'll show you, reacted to it.
8:39
Speaker 2: It was on The View. Obviously, it's the A Block website.
8:42
Speaker 2: of our show. If you have heard any warning from
8:46
Speaker 2: the AI industry outside of, let's say, some 2025 pronouncements
8:50
Speaker 2: from Dario Amadei, probably a year or so now at
8:53
Speaker 2: this point where he's said many similar things, you and I,
8:56
Speaker 2: we cover this day in and day out. None of
8:58
Speaker 2: this was surprising. So the story here is about how
9:01
Speaker 2: much this breach containment. And I also think how much
9:04
Speaker 2: that this has also been agreed upon by the very
9:07
Speaker 2: people in the industry itself. So let's put the next
9:10
Speaker 2: one up here on the screen because you could say, okay,
9:12
Speaker 2: Jacob Coxon, the guy barely worked there for a couple
9:15
Speaker 2: of weeks. He worked at OpenAI. Oh, he's just a
9:18
Speaker 2: agitator or an activist or something like that. Well, here
9:21
Speaker 2: is Evan Hubinger. He is the current alignment science lead
9:26
Speaker 2: at Anthropic, right? So a current employee at the company,
9:30
Speaker 2: effectively his boss. He says, Jacob is correct here. We
9:34
Speaker 2: really do earnestly believe AI could kill all humans. I
9:38
Speaker 2: personally think it is a greater than 10% chance within
9:42
Speaker 2: the next decade. I believe Anthropic is trying its best,
9:46
Speaker 2: but we do not yet have a plan to solve
9:48
Speaker 2: alignment for superintelligence and are not clearly on track to. Now,
9:54
Speaker 2: this is extraordinary. This is a company which is going
9:57
Speaker 2: public in the next six months. Many people in the
10:00
Speaker 2: AI industry were saying, oh, this is a comms disaster. No,
10:04
Speaker 2: this is a reality disaster. And in fact, it is
10:07
Speaker 2: a reality disaster for which has been in plain English
10:10
Speaker 2: from the CEO of Anthropic, from the CEO of OpenAI,
10:14
Speaker 2: from the employees themselves. Ryan and I covered the Anthropic's
10:19
Speaker 2: total addressable market where they said it's $ 30 trillion. That
10:22
Speaker 2: is the entire US economy where they said, our plan
10:26
Speaker 2: for our total addressable market is $ 30 trillion, effectively replace
10:30
Speaker 2: all aspects of commerce in human society. So you're like, okay.
10:34
Speaker 2: And then you continue to look not just at the
10:37
Speaker 2: warnings within the company, not just at the past statements,
10:41
Speaker 2: not just at Jacob, but his bosses of his counterparts
10:45
Speaker 2: at OpenAI, and all of them ultimately come back to
10:48
Speaker 2: the same thing. It all seems like we're on this
10:51
Speaker 2: track to make it so that we will create this
10:54
Speaker 2: technology of which we have no control. And I do think, though,
10:58
Speaker 2: and I would really love, we've tried to make contact
11:00
Speaker 2: with Jacob. Jacob, if you're listening, please, we really would
11:03
Speaker 2: love to interview you more than five minutes of Bret
11:06
Speaker 2: Baier talk. With the stick mic.
11:07
Speaker 3: I'm in touch with him.
11:08
Speaker 2: Mike, come on. Okay. So, I mean, here's the thing.
11:10
Speaker 2: We really need a longer conversation because you know what?
11:14
Speaker 2: If you're going to make extraordinary claims, require extraordinary evidence,
11:17
Speaker 2: so you say you're going to kill humanity. How? Like
11:19
Speaker 2: what is the exact mechanism? Is it taking down the
11:22
Speaker 2: energy grid? Is it bioengineering a targeted virus? Is it
11:28
Speaker 2: a nuclear weapons launch, which are all controlled by technology?
11:32
Speaker 2: Like when you're going to say something like this, you
11:34
Speaker 2: actually owe it to us to tell us like, how
11:35
Speaker 2: exactly do you mean it would kill all of us?
11:38
Speaker 2: But the second thing, and this is why we focus
11:40
Speaker 2: so much on the cybersecurity part, is it is so
11:43
Speaker 2: obvious now through multiple incidents that the infrastructure of which
11:47
Speaker 2: our modern lives depend on is not prepared for this whatsoever.
11:51
Speaker 2: And that the company's who allegedly have all these safety
11:54
Speaker 2: and alignment and all these things, these models continue to
11:58
Speaker 2: break containment. And that breaking containment is actually a feature,
12:02
Speaker 2: not a bug, because it means that it is developing
12:05
Speaker 2: this recursive intelligence and its ability to improve itself, to reason,
12:10
Speaker 2: to work outside the bounds of what humans tell it
12:13
Speaker 2: to do so that it can break containment and it
12:15
Speaker 2: can reach further. Ultimately, that's what we have decided the
12:18
Speaker 2: value here in the United States is all about.
12:21
Speaker 4: Why?
12:22
Speaker 2: Because of brute capitalist enterprise. The last thing I just
12:25
Speaker 2: want to say, and this is a much longer conversation
12:27
Speaker 2: around China, is everybody says we need to beat China.
12:30
Speaker 2: So I actually spent a lot of time reading primary
12:33
Speaker 2: source documents and translations. What do the Chinese believe? They
12:38
Speaker 2: are not, quote, AGI-pilled. And I'm quoting directly from one
12:41
Speaker 2: of their top analysts who says, the Chinese Communist Party
12:46
Speaker 2: is looking at this with deep suspicion, with some interest.
12:50
Speaker 2: They have a absolute bedrock philosophy as if this thing
12:54
Speaker 2: ever got to a point where it could supplant the
12:56
Speaker 2: Communist Party. We're shutting this shit down, period. Their way
13:00
Speaker 2: that they are currently developing their AI industry is DeepSeek
13:04
Speaker 2: is their only company which is deeply interested in AGI.
13:08
Speaker 2: And even then, the DeepSeek CEO is very different than
13:12
Speaker 2: our top CEOs. He says, we do not believe in profit.
13:16
Speaker 2: We do not believe that this is a capitalist enterprise.
13:19
Speaker 2: They look at it a little bit more as a
13:20
Speaker 2: research project, I guess is probably the best way to
13:23
Speaker 2: describe it. They also are willing to sacrifice profit by
13:26
Speaker 2: making sure all of their models are open source. And
13:29
Speaker 2: a lot of it is resource competition with the United
13:31
Speaker 2: States over our continued export controls on their GPUs. But
13:35
Speaker 2: the second thing is that they don't have a counterpart,
13:38
Speaker 2: at least from what I could tell, of all of
13:40
Speaker 2: these other people who are racing towards AGI. The way
13:43
Speaker 2: that the Chinese are developing their AI industry is they
13:46
Speaker 2: have given a diktat basically to the entire market where
13:51
Speaker 2: they're like, your job is to create AI that has
13:55
Speaker 2: capabilities to enhance directives which are important to the state,
13:59
Speaker 2: aka manufacturing, aka robotics. And even then, I don't know
14:04
Speaker 2: if you saw this, but The humanoid robots, which a
14:06
Speaker 2: lot of people have been very worried about, they just
14:09
Speaker 2: smashed that industry over the head. They were like, this
14:11
Speaker 2: seems to be going too far. This is creating some problems.
14:14
Speaker 2: Just yesterday, they issued a regulatory authority basically wiping out
14:18
Speaker 2: or at least dramatically restricting a lot of the humanoid
14:21
Speaker 2: robot development that's been happening inside of China. So they
14:23
Speaker 2: have a much, much more, they have a very different system.
14:28
Speaker 2: They're not going towards this artificial and recursive super intelligence,
14:31
Speaker 2: the ability for the models to train themselves. They don't
14:34
Speaker 2: seem to care about that as much. They're trying to
14:36
Speaker 2: use AI to enhance the actual capabilities of the Chinese
14:40
Speaker 2: economy in manufacturing and in a productivity sense. So they're
14:45
Speaker 2: very much in the school of develop the tools, not
14:48
Speaker 2: the supplantation. We are in the opposite, right, where we
14:51
Speaker 2: have not seen as much of the enterprise application yet.
14:56
Speaker 2: for AI. And instead, we're racing forward with all these
14:59
Speaker 2: data centers and we're going sprinting towards AGI with some
15:02
Speaker 2: belief that it'll magically create some great new economic opportunity.
15:06
Speaker 4: Yeah.
15:06
Speaker 1: So their approach is to actually focus on the applications.
15:09
Speaker 1: And there was actually a big article about how automation
15:11
Speaker 1: is replacing all of these blue collar workers in China
15:13
Speaker 1: and they have an economic problem on their hands because
15:15
Speaker 1: of that. But that's been much more of the focus.
15:18
Speaker 1: The other thing that I've talked to Garrison about, and
15:20
Speaker 1: then he's made the case for here too, is what
15:22
Speaker 1: the Chinese are able to do very effectively is fast follow.
15:25
Speaker 1: So wherever we push the bleeding edge of the frontier,
15:28
Speaker 1: they're able to follow behind very quickly. Well, what does
15:31
Speaker 1: that mean? That means that the further we push the frontier,
15:34
Speaker 1: the more capability the Chinese are going to have to.
15:36
Speaker 1: So the whole like, well, the Chinese will do it
15:38
Speaker 1: thing really misunderstands the nature and the dynamic of what
15:41
Speaker 1: is happening here. In any case, Getting back to Jacob,
15:45
Speaker 1: I think the reason why this broke through at this
15:47
Speaker 1: moment is because there had been enough concern and coverage
15:49
Speaker 1: over the hugging face hack and crime spree and civilizational
15:55
Speaker 1: buildup of these AIs that people were kind of paying attention.
15:59
Speaker 1: So that when he said, OK, guys, I really have
16:01
Speaker 1: concerns so much so that and I again, I want
16:04
Speaker 1: to reiterate this. He's leaving and he is giving up
16:09
Speaker 1: all of the money that he could have made in
16:11
Speaker 1: the anthropic IPO. Like he was due for a massive payday,
16:16
Speaker 1: one of the largest in the history of capitalism. And
16:19
Speaker 1: he walked away from all of that in order to
16:22
Speaker 1: sound this alarm. So again, once again, the objections I'm hearing,
16:26
Speaker 1: which are sort of unfalsifiable and have just really are
16:30
Speaker 1: increasingly devolving to the land of QAnon conspiracy theories. is
16:34
Speaker 1: that this is all some elaborate marketing hype play, okay?
16:39
Speaker 3: There is, first of all, no evidence for that.
16:41
Speaker 1: Second of all, we've had many whistleblowers leave the companies
16:44
Speaker 1: and give up their paydays to come out and say,
16:46
Speaker 1: I'm genuinely deeply concerned. Have we had any whistleblowers come
16:50
Speaker 1: out and say, I'm coming forward because they're lying to
16:53
Speaker 1: you about the capabilities of this? They're lying to you
16:55
Speaker 1: about the labor replacement. They're lying to you about the
16:58
Speaker 1: existential risk. No, we see no evidence for that whatsoever.
17:02
Speaker 1: And as we discussed earlier this week, that whole idea,
17:05
Speaker 1: this is a hype cycle and it's just marketing. It
17:08
Speaker 1: should have been put to bed by the fact that
17:10
Speaker 1: we now know open AI tried to cover up some
17:13
Speaker 1: of the safety incidents and AI agents escaping the sandbox
17:18
Speaker 1: and make sure that that did not become public. Their
17:21
Speaker 1: hand was forced. And they've continued to constrain access to
17:25
Speaker 1: researchers to be able to tell exactly what the hell
17:27
Speaker 1: has been going on at their company. So if it
17:29
Speaker 1: was all a hype cycle, wouldn't they want to get
17:31
Speaker 1: all of this out there? Oh, my God, our AIs
17:33
Speaker 1: are so dangerous. No, that is not the way they've
17:35
Speaker 1: been behaving. So I need people who had that view, which, again,
17:39
Speaker 1: I understand where it came from, because you should not
17:41
Speaker 1: trust AI executives. You should not trust these tech oligarchs whatsoever.
17:47
Speaker 1: But I really would implore you to reconsider your priors
17:50
Speaker 1: if that is where you have been, because we now
17:53
Speaker 1: have evidence that contradicts that. You had not just, you know,
17:59
Speaker 1: Jacob's boss effectively come out and say, yeah, he's right.
18:02
Speaker 1: We do actually think there is a 10 percent chance
18:06
Speaker 1: of that these will lead to human extinction. You have
18:10
Speaker 1: a huge number of people from within the industry. Go
18:13
Speaker 1: ahead and put A4 up on the screen. This Twitter user,
18:17
Speaker 1: Taylor Populars, did the Lord's work here and compiled a
18:20
Speaker 1: bunch of quotes from people, leading scientists in the industry
18:23
Speaker 1: coming out and basically saying, yeah, we agree. So he
18:26
Speaker 1: starts with Jacob Coxon. Then it goes to Samuel Marks,
18:29
Speaker 1: scalable oversight lead at Anthropic. Quote, AI developers believe their
18:33
Speaker 1: tech could cause human extinction or similarly bad outcomes. This
18:36
Speaker 1: could happen in the next few years. Next one, Evan Hubinger.
18:39
Speaker 1: We read that one earlier. Alex Turner, former research scientist
18:42
Speaker 1: at Google DeepMind, says, yes, I left Google DeepMind in June.
18:46
Speaker 2: Jacob is right.
18:47
Speaker 1: Many researchers believe they're building something that could kill everyone
18:50
Speaker 1: on the planet. It was literally my day job to
18:51
Speaker 1: think about how to stop that. Another chief scientist at OpenAI,
18:55
Speaker 1: Jacob Pachocchi, this is a time that calls for extreme caution.
18:59
Speaker 1: I'm concerned no one is prepared for the consequences of
19:01
Speaker 1: a continued rapid rise in machine intelligence. Jason Wolf, a
19:04
Speaker 1: researcher at OpenAI, I don't know what my probabilities are
19:07
Speaker 1: on literal extinction, but I think there are a number
19:09
Speaker 1: of ways AI could go poorly for humanity. At the current, frankly,
19:12
Speaker 1: terrifying pace, humanity will be quite lucky if we manage
19:15
Speaker 1: to find and stay on the narrow path between all
19:17
Speaker 1: the bad outcomes. Jonathan Richard Swartz, former senior research scientist
19:21
Speaker 1: at Google DeepMind, I left DeepMind after seven years and
19:24
Speaker 1: since rejected offers to join the other two labs due
19:27
Speaker 1: to severe concerns about the concentration of power these labs represent.
19:30
Speaker 1: What is currently happening in this field is deeply unhealthy
19:33
Speaker 1: for society. Go to the next page. You can see
19:36
Speaker 1: just an image of some of the tweets. that have
19:39
Speaker 1: put out, and every time I refresh my timeline on Twitter,
19:42
Speaker 1: I see some new researcher, open AI, anthropic, deep mind
19:46
Speaker 1: coming out and saying, yes, the risks are real, we
19:48
Speaker 1: are genuinely concerned about this. This isn't just some marketing
19:52
Speaker 1: hype cycle. So then the natural question is, what the
19:55
Speaker 1: hell are you people doing? Like, why are we doing this?
19:58
Speaker 1: And I think that Jacob is taking the right course
20:00
Speaker 1: of action here by walking away and say, I'm not
20:02
Speaker 1: going to be involved. However, it is not hard to
20:04
Speaker 1: imagine the very human and very capitalist driven psychology of
20:09
Speaker 1: that leads to them staying.
20:10
Speaker 3: We've seen a version of it in the Trump administration.
20:12
Speaker 1: Well, if it's not me there and I'm genuinely concerned
20:15
Speaker 1: and I'm a responsible actor, I can help shape things
20:18
Speaker 1: from the inside. Then someone else who has less scruples
20:20
Speaker 1: is going to come in and they're just going to
20:22
Speaker 1: do the same thing. Or the other thing that you
20:24
Speaker 1: hear from politicians all.
20:26
Speaker 3: The time is, ah, but if not us, then China.
20:29
Speaker 1: So they are locked in this very obvious prisoner's dilemma. Again,
20:34
Speaker 1: I think the right thing to do would be to
20:36
Speaker 1: come forward the way that Jacob has, but it is
20:38
Speaker 1: not hard to imagine the human psychology that has led
20:42
Speaker 1: to this completely bizarre and insane situation where the majority
20:47
Speaker 1: of the people who are building the product think the
20:49
Speaker 1: product could have an absolutely catastrophic outcome.
20:52
Speaker 4: Right.
20:52
Speaker 2: And the one thing that you didn't add in there,
20:54
Speaker 2: which is in the White House case of politicians or others,
20:59
Speaker 2: that's just power. Here we're talking about generational wealth. We're
21:02
Speaker 2: talking about the ability where you can stay, invest your shares,
21:07
Speaker 2: and literally not have to work again a day in
21:09
Speaker 2: your life. I'm pretty sure Jacob is like 2017. I mean,
21:12
Speaker 2: I don't know the exact number. I'd be willing to
21:14
Speaker 2: bet it's tens of millions of dollars at a minimum
21:17
Speaker 2: that he probably walked away from. So, I mean, ask
21:20
Speaker 2: yourself that question, too, from a family perspective. I'm like,
21:23
Speaker 2: is it that easy to walk away from something like that?
21:27
Speaker 2: Or what if you never had to work again? What
21:30
Speaker 2: would you be willing to sacrifice?
21:31
Speaker 3: I wonder if there'll be a mass exodus after the
21:33
Speaker 3: Sheriff's Fest.
21:34
Speaker 2: Yeah, probably. I mean, there almost always is in normal times.
21:36
Speaker 2: So then I imagine with something like this. Let's go
21:39
Speaker 2: to the next part, A6, please, just to show you.
21:43
Speaker 2: This actually came on the heels of Jacob's announcement. In
21:48
Speaker 2: my opinion, that's got nearly the amount of coverage it deserves,
21:51
Speaker 2: which is Anthropic literally on the same day had to
21:54
Speaker 2: put out a statement, quote, an alignment assessment of recent
21:57
Speaker 2: cybersecurity incidents. And you will be shocked to learn that
22:02
Speaker 2: what they discovered after auditing some of their systems is
22:06
Speaker 2: is that they had multiple incidents of a scan of
22:08
Speaker 2: their transcripts where Claude could have obtained internet access through
22:12
Speaker 2: a cyber evaluation. Given the volume, this missed a set
22:16
Speaker 2: of transcripts, et cetera, and we scanned these and identified
22:19
Speaker 2: a fourth incident from January of 2026 involving an early
22:23
Speaker 2: version of Claude, and we have notified all effective parties.
22:27
Speaker 2: They basically went through and audited all of their systems,
22:30
Speaker 2: and they discovered that even though they had told these AIs,
22:34
Speaker 2: which were training, in an allegedly non-internet-connected system that it
22:39
Speaker 2: was able to break out and hack into different parts,
22:43
Speaker 2: or at the very least, hack their own internal system
22:46
Speaker 2: and gain access to the internet to be able to
22:48
Speaker 2: do whatever it wanted. I mean, these, again, are the
22:51
Speaker 2: frontier models where you have, I mean, these people are
22:55
Speaker 2: paid exorbitant amounts of money. I can't even describe to
23:00
Speaker 2: you what the money race right now in Silicon Valley is.
23:03
Speaker 2: To put it in perspective, The second tier players are
23:06
Speaker 2: making $ 10 to $ 15 to $ 20 million a year. Imagine
23:09
Speaker 2: what these guys are making who are on the actual
23:12
Speaker 2: cutting edge, 150 IQ machine learning engineers. I mean, these
23:16
Speaker 2: people are printing to the, I can't even, again, like,
23:20
Speaker 2: and I'm talking just cash, not even a let alone
23:22
Speaker 2: the shares of what they are being able to take away.
23:25
Speaker 2: That is the extent of what we're dealing with. And
23:28
Speaker 2: the talent pool who are around this entire thing, who
23:31
Speaker 2: are trying their best to align the technology or any
23:36
Speaker 2: of that. But I think what they're trying to tell
23:37
Speaker 2: us is that it is unalignable in the current fashion
23:41
Speaker 2: without government regulation. I just come back to there is
23:45
Speaker 2: literally no read. I mean, again, my China reading, every
23:48
Speaker 2: model has to be registered and evaluated by the Chinese government.
23:53
Speaker 2: The Chinese government itself, I encourage everybody to go read
23:55
Speaker 2: Dan Wang's book, to the extent that the CCP and
23:59
Speaker 2: the China is a society of elites, its elites are engineers.
24:03
Speaker 2: There are many pluses and minuses to that. But one
24:06
Speaker 2: of the pluses within the Chinese system is they have
24:09
Speaker 2: genuine expertise at the government regulatory level and their ability
24:13
Speaker 2: to inform the Chinese political leaders so that the two
24:15
Speaker 2: of them come up with a strategy and their regulators
24:18
Speaker 2: are just as knowledgeable as many of the researchers, at
24:22
Speaker 2: least again, from what I understand, Many of their Pentagon bureaucracy, again, others,
24:26
Speaker 2: they don't have democracy. They have permanent government so these
24:29
Speaker 2: people can stay there for 15, 20 years. Their deep state,
24:32
Speaker 2: as long as it's intelligent, it's a very well-run system.
24:35
Speaker 2: So for them, they're able to guide the technology in
24:38
Speaker 2: the way that they want, look with interest, read the
24:42
Speaker 2: research papers, register the models. We have none of that, zero,
24:46
Speaker 2: whenever it comes to our models. That's the last part
24:49
Speaker 2: here is that Jacob called for international cooperation. cooperation. I mean,
24:54
Speaker 2: in a sane world, Jacob would be called along with
24:59
Speaker 2: all the CEOs of these companies to testify for hours
25:04
Speaker 2: before Congress. They would testify. There would be a working
25:08
Speaker 2: group immediately to develop bipartisan legislation. This would be with
25:14
Speaker 2: guidance from the White House, from industry leaders. And we
25:17
Speaker 2: would all, you know, there would be some democratic input.
25:20
Speaker 2: None of that is happening. Trump is out here being like,
25:22
Speaker 2: I'm not that worried about it. What did he say?
25:24
Speaker 2: He said it'd all work out in the end. Something
25:26
Speaker 2: like that. I mean, it's great. Yeah, we have this thought.
25:29
Speaker 2: Take a listen.
25:29
Speaker 6: A8.
25:30
Speaker 1: Some people say the worst case scenario with AI is
25:32
Speaker 1: that the robots, the machinery learns to.
25:35
Speaker 3: Obviously, it thinks for itself.
25:36
Speaker 2: That's what it does.
25:38
Speaker 3: And that could turn against humanity. Do we have the guardrails?
25:43
Speaker 3: It's going to be fine. We'll always have something to
25:45
Speaker 3: stop them, right?
25:46
Speaker 7: We'll have a little gear.
25:47
Speaker 3: I really hope so. I really don't like that robot.
25:50
Speaker 2: We'll stop. Yeah, see, I mean, that's the extent of
25:52
Speaker 2: the thinking of Trump. Look, geriatric leader who doesn't stand technology,
25:56
Speaker 2: maybe we could say, oh, that's just a one-off. That's
25:59
Speaker 2: the whole system, literally, what Chuck Grassley is going to
26:01
Speaker 2: regulate AI? That guy's like 90 years old.
26:04
Speaker 1: I mean, in fairness, Bernie has been the most intelligent
26:07
Speaker 1: on this, and he is also very aged.
26:09
Speaker 3: But he is an outlier.
26:11
Speaker 2: Extreme outlier to the norm.
26:13
Speaker 1: I mean, Trump sold his presidency to the AI tech.
26:17
Speaker 2: Oligarchs.
26:17
Speaker 1: That is entirely clear at this point. Elon giving him
26:21
Speaker 1: $ 250 million for his campaign. Remember at the inauguration who
26:25
Speaker 1: was standing behind him? It wasn't the cabinet members or
26:29
Speaker 1: elected politicians. It was the richest tech overlords on the planet.
26:33
Speaker 1: That is who is running the show. He's going to
26:35
Speaker 1: do what they want him to do. And that's what
26:37
Speaker 1: his presidency is. So that's part of why it's so
26:41
Speaker 1: important to just slow things down because we're not getting
26:45
Speaker 1: anything out of the Trump administration.
26:47
Speaker 3: They've set the course. They're all on board.
26:49
Speaker 1: For the like no holds barred, no regulation whatsoever. And so,
26:55
Speaker 1: I mean, my final plea to people would just be,
26:58
Speaker 1: let's say that you feel like maybe there's a 1%
27:03
Speaker 1: chance that some fraction of this destruction occurs. Okay, well,
27:08
Speaker 1: what's the cost from just slowing down and allowing a
27:10
Speaker 1: democratic process to play out and putting some safeguards in place?
27:13
Speaker 3: What's the harm in doing that? You don't lose anything.
27:16
Speaker 3: from doing that.
27:17
Speaker 1: Whereas on the other side, you know, the risk that's
27:20
Speaker 1: being sketched out here is so catastrophic. And to Sagar's point,
27:23
Speaker 1: I've seen a lot of people say, okay, well, how
27:24
Speaker 1: exactly would it happen?
27:25
Speaker 3: You know, it's very hard to know, right?
27:27
Speaker 1: You can't exactly predict the future, but there has been
27:30
Speaker 1: a lot of thinking about scenarios that could theoretically unfold.
27:34
Speaker 1: And so if you want to actually dig into one
27:36
Speaker 1: of the use cases, We had on one of the
27:40
Speaker 1: co-authors of the book, If Anyone Builds It, Everyone Dies,
27:43
Speaker 1: and they lay out, okay, they lay out one potential
27:46
Speaker 1: scenario for how this could all unfold, just to help
27:48
Speaker 1: you wrap your mind around it, because I understand it
27:50
Speaker 1: sounds insane. Like, I feel insane talking about it. It
27:53
Speaker 1: genuinely sounds like sci-fi. But just ask yourself the question,
27:56
Speaker 1: you know, when has it ever been the case that
27:59
Speaker 1: you've had a superior intelligence that really cares about the
28:04
Speaker 1: lower intelligence creature and like, you know, allows them to
28:07
Speaker 1: continue to do their thing and doesn't impose their will
28:10
Speaker 1: on it? What we're trying to create, what they're trying
28:13
Speaker 1: to create is something that surpasses human beings in terms
28:16
Speaker 1: of intelligence in all meaningful ways. That is a completely
28:21
Speaker 1: different situation that we have never experienced on this planet before.
28:25
Speaker 1: What is the likely fallout of that? I mean, at
28:27
Speaker 1: the very least, we should take some time to think
28:29
Speaker 1: that through. We should have requirements. Anthropic puts out this
28:33
Speaker 1: analysis of this latest safety incident. They do it out of,
28:36
Speaker 1: I guess, the goodness of their hearts. We're depending on
28:38
Speaker 1: these people to come clean. And we already have instances
28:41
Speaker 1: where they haven't. So what else has happened we don't
28:43
Speaker 1: even know about?
28:45
Speaker 8: How could it be that it is easier to build
28:50
Speaker 8: a potentially job and life and quality of life threatening
28:54
Speaker 8: AI than it is to open a food truck, which
28:57
Speaker 8: is the example that Greg Kazar gave.
28:59
Speaker 3: So, you know, that's that's where we are.
29:01
Speaker 1: And, you know, there have been there is starting to
29:04
Speaker 1: be what I'm encouraged by is the level of attention
29:08
Speaker 1: and concern, because previously these warnings have just sort of
29:11
Speaker 1: flown under the radar. They haven't reached a mass population.
29:15
Speaker 1: So I'm glad to see that this is being taken
29:17
Speaker 1: seriously now.
29:18
Speaker 3: But both on the.
29:19
Speaker 1: Domestic front and on the international cooperation front, you know,
29:22
Speaker 1: it's very discouraging where we are at the moment.
29:24
Speaker 2: Let me read, because we talk a lot about the
29:27
Speaker 2: doom case. Let me read about the positive case. This
29:29
Speaker 2: is a friend of mine, Tanner Greer, goes by Scholar Stage,
29:32
Speaker 2: really insightful guy here in Washington. And he wrote this,
29:36
Speaker 2: most politicos do not realize that the, quote, good outcome,
29:39
Speaker 2: the outcome that does not involve extinction, instead involves the following.
29:43
Speaker 2: A transhumanist future for our species, see Altman's The Merging
29:48
Speaker 2: blog post, where we meld with machines, a future where
29:51
Speaker 2: all meaningful agency is given to machines on whose grace
29:55
Speaker 2: we will thereafter depend. the general eclipse of humanity by
29:58
Speaker 2: a successor species. But they will realize these things and
30:02
Speaker 2: will realize them soon. What the industry does not realize
30:04
Speaker 2: is that to most Americans, all of those futures are
30:06
Speaker 2: horrific and disgusting. Few outside of the valley want gilded slavery.
30:11
Speaker 2: Fewer still yearn to merge with the machine. And I
30:14
Speaker 2: think that the point that he makes there is we've
30:16
Speaker 2: just spent an entire 30 minutes talking about destruction and
30:19
Speaker 2: the cybersecurity and, you know, bio labs. It's like, yeah,
30:23
Speaker 2: but this is the bull case. The bull case for
30:25
Speaker 2: it is that we do become these transhumanists who, you know,
30:29
Speaker 2: are able to turn our entire lives over to these agents.
30:33
Speaker 2: Their theory is that this will unleash human flourishing, whereas
30:37
Speaker 2: all evidence exists to the contrary about how much it
30:41
Speaker 2: will actually supplant what it means to be human itself.
30:44
Speaker 2: So I wanted to put that in there too. People say, oh,
30:46
Speaker 2: you don't cover the bull case of AI. That's the
30:48
Speaker 2: bull case right there. I don't want that. I certainly don't.
30:51
Speaker 2: And you shouldn't either. All right, let's get to the
30:53
Speaker 2: next part about math.
30:57
Speaker 1: All right, so let's talk a little bit about the
31:00
Speaker 1: ethical conduct of these AI companies today, focusing in on
31:04
Speaker 1: OpenAI in particular. So put B1 up on the screen.
31:08
Speaker 1: They put out this big announcement that they solved a
31:11
Speaker 1: new frontier math problem They say, we're sharing a solution
31:16
Speaker 1: to the Navier-Stokes Millennium Prize problem, one of the deepest
31:20
Speaker 1: problems at the frontier of mathematics. The proof was produced
31:22
Speaker 1: by a group of agents using an open AI next
31:25
Speaker 1: generation model significantly more capable than GPT-6 Astra. That's the
31:29
Speaker 1: one they just put out. The problem concerns whether the
31:31
Speaker 1: description of smooth three-dimensional fluid motion modeled by the Navier-Stokes
31:35
Speaker 1: equations can break down. It has remained unresolved. for roughly
31:39
Speaker 1: 90 years. So we have some other instances of AI
31:43
Speaker 1: solving some of these very challenging, longstanding math problems. So
31:48
Speaker 1: first glance, oh, wow, okay, another breakthrough. Not so fast.
31:52
Speaker 1: This one seems to have some caveats involved.
31:55
Speaker 3: Go ahead and put B2.
31:56
Speaker 1: Up on the screen. So this is from TechCrunch. I'm
31:59
Speaker 1: going to read you a little bit of this article.
32:02
Speaker 1: So there were two other mathematicians, among many others, but
32:05
Speaker 1: there were two other mathematicians who were working on this
32:07
Speaker 1: particular Navier-Stokes problem and were making significant progress towards a proof.
32:15
Speaker 1: So the allegation here from TechCrunch is, opening, I fought
32:18
Speaker 1: dirty on career-making math problems, says NYU mathematician. This math professor,
32:23
Speaker 1: Tristan Buckmaster, announced three proofs on Tuesday with a preliminary
32:26
Speaker 1: finding on one of the major unsolved problems in theoretical mathematics.
32:30
Speaker 1: The findings, made in collaboration with anthropic mathematician Levin Alpoge,
32:35
Speaker 1: and using both Codex, which is OpenAI, keep that in mind,
32:38
Speaker 1: and Claude AI models are significant in themselves, but they're
32:42
Speaker 1: also accompanied by an unusual controversy surrounding OpenAI's attempts to
32:45
Speaker 1: solve the same problem. There is another part of this story,
32:49
Speaker 1: Buckmaster wrote in a statement announcing the proofs. One that honestly,
32:52
Speaker 1: I very much wish I did not have to be
32:54
Speaker 1: concerned with. According to the statement, a parallel effort by
32:58
Speaker 1: OpenAI built on their work before it became public led
33:01
Speaker 1: to a tangle of academic rivalries and conflicting claims.
33:05
Speaker 3: So here is the TLDR.
33:07
Speaker 1: These two mathematicians were using Cloud AI and Codex, which
33:12
Speaker 1: is again, OpenAI. They were using this to work on
33:16
Speaker 1: a direction of the problem that other mathematicians had sort
33:18
Speaker 1: of pushed on and they were pushing forward using this AI.
33:22
Speaker 1: very close to completion on this. And suddenly they find
33:26
Speaker 1: out that OpenAI is announcing that they have solved this
33:30
Speaker 1: problem and they used the same direction to solve this
33:35
Speaker 1: problem that these two mathematicians were using. And according to
33:39
Speaker 1: the math world, this was a very unusual way to
33:43
Speaker 1: go about solving this particular problem. So lo and behold,
33:47
Speaker 1: they get in this exchange with OpenAI where they're saying
33:51
Speaker 1: basically like what's going on here and asking them, hey,
33:55
Speaker 1: did you use our sessions on your product? Did you
33:59
Speaker 1: like take those and then throw what apparently now we
34:02
Speaker 1: know was like $ 20 million of compute? at pushing through
34:06
Speaker 1: the final stages of this problem to be able to
34:10
Speaker 1: claim yourself that you did it. And of course, they
34:12
Speaker 1: deny that that is the case, although they say they
34:14
Speaker 1: cannot rule out that the model didn't use their work
34:19
Speaker 1: for its own training so that then, and this is,
34:22
Speaker 1: by the way, just like a thing that you have
34:24
Speaker 1: to sign up for when you use these products, that
34:26
Speaker 1: what you do on their platforms will be used.
34:29
Speaker 3: In their training.
34:30
Speaker 1: So they said, well, we can't say that they didn't
34:31
Speaker 1: use the training of your work on our platform in
34:36
Speaker 1: order to come up with this answer. So this is
34:40
Speaker 1: where things effectively stand. There's some wild things in here too, Saver.
34:43
Speaker 1: Like for example, when they were kind of going back
34:46
Speaker 1: and forth and there was the mathematician Tristan, he was like, well, look,
34:52
Speaker 1: I'm going to go forward publicly with all of this.
34:54
Speaker 1: And this open AI guy responds to him, well, I
34:58
Speaker 1: don't know why you'd want to ruin your career like that.
35:00
Speaker 1: I mean, just sort of like mob boss threatening tactics.
35:04
Speaker 1: There was also a lot of upset over the fact
35:06
Speaker 1: that the other mathematician involved works for Anthropic. Now, they
35:09
Speaker 1: say he wasn't doing this in any sort of official
35:11
Speaker 1: capacity for Anthropic. They were doing this on the side.
35:14
Speaker 1: There was no contract with Anthropic or anything like that.
35:16
Speaker 1: But that was a problem for OpenAI as well. So
35:19
Speaker 1: it's pretty wild situation being alleged here that they either
35:24
Speaker 1: saw that this work was going on, that it was
35:26
Speaker 1: almost close to completion, and then they threw tens of
35:28
Speaker 1: millions of dollars of compute at it to get to
35:30
Speaker 1: the problem, or they didn't realize that their.
35:34
Speaker 3: Platforms were stealing the work. of these mathematicians.
35:38
Speaker 1: And so when they asked, okay, you know, let's work
35:40
Speaker 1: on the Navy or Stokes problems, like, oh, these guys
35:42
Speaker 1: are doing it. I'll just use their work and we
35:44
Speaker 1: can finish it this way.
35:45
Speaker 2: Let's put B3 on the screen. Tristan Buckmaster actually released
35:49
Speaker 2: some of this in a written statement. So here's what
35:53
Speaker 2: he says. He says, I asked whether the model had
35:56
Speaker 2: been trained on or had access to our sessions in Codex,
36:00
Speaker 2: into which we've been putting all of our drafts for
36:02
Speaker 2: the whole of this project. I was told the model
36:04
Speaker 2: did not look up user data. I asked again about training.
36:07
Speaker 2: I did not get an answer. The first two proposals
36:11
Speaker 2: were offered to me. The first was that we post
36:13
Speaker 2: our result and that OpenAI post its result the next day.
36:16
Speaker 2: The second was that after posting, I alone write a
36:19
Speaker 2: paper representing the result, acknowledging that an internal OpenAI model
36:23
Speaker 2: had resolved it. They twice asserted that he wanted them
36:27
Speaker 2: to remove from authorship one of his co-authors and said
36:29
Speaker 2: it would be simple if only it were not the
36:31
Speaker 2: case and that it was so annoying to the other
36:33
Speaker 2: person he was working with works at Anthropic. It was
36:36
Speaker 2: also said if OpenAI posted after us, they would say
36:38
Speaker 2: that we deserve the prize and we're the closest humans
36:41
Speaker 2: to the problem. I declined both of these offers. I
36:44
Speaker 2: said that if OpenAI released its results in the way proposed,
36:48
Speaker 2: I would go public with what happened. The reply was,
36:51
Speaker 2: why would you ruin your career? I replied, I am
36:54
Speaker 2: an academic and asked why he thought going public would
36:57
Speaker 2: ruin my career. The reply was, If you don't want
36:59
Speaker 2: me to be nice, then I don't have to be nice.
37:03
Speaker 2: I mean, yeah, this is genuine, genuine like thuggery. And look,
37:10
Speaker 2: it's a math thing. No offense to the math people.
37:13
Speaker 2: I don't think my life is going to change very
37:14
Speaker 2: much because this problem has been solved. But, you know,
37:18
Speaker 2: as a history guy, this is all of the hallmarks.
37:22
Speaker 2: of this is how I know we are truly back
37:25
Speaker 2: in the Gilded Age. Because the hallmarks of the Vanderbilt,
37:29
Speaker 2: I encourage everybody to go read it. The First Tycoon,
37:31
Speaker 2: one of my favorite books I've ever read. It is
37:33
Speaker 2: a biography of Cornelius Vanderbilt. It literally is just a
37:37
Speaker 2: tale of these Gilded Age cutthroat barbarians who just like
37:42
Speaker 2: steal each other's stuff. They're constantly price cutting and warring
37:47
Speaker 2: and stealing each other's employees and then coming in and
37:51
Speaker 2: threatening the other person and suing them in court. Some
37:54
Speaker 2: of our most important court decisions actually stem from their
37:57
Speaker 2: thievery of each other and from breaking the law. I mean,
38:00
Speaker 2: that was how the steamship and eventually railroad industry itself
38:05
Speaker 2: So when I've given that famous example of railroaded, these
38:08
Speaker 2: people weren't that smart. They were ultimately, in many cases,
38:11
Speaker 2: they made titanic mistakes, but they were just cutthroat, ruthless
38:15
Speaker 2: men who built it on the backs of a lot
38:18
Speaker 2: of their workers. When I saw this, I said, wow. Like,
38:23
Speaker 2: I know exactly where this comes from. Like, this, like,
38:26
Speaker 2: why would you want to ruin... Be an awful bad thing,
38:29
Speaker 2: you know, for that to happen.
38:30
Speaker 3: Here's the thing.
38:31
Speaker 1: If you're a mathematician or you're a scientist, if you're
38:34
Speaker 1: a writer and you're running drafts through these things, like,
38:37
Speaker 1: they're stealing your shit.
38:38
Speaker 3: I mean, that's what it... That's what it appears. They
38:40
Speaker 3: deny it.
38:40
Speaker 2: Yeah. Allegedly. But.
38:42
Speaker 1: I mean, but they don't even really deny it because
38:44
Speaker 1: we know that they are using whatever you input. They're
38:46
Speaker 1: using that for training. That's how these things have been
38:49
Speaker 1: built is off of the sum total of human knowledge.
38:53
Speaker 1: So that you should think about that moving forward. And
38:57
Speaker 1: I don't know how to get around it because. They
39:00
Speaker 1: really are powerful tools in terms of, you know, if
39:02
Speaker 1: you're doing something that is brute force, requires a lot
39:05
Speaker 1: of data or trying a whole bunch of different solutions
39:09
Speaker 1: and just sort of like going through them and putting
39:11
Speaker 1: them in one by one and seeing which one works out.
39:14
Speaker 3: There's really no other way to go about it.
39:16
Speaker 1: But they I doubt this will be the last case
39:19
Speaker 1: and it may not even be the first case. Where
39:21
Speaker 1: they either intentionally or accidentally just take your thing and
39:25
Speaker 1: are like, oh, we got it. We did it over
39:26
Speaker 1: here with our AI and no real human involvement. So
39:30
Speaker 1: congratulations to us.
39:32
Speaker 3: It's really wild.
39:33
Speaker 1: I was saying it's sort of like if you wrote
39:36
Speaker 1: a novel in Microsoft Word and then you're like on
39:39
Speaker 1: the last chapter and Microsoft announces like, hey, guys, we
39:43
Speaker 1: wrote this incredible novel.
39:45
Speaker 3: Isn't it great?
39:46
Speaker 2: Whoa, whoa.
39:48
Speaker 3: Yeah.
39:48
Speaker 2: Whoa, whoa, whoa. How did that happen?
39:50
Speaker 4: Yeah.
39:50
Speaker 1: So that's kind of what was being into that here
39:54
Speaker 1: or alleged here is kind of the idea.
39:56
Speaker 3: So it really is.
39:58
Speaker 1: It's it's a very wild situation and it's hard to
40:00
Speaker 1: imagine what will prevent.
40:02
Speaker 3: Future situations like this also in the future.
40:04
Speaker 2: I can't, yeah, I mean, who knows? I also was
40:06
Speaker 2: thinking about it in terms of, because my parents are academics,
40:09
Speaker 2: so I totally get it, where these guys, it's not
40:11
Speaker 2: like you make a lot of money being a mathematician.
40:13
Speaker 2: In fact, the only way that you get to keep
40:15
Speaker 2: your job is that you go out and you get
40:19
Speaker 2: billionaires and companies like OpenAI and Anthropic to give you
40:23
Speaker 2: money to fund your research. And I'm assuming that's kind
40:26
Speaker 2: of what he really was saying there. was, hey, the
40:29
Speaker 2: math billionaire world isn't that big. Maybe I'm just going
40:33
Speaker 2: to dry up all your funding. I mean, that's scary stuff.
40:37
Speaker 2: If you want an analogy, it's very similar to what
40:39
Speaker 2: happened with gain of function, where any of the academics
40:43
Speaker 2: that were warning Fauci about gain of function or who
40:46
Speaker 2: were whistleblowing about the dangers, how because Fauci was the
40:49
Speaker 2: head of the NIAID, which controls all of the scientific funding,
40:53
Speaker 2: that he could have just cut his own opponents off.
40:56
Speaker 2: Like when you have that ability, this spigot, you know,
40:59
Speaker 2: which funds all of these projects and different things that
41:02
Speaker 2: are happening in academia, you have massive control over people's careers.
41:06
Speaker 2: So just again, to give you examples. So yeah, I
41:09
Speaker 2: see very much hints of, gilded age in all of this.
41:13
Speaker 2: And that's how you know these people really are ready
41:16
Speaker 2: to play. Not a good thing for humanity. Let's get
41:19
Speaker 2: to data centers. Turning now to data centers, there's a
41:25
Speaker 2: lot of talk about data centers. There's a lot of
41:26
Speaker 2: propaganda you might have heard. Oh, really helps all the
41:30
Speaker 2: local communities, their property taxes go down, they're flooded with money.
41:35
Speaker 2: These are all the things that we're told all of
41:37
Speaker 2: it was from industry sources. Well, luckily for our purposes,
41:41
Speaker 2: there's actually now been a systematic review of all the evidence.
41:44
Speaker 2: Let's put it up here on the screen. I read
41:47
Speaker 2: this entire paper. I encourage you to, if you want to,
41:49
Speaker 2: it's in the social sciences journal. It's literally publicly, maybe
41:54
Speaker 2: we'll even put a link In the description, you can
41:56
Speaker 2: read it for yourself. So it's called Capital Without Labor,
41:59
Speaker 2: Data Centers, and the Local Economy. So let me read
42:02
Speaker 2: you some of the findings. What they ultimately found through, again,
42:05
Speaker 2: this review, and I'll even explain the methodology because it
42:08
Speaker 2: is actually pretty interesting. So the takeaway is basically that
42:13
Speaker 2: they were able to look at counties with data centers
42:16
Speaker 2: and then compare them to counties without. But you can't
42:19
Speaker 2: just compare apples to oranges. You want to try and
42:21
Speaker 2: compare apples to apples. So they actually looked at counties
42:25
Speaker 2: with data centers and then compared them to counties which
42:28
Speaker 2: had been heavily considered for data centers, but ultimately the
42:31
Speaker 2: project did not go forward. So we know that it
42:33
Speaker 2: was a desirable place to put a data center. They
42:36
Speaker 2: looked at this over, I think, a 20-year period. Now, obviously,
42:40
Speaker 2: their previous data centers, like the YouTube-type data centers, the
42:42
Speaker 2: smaller ones, but they ended their data set in 2025.
42:46
Speaker 2: The data center boom began in 2023. So a lot
42:47
Speaker 2: of the data is actually pretty good. I will give
42:49
Speaker 2: you some of the caveats later, but I'm just setting
42:52
Speaker 2: it up. So capital without labor. Here are the key
42:54
Speaker 2: findings as I summarize them. Number one, there was no
42:57
Speaker 2: increase in the overall financial health of these communities. Okay.
43:02
Speaker 2: Number two, there was no increase in local employment or
43:06
Speaker 2: new business formation. I actually think that's the single most
43:08
Speaker 2: important one. Because network effects of companies coming in to
43:13
Speaker 2: your county is the single reason why you want it.
43:16
Speaker 2: And everybody knows this. So, for example, whenever factories would
43:19
Speaker 2: close down in Michigan. One of the things that would
43:23
Speaker 2: devastate the town was not that the factory was gone
43:26
Speaker 2: and the jobs. It was the local supplier to the
43:29
Speaker 2: factory also went away. It was the laundromat which does
43:33
Speaker 2: all of the laundry for the guys who work there
43:37
Speaker 2: closes down. It was the seat leather manufacturer 20 miles away.
43:41
Speaker 2: He had no longer, like that's what devastates a community.
43:46
Speaker 2: In this case, and that's why the paper is called
43:49
Speaker 2: Capital Without Labor, is they find that while it may
43:51
Speaker 2: take 1,000 plus people to build a data center, the
43:54
Speaker 2: permanent number of employees there is only 20 to 30 people.
43:57
Speaker 2: And so the network effect of having it there doesn't
44:00
Speaker 2: do very much. Okay, so that's number two. Number three,
44:03
Speaker 2: and this was the most interesting finding to me, is
44:05
Speaker 2: that local government borrowing costs actually went up. And that
44:09
Speaker 2: is because housing price growth in the towns where they
44:13
Speaker 2: build one, it goes down because people don't want to
44:16
Speaker 2: live around them. And that when people don't want to
44:19
Speaker 2: live around them, and housing prices don't go down, then
44:23
Speaker 2: schools actually end up having to squeeze more from the
44:26
Speaker 2: property tax base. Because if your house value is not
44:30
Speaker 2: going to continue to rise in the way that it
44:32
Speaker 2: normally has, that means you need to increase the overall
44:34
Speaker 2: proportion of property tax on the not decreasing asset, but
44:38
Speaker 2: the less appreciating asset of housing. So Ultimately, put it
44:42
Speaker 2: all together. No increase in the overall financial health of
44:45
Speaker 2: the county, no increase in local employment, local government borrowing
44:48
Speaker 2: costs increase, housing price growth slows, schools end up squeezing
44:52
Speaker 2: more from property tax. There's a water element to this,
44:54
Speaker 2: which I want to note. It's a little complicated, but
44:57
Speaker 2: basically in the counties where there is water scarcity. The
45:01
Speaker 2: borrowing and the bond yields on water bonds specifically went up,
45:06
Speaker 2: meaning that they saw more risk to the overall water infrastructure.
45:10
Speaker 2: I'm being very careful in my language. I'm not saying
45:12
Speaker 2: they're guzzling up all the water or any of that. Basically,
45:15
Speaker 2: they're putting strain on the infrastructure, requiring more money to
45:19
Speaker 2: be built in. Now, here are the caveats of which
45:23
Speaker 2: I would be foolish not to mention. this is a
45:25
Speaker 2: cutoff in 2025. We don't have long-term data. The promises
45:30
Speaker 2: of the industry will take five to 10 years to materialize.
45:34
Speaker 2: But I am just saying that, and by the way,
45:37
Speaker 2: not a single data center proponent came and said, this
45:42
Speaker 2: is BS, this is bad research, or any of that.
45:45
Speaker 2: They gave the caveats that I just gave you about
45:47
Speaker 2: the timeframe and about how it doesn't capture all that stuff.
45:50
Speaker 2: And there are some minute things that you can quibble with,
45:52
Speaker 2: but nobody is disputing that this is the finding as
45:56
Speaker 2: of right now. Okay. So, I mean, I feel like
45:58
Speaker 2: this is one of the most important papers for the discourse.
46:02
Speaker 2: And instead, everybody just wants to cite New York Times
46:05
Speaker 2: articles about how one place, Loudoun County. is doing so well.
46:08
Speaker 2: And it's like, well, here's the evidence from thousands of
46:11
Speaker 2: data centers across the nation. I have it right here
46:13
Speaker 2: in front of you.
46:14
Speaker 3: Loudoun, already a wealthy county, by the way.
46:16
Speaker 2: These people are filthy rich defense contractors. It's crazy.
46:20
Speaker 1: Worth noting, too, that they may not be representative of
46:24
Speaker 1: the average data center experience.
46:26
Speaker 3: It's almost like you're saying.
46:28
Speaker 1: that if you support data centers in your area, you
46:31
Speaker 1: want to be backwards and poor.
46:32
Speaker 3: It's almost like you're saying that.
46:34
Speaker 1: No, I mean, it's pretty devastating when you see this,
46:36
Speaker 1: because this is the one thing that the data center
46:38
Speaker 1: proponents have really had. And I think a lot of
46:40
Speaker 1: people are just sort of looking at it like, well,
46:43
Speaker 1: what would be the problem? Why wouldn't you want this
46:45
Speaker 1: economic job creation in your town and all of this
46:48
Speaker 1: additional tax revenue? And they say they're going to deal
46:51
Speaker 1: with the electricity issue.
46:52
Speaker 3: Don't worry about it. The water issues are overblown. This
46:55
Speaker 3: is just economic development.
46:57
Speaker 1: And it's kind of amazing how sometimes the just natural
47:00
Speaker 1: gut instinct of the American people is more aligned with reality,
47:04
Speaker 1: not always, but sometimes, than what the intelligentsia, the beltway intelligentsia,
47:12
Speaker 1: want to sell to you. I wanted to also pull up,
47:15
Speaker 1: there's some polling here about Bernie and Greg Kassar have
47:19
Speaker 1: put forward this bill to pause AI development, which was
47:22
Speaker 1: framed as this very radical idea in the Beltway and
47:26
Speaker 1: sort of like legacy media and among intellectuals here in D.C. Well,
47:31
Speaker 1: as I predicted, and I'm sure you would have expected
47:33
Speaker 1: as well, Sagar, this is a very common sense proposal
47:37
Speaker 1: to the vast majority of Americans who are like, I
47:40
Speaker 1: don't know where this is leading, but I know that
47:42
Speaker 1: I don't trust these tech oligarchs. So maybe we should
47:46
Speaker 1: just take a pause and figure this thing out. And
47:48
Speaker 1: what is going on? What is the impact going to be?
47:50
Speaker 1: And who is going to have control? And is there
47:52
Speaker 1: going to be some public ownership? How is this all
47:54
Speaker 1: going to work? So that specific bill to pause AI
47:57
Speaker 1: development and permanently ban the creation of AI superintelligence is
48:01
Speaker 1: supported by 68% of voters, 72% of Democrats, 70% of independents,
48:08
Speaker 1: and even 63% of Republicans. You only have 25% of
48:15
Speaker 1: voters who oppose this, and the remainder are undecided. So 68% say, yeah,
48:23
Speaker 1: for the pause, I don't know why we would want
48:25
Speaker 1: to create superintelligence.
48:26
Speaker 3: I'm good with that.
48:27
Speaker 1: And only 25% who are opposed, and only 9%, by
48:31
Speaker 1: the way, who strongly oppose. So that's where the politics
48:35
Speaker 1: of this thing reside right now. And it is one
48:38
Speaker 1: more issue where there is just such a massive disconnect
48:42
Speaker 1: between DC elected officials and the broad sensibilities of the public.
48:48
Speaker 1: It is just hanging out there as this incredibly polarized,
48:52
Speaker 1: incredibly powerful issue for someone to be able to grab
48:56
Speaker 1: onto and really speak to the concerns of the American people.
48:59
Speaker 1: And I think it ties in With the data center concerns,
49:02
Speaker 1: although people have a variety of.
49:03
Speaker 3: Issues and concerns.
49:04
Speaker 2: Yeah. And I also wanted to just show the wide
49:06
Speaker 2: range of how rapidly things are now moving in America.
49:09
Speaker 2: So let's start with Ro Khanna, who immediately put out
49:13
Speaker 2: a long statement about multiple proposals in reaction, both not
49:18
Speaker 2: just the data center issue, but about AI. Let's take
49:20
Speaker 2: a listen to him.
49:22
Speaker 7: We need action. There are five things we must do. First,
49:28
Speaker 7: I've been calling for an AI agency, a federal AI
49:34
Speaker 7: safety agency. just like we have an agency to regulate
49:38
Speaker 7: nuclear power and to regulate our airlines. Second, there has
49:44
Speaker 7: to be a safety check to make sure that any
49:48
Speaker 7: agentic AI system is contained before it goes on the
49:53
Speaker 7: open network.
49:54
Speaker 2: So you could see that's a flavor of Congressman Ro
49:57
Speaker 2: Khanna to give you.
49:58
Speaker 3: Always worth remembering, this man represents Silicon Valley.
50:01
Speaker 2: Right. Okay.
50:03
Speaker 1: So if he can take a stand on this, no
50:06
Speaker 1: one else has an excuse in Washington whatsoever because the
50:10
Speaker 1: tech oligarchs and many of the workers at these companies
50:12
Speaker 1: too are literally his constituents.
50:15
Speaker 2: And then just to give you a taste too, here's
50:17
Speaker 2: one of the most pro-free market conservatives, Ted Cruz, reacting
50:21
Speaker 2: actually to this. C7, let's take a listen.
50:24
Speaker 6: I can tell you there's legislation that I'm working on
50:27
Speaker 6: with Amy Klobuchar and John Thune to try to address
50:31
Speaker 6: catastrophic risk. It is dangerous.
50:33
Speaker 2: It is frightening. How? Explain how that would happen.
50:36
Speaker 6: For AI that would implicate biological issues or nuclear issues
50:42
Speaker 6: that have the potential to have catastrophic risk, it would
50:44
Speaker 6: require government supervision and approval before they can be implemented.
50:48
Speaker 2: Will it pass?
50:50
Speaker 6: It's not clear. So so Amy and John are negotiating
50:53
Speaker 6: right now. I mean, look, this is it is possible
50:56
Speaker 6: to move bipartisan legislation. And actually, the Commerce Committee, which
50:59
Speaker 6: I chair, has moved more bipartisan legislation than any committee
51:02
Speaker 6: in the Senate.
51:03
Speaker 2: So it is possible.
51:04
Speaker 6: But it ain't easy.
51:05
Speaker 2: This is a very divided time. And is it is
51:07
Speaker 2: the is the horse out of the barn already? Look, whoopee,
51:10
Speaker 2: that is a good question.
51:12
Speaker 6: And one of the real challenges is China is going
51:14
Speaker 6: going full speed ahead. And whatever we do here. China's
51:18
Speaker 6: not going to stop. So I've said somewhat tongue in cheek,
51:21
Speaker 6: but not entirely. If they're going to be killer robots,
51:25
Speaker 6: I'd rather they be American killer robots than Chinese killer robots.
51:30
Speaker 4: All right.
51:30
Speaker 1: So you could see still the Chinese killer robot. I
51:34
Speaker 1: don't really care whether I get murdered by a Chinese
51:36
Speaker 1: or an American killer robot.
51:38
Speaker 3: Personally, it all seems bad to me, Sagar.
51:41
Speaker 4: Right.
51:41
Speaker 2: And I mean, that's my last point before we move
51:45
Speaker 2: to the midterms. is there is not actual evidence that
51:50
Speaker 2: the Chinese want any sort of, quote, autonomous robot. In fact,
51:55
Speaker 2: there is only one paper that has ever been published
51:58
Speaker 2: by the People's Liberation Army talking about AI warfare. I
52:02
Speaker 2: read it yesterday. And in it, it's honestly very nuanced.
52:06
Speaker 2: They're like, yeah, this will present us opportunities. Autonomous weapons
52:10
Speaker 2: and autonomous drone systems have tremendous capabilities on the battlefield,
52:14
Speaker 2: as we've seen in Ukraine. However, significant questions remain about
52:18
Speaker 2: control and about our ability to make sure that we
52:21
Speaker 2: have the ability to have any sort of say. And
52:25
Speaker 2: for these weapon systems to then turn on... This is
52:27
Speaker 2: the official, again, the official People's Liberation Army. It's like
52:30
Speaker 2: two colonels, I think, who are writing about this in Chinese.
52:34
Speaker 2: Any of you can go and read it right now.
52:35
Speaker 2: There's an English translation, which has been made available by
52:39
Speaker 2: the RAND Corporation. And reading through it, I was like,
52:42
Speaker 2: I don't see Iranian endorsement of autonomous robots. I mean, look,
52:45
Speaker 2: I could be naive in that they're going all in,
52:48
Speaker 2: but I look at the way that they're controlling the system,
52:50
Speaker 2: the fact that that is the only published piece yet
52:53
Speaker 2: so far I just feel like we've created this Schrodinger's China.
52:58
Speaker 2: where it's like, you know, it is or isn't this
53:01
Speaker 2: dramatic threat, and nobody is ever, nobody's actually looking at, like,
53:04
Speaker 2: what are they doing? You know, they're not racing towards,
53:08
Speaker 2: or they're not in some mission to race towards AGI.
53:12
Speaker 2: They have one company, which is. They have a lot
53:14
Speaker 2: of companies, which aren't. Their government seems very skeptical and
53:18
Speaker 2: semi-curious about the way that things are happening, but very
53:21
Speaker 2: ready to shut things down. I don't see a lot
53:23
Speaker 2: of evidence that it's some Sputnik, Soviet-style Star City, if
53:28
Speaker 2: you watch that show, type of program where they're like,
53:31
Speaker 2: we're going to pursue this at all costs. I do
53:34
Speaker 2: not see that whatsoever.
53:35
Speaker 1: Just let's put C3B up on the screen. This is
53:37
Speaker 1: my last point with regard to the politics. Right now,
53:40
Speaker 1: Democrats are being far more outspoken than Republicans. Not hard
53:44
Speaker 1: to understand why. Number one, Trump is all in on AI.
53:48
Speaker 1: So this is a tabulation.
53:49
Speaker 3: At least 22 politicians responded directly to.
53:51
Speaker 1: The tweet about the AI researcher who resigned and said, like,
53:56
Speaker 1: this is an existential threat. You had 22 politicians respond.
54:00
Speaker 1: Three of them were Republicans. The rest of them were Democrats.
54:03
Speaker 1: And it's not hard to understand why you have Trump,
54:06
Speaker 1: who was all in on this. You have a Republican
54:08
Speaker 1: Party that is very much in league with him. And
54:11
Speaker 1: you also have, you know, the tech world used to
54:14
Speaker 1: be seen as very, like, liberal Democratic, especially Obama era.
54:18
Speaker 1: They were all giving money to Obama and doing stuff
54:20
Speaker 1: at the White House and partnering with him even in
54:22
Speaker 1: his campaign.
54:23
Speaker 3: They have completely switched and gone all in for Trump.
54:27
Speaker 1: And so now the top echelons of the tech companies
54:30
Speaker 1: are very much their right wing.
54:32
Speaker 3: They're right wing coded like that is.
54:34
Speaker 1: So they have less sway with both the Democratic Party
54:37
Speaker 1: elites because they're not getting the money they used to
54:39
Speaker 1: get from tech world and also with the Democratic Party base.
54:43
Speaker 3: So if Democrats do.
54:44
Speaker 1: Take control of the House and or the Senate, then
54:48
Speaker 1: it's going to be, OK, how does the rubber meet
54:50
Speaker 1: the road? They're saying they're going to do a lot
54:52
Speaker 1: of investigations into this, that there are going to be hearings.
54:55
Speaker 1: The bigger question is, OK, who runs in 2028? What
54:57
Speaker 1: does the platform look like there? But, you know, that's
55:00
Speaker 1: going to be a major political question going forward is
55:04
Speaker 1: whether or not the Democratic Party actually has the stones
55:08
Speaker 1: to stand up to these titans of oligarchy and and
55:13
Speaker 1: actually push back on the major capital class at this
55:16
Speaker 1: point because that's what is going to be required in
55:20
Speaker 1: order to rein this stuff in. Even as some of
55:22
Speaker 1: them pose like, oh, we would so love regulation. Look
55:25
Speaker 1: at how they spend their lobbying cash. Look at who
55:27
Speaker 1: the money goes to from their super PACs. And it
55:30
Speaker 1: paints a very different picture of what they actually want
55:32
Speaker 1: out of Washington.
55:33
Speaker 2: Very, very true.