Latest / Breaking Points with Krystal and Saagar / 8/31/26: Don Jr Defends Prediction Markets, Milo Deported, AI Hacking Spree

Transcript

8/31/26: Don Jr Defends Prediction Markets, Milo Deported, AI Hacking Spree · Breaking Points with Krystal and Saagar

• 0:00
  Speaker 1: Hey guys, Sagar and Crystal here.
• 0:01
  Speaker 2: Independent media just played a truly massive role in this election,
• 0:05
  Speaker 2: and we are so excited about what that means for
• 0:07
  Speaker 2: the future of this show.
• 0:08
  Speaker 1: This is the only place where you can find honest
• 0:10
  Speaker 1: perspectives from the left and the right that simply does
• 0:13
  Speaker 1: not exist anywhere else.
• 0:14
  Speaker 2: So if that is something that's important to you, please
• 0:17
  Speaker 2: go to breakingpoints.com, become a member today, and you'll get
• 0:20
  Speaker 2: access to our full shows, unedited, ad-free, and all put
• 0:23
  Speaker 2: together for you every morning in your inbox.
• 0:25
  Speaker 1: We need your help to build the future of independent
• 0:27
  Speaker 1: news media, and we hope to see you at BreakingPoints.com.
• 0:33
  Speaker 1: Turning now to the betting markets. Betting. See, I use
• 0:37
  Speaker 1: that word importantly. That's what Kalshi and Polymarket have said
• 0:41
  Speaker 1: that they are not doing. They're offering event contracts. They're
• 0:43
  Speaker 1: not betting, guys. Let's put this up here on the screen.
• 0:47
  Speaker 1: Music to my ears. Prediction markets should be regulated as
• 0:51
  Speaker 1: gambling in appeals court now says this outcome contradicts an
• 0:55
  Speaker 1: earlier appeals court decision. Federal regulators said that the split
• 0:58
  Speaker 1: rulings called for a resolution by the Supreme Court. So
• 1:01
  Speaker 1: let me just take a step back and explain it.
• 1:03
  Speaker 1: So the U.S. Court of Appeals for the Ninth Circuit
• 1:05
  Speaker 1: in San Francisco rejected Kalshi's request for relief against Nevada
• 1:10
  Speaker 1: gambling laws after the state sued the company in February for, quote,
• 1:14
  Speaker 1: unlicensed operations. Remember, Kalshi continues to say they are not
• 1:19
  Speaker 1: a sports betting or betting platform. They are an event
• 1:22
  Speaker 1: contracts business, that event contracts business should be regulated solely
• 1:26
  Speaker 1: by the CFTC and that states, importantly, who have been
• 1:29
  Speaker 1: given the authority to regulate gambling by both the pass
• 1:32
  Speaker 1: of a decision by the Supreme Court, what it means
• 1:36
  Speaker 1: is that they are able, under their purview, to regulate,
• 1:39
  Speaker 1: let's say, casinos. They're able to regulate sportsbooks. They're able
• 1:42
  Speaker 1: to put taxes in place to be able to capture
• 1:45
  Speaker 1: some of that revenue. Kalshi has rejected this from the get-go,
• 1:48
  Speaker 1: saying you have no right to regulate us And the
• 1:51
  Speaker 1: CFTC under Donald Trump has gone to bat a number
• 1:54
  Speaker 1: of times to help support the company. Remember that in
• 1:57
  Speaker 1: one case that the CFTC issued emergency authority, which had
• 2:02
  Speaker 1: not been seen in years, ordering Kalshi to continue operating.
• 2:06
  Speaker 1: in New York State after a New York court said
• 2:08
  Speaker 1: that they were an unlicensed illegal gambling platform. So again,
• 2:12
  Speaker 1: the basics of all of this is CalSheet doesn't want
• 2:14
  Speaker 1: to be regulated by the state. Number one, they don't
• 2:16
  Speaker 1: want to pay state taxes. Number two, they want the
• 2:18
  Speaker 1: feds to continue to protect them because it means that
• 2:21
  Speaker 1: they can remain legal in Texas and California, the crown
• 2:24
  Speaker 1: jewels of the legal gambling market, which both do not
• 2:26
  Speaker 1: have legalized gambling at a sportsbook level, but are able
• 2:29
  Speaker 1: to bet on CalSheet and on PolyMarket, which of course
• 2:32
  Speaker 1: they want to capture all of that revenue. They are
• 2:34
  Speaker 1: heading into an NFL season, which by all accounts is
• 2:38
  Speaker 1: going to be an absolute bombshell record. By the way,
• 2:41
  Speaker 1: you're a tennis fan, right? Yeah. You'll be happy to
• 2:43
  Speaker 1: know the U.S. Open just partnered with Cal State. They
• 2:45
  Speaker 1: had not had a betting partner previously, but everyone will
• 2:49
  Speaker 1: be happy to know as things kick off and flushing
• 2:52
  Speaker 1: that the U.S. Open storied institution has decided to partner
• 2:54
  Speaker 1: with Cal State. Along with a number of MLB teams.
• 2:57
  Speaker 1: So everyone's just, you know, really just it's a beautiful
• 3:00
  Speaker 1: thing to watch, actually. But what's happened here is that
• 3:04
  Speaker 1: the judge has decided to actually just read very, very,
• 3:09
  Speaker 1: very few times that you see this. And the judge said, well,
• 3:13
  Speaker 1: you know, I think you are a legal sports betting
• 3:16
  Speaker 1: platform because and let's put this up here on the screen.
• 3:19
  Speaker 2: Love this.
• 3:20
  Speaker 1: Circuit Judge. CalSheX LLC advertises itself, quote, as the first
• 3:25
  Speaker 1: app for legal sports betting in all 50 states. The
• 3:29
  Speaker 1: judge continues. As the volume of activity on CalShe's sports
• 3:33
  Speaker 1: betting platform ballooned, the Nevada Gaming Control Board sent a
• 3:36
  Speaker 1: cease and desist letter notifying CalShe was violating Nevada statute
• 3:40
  Speaker 1: and gaming regulations. Kalshi sought injunctive relief, arguing it is
• 3:44
  Speaker 1: not a legal sports betting platform, but a designated contract
• 3:47
  Speaker 1: market under the Commodity and Exchange Act offering legal sports
• 3:50
  Speaker 1: events contract. Because we disagree with Kalshi's overly broad reading
• 3:55
  Speaker 1: of the CEA, and because CFTC regulations currently prohibit offering
• 3:58
  Speaker 1: contracts related to gaming on prediction markets, we affirm the
• 4:01
  Speaker 1: district court order dissolving the injunction to sports events contracts,
• 4:05
  Speaker 1: we were manned for the district court to consider Kalshi's
• 4:07
  Speaker 1: election contract. So there's a little bit of a difference there.
• 4:10
  Speaker 1: But I just love that all the judge does is
• 4:13
  Speaker 1: cite their advertisement saying we are the first app for
• 4:16
  Speaker 1: legal sports betting in all 50 states. By the way,
• 4:18
  Speaker 1: Kalshi says that they've pulled the ad and they no
• 4:20
  Speaker 1: longer support it. But anyway, just saying. I mean, the
• 4:25
  Speaker 1: whole thing is just, it's ridiculous because we all know
• 4:29
  Speaker 1: what it is. And just because you're not a sports
• 4:32
  Speaker 1: book who's setting the odds in the way that DraftKings
• 4:36
  Speaker 1: or FanDuel are, does not mean that you are not
• 4:39
  Speaker 1: offering a means to gamble. In fact, I saw a
• 4:41
  Speaker 1: perfect corollary to this. Do you know how horse betting works?
• 4:46
  Speaker 1: A lot of people don't. So it's kind of interesting, actually.
• 4:49
  Speaker 1: It's parimutuel, which means that the odds are not set
• 4:53
  Speaker 1: by the house. The odds are set by the amount
• 4:56
  Speaker 1: of money that's in the pool. So if an overwhelming
• 4:58
  Speaker 1: favorite is there, you can take the other side of that.
• 5:01
  Speaker 1: And the odd does not actually lock until the horse
• 5:05
  Speaker 1: goes from, I don't know a ton about it, but
• 5:06
  Speaker 1: when they release from the gate, I'll call it that,
• 5:08
  Speaker 1: the horse people will get mad at me. That is
• 5:11
  Speaker 1: still considered gambling, even though the book is not setting
• 5:15
  Speaker 1: the price. And that's one of the main distinctions that
• 5:17
  Speaker 1: Kalshi makes as to why they are not gambling. But
• 5:20
  Speaker 1: it's regulated as gambling because everybody knows. In fact, that's
• 5:25
  Speaker 1: the oldest form of gambling in the United States is
• 5:27
  Speaker 1: gambling on horses. But same thing. It's not a casino
• 5:30
  Speaker 1: being like, oh, this horse has eight to one Kalshi
• 5:33
  Speaker 1: is very similar. It's an event contract market, which is
• 5:37
  Speaker 1: live traded with two sides on the other. It's not
• 5:39
  Speaker 1: exactly like horse betting, but.
• 5:41
  Speaker 2: Horse betting- It's pretty analogous, though.
• 5:43
  Speaker 1: It's analogous enough, I think, to the- There's some technical
• 5:46
  Speaker 1: breakdown in terms of the exact market mechanism, but everybody
• 5:49
  Speaker 1: understands that horse betting is betting, is gambling regulated by
• 5:53
  Speaker 1: the state. It's effectively the same mechanism.
• 5:55
  Speaker 2: It's also very much the same way that people look
• 5:57
  Speaker 2: at Polymer.
• 5:58
  Speaker 1: Exactly. Absolutely.
• 6:00
  Speaker 2: People's understanding of what they're doing when they're putting money
• 6:02
  Speaker 2: down on whether or not it's going to rain on
• 6:04
  Speaker 2: Sunday or whatever they're betting on. So in terms of
• 6:07
  Speaker 2: the court mechanics here, this is a circuit court ruling,
• 6:10
  Speaker 2: which is very significant. But it sets up a conflict
• 6:14
  Speaker 2: between there was a different circuit court ruling that went
• 6:17
  Speaker 2: in the other direction, makes it even more likely that
• 6:20
  Speaker 2: this is ultimately headed to the Supreme Court. Where I
• 6:23
  Speaker 2: am not at all confident.
• 6:24
  Speaker 1: Oh, no, we're hosed.
• 6:25
  Speaker 2: They're going to rule in the correct manner because, I mean,
• 6:29
  Speaker 2: they you know, the conservative justices have found themselves at
• 6:33
  Speaker 2: odds over a number of things. They do not typically
• 6:36
  Speaker 2: find themselves at odds over deregulating business and letting business
• 6:39
  Speaker 2: do whatever they want. And they've also shown themselves to
• 6:42
  Speaker 2: be pretty slavishly concerned about. about what this administration wants
• 6:47
  Speaker 2: from them, going out of their way to give this
• 6:50
  Speaker 2: administration what it wants, frequently through the shadow docket and
• 6:53
  Speaker 2: trying not to provoke their ire as much as possible.
• 6:57
  Speaker 2: And the Trump administration and Don Jr., have very specifically
• 7:01
  Speaker 2: made it clear that they do not want states to
• 7:04
  Speaker 2: be able to regulate these markets. Don Jr. is involved
• 7:09
  Speaker 2: both with PolyMarket and with CalSheet. You've had states, both
• 7:13
  Speaker 2: red and blue states, almost all of them, 44 states,
• 7:16
  Speaker 2: signed a letter saying, hey, we want the ability to
• 7:18
  Speaker 2: be able to regulate these things. We disagree with court
• 7:21
  Speaker 2: decisions that go in the other direction. Don Jr. at
• 7:25
  Speaker 2: a meeting of Republican state officials said, came in and said, hey,
• 7:28
  Speaker 2: we want you to, I'm paraphrasing, hey, we want you
• 7:31
  Speaker 2: to back off. We want this thing to go forward,
• 7:34
  Speaker 2: and we do not want you standing in the way.
• 7:36
  Speaker 2: So they've made it incredibly clear where they stand on
• 7:38
  Speaker 2: this issue.
• 7:39
  Speaker 1: Yeah, I actually have that, D5. This is an important
• 7:42
  Speaker 1: point of the story. And Kalshi, by the way, is
• 7:44
  Speaker 1: fully prepared for Democrats to take power. So don't be
• 7:48
  Speaker 1: surprised that they're not just in the bed with Trump Jr.
• 7:51
  Speaker 1: But to your point, he gave a speech to a
• 7:53
  Speaker 1: bunch of Republican state attorney general. Remember, this is just
• 7:56
  Speaker 1: the child of the president of the United States who
• 7:59
  Speaker 1: just decides to give a speech at a gathering to
• 8:02
  Speaker 1: these Republican attorney generals and basically ask them, hey, go
• 8:05
  Speaker 1: ahead and step off of these prediction markets. Obviously, he
• 8:09
  Speaker 1: has a dramatic conflict of interest in that recommendation since
• 8:13
  Speaker 1: he's literally on the board and or a paid advisor.
• 8:16
  Speaker 1: to them. But Kalshi has also prepared themselves by hiring
• 8:20
  Speaker 1: people like the ex-Biden aide, Stephanie Cutter, a number of
• 8:23
  Speaker 1: other Democratic operatives at the state level. They are dramatically
• 8:28
  Speaker 1: lobbying and giving money out, including from their investors, to
• 8:34
  Speaker 1: the tune of tens of millions of dollars. And I mean,
• 8:38
  Speaker 1: I think with this story, if you want to go
• 8:40
  Speaker 1: and read it, it's fascinating because it breaks down exactly
• 8:43
  Speaker 1: what I'm talking about. It goes to the breakdown of
• 8:46
  Speaker 1: of states regulated gambling versus the legal framework. And the
• 8:50
  Speaker 1: reason why everyone should care is the government, is the
• 8:54
  Speaker 1: federal government stepped in and basically legalized gambling in all
• 8:58
  Speaker 1: 50 states without even the express approval. Now, as people know,
• 9:01
  Speaker 1: I'm opposed to all betting, period. I don't think there
• 9:03
  Speaker 1: should be any online gambling whatsoever. However, if the people
• 9:07
  Speaker 1: of the state want it, it's my job to like
• 9:10
  Speaker 1: go and convince the actual voters of the state. But
• 9:13
  Speaker 1: in this case, it's a regulatory loophole that they're using
• 9:17
  Speaker 1: to the tune of billions to dramatically enrich Tarek Mansour,
• 9:21
  Speaker 1: the CEO and his co-founder, who is one of the
• 9:23
  Speaker 1: youngest female billionaires in America. True, a true American dream story,
• 9:27
  Speaker 1: why everyone wants to go to MIT so they can
• 9:29
  Speaker 1: become a billionaire on the backs of a bunch of
• 9:32
  Speaker 1: gambling addicts. But what you can actually see in the
• 9:35
  Speaker 1: story is not just the regulatory capture from Kalshi and
• 9:39
  Speaker 1: they're going together, but it's the sports leagues too. That's
• 9:41
  Speaker 1: why I brought up US Open, NFL, MLB. I mean,
• 9:45
  Speaker 1: we are going into a cycle where this NFL season
• 9:50
  Speaker 1: will eclipse everything in modern history, almost certainly combined, it
• 9:55
  Speaker 1: will eclipse every other developed country in the world. And
• 9:59
  Speaker 1: we became, a nation of gambling addicts, eclipsing the amount
• 10:03
  Speaker 1: bet than any other form of entertainment combined in the
• 10:06
  Speaker 1: entire United States in a matter of seven years. Seven
• 10:09
  Speaker 1: years is all it took for us to overtake. I mean,
• 10:12
  Speaker 1: gambling's been legal, sports betting and all this in Europe
• 10:14
  Speaker 1: for decades. And that's all it took to basically take
• 10:18
  Speaker 1: over the minds of so many of these young guys.
• 10:21
  Speaker 1: And the damage will be devastating. We're watching it already
• 10:26
  Speaker 1: sweep across the nation. I just wanted to have my
• 10:28
  Speaker 1: brief victory lap because I am almost 100% sure. the
• 10:31
  Speaker 1: court is going to side with Cal State and Polymarket.
• 10:33
  Speaker 1: And so everybody just needs to get ready. This is
• 10:34
  Speaker 1: what it's going to look like. It's going to be
• 10:35
  Speaker 1: legal everywhere. They're going to side on behalf of these companies.
• 10:38
  Speaker 1: And the one saving grace that we do have, Congress
• 10:41
  Speaker 1: can still act. All Congress has to do is say,
• 10:44
  Speaker 1: you are now regulated as gambling, and it can end.
• 10:47
  Speaker 1: But of course, they're pouring billions, literally billions combined, their industry,
• 10:52
  Speaker 1: the crypto industry, the tech industry, into all of these
• 10:55
  Speaker 1: seats to the point where even a lot of these
• 10:58
  Speaker 1: Democratic senators, they got nothing to say whenever it comes
• 11:01
  Speaker 1: to calci or polymarket because they want the money.
• 11:03
  Speaker 2: Yeah, I mean, we've pretty successfully made AIPAC really toxic
• 11:06
  Speaker 2: to the point that even Mike Rogers in Michigan was like, hey,
• 11:10
  Speaker 2: you know, give me money, but do it like kind
• 11:12
  Speaker 2: of under the table because I can't be associating with
• 11:14
  Speaker 2: you people. That has really broken through to the American
• 11:17
  Speaker 2: people in a way that has, I think, made it
• 11:19
  Speaker 2: more difficult for them. Obviously, they're still spending tons of money.
• 11:22
  Speaker 2: But it's taking much more money for them to succeed.
• 11:24
  Speaker 2: And they are even when they throw tens of millions
• 11:27
  Speaker 2: of dollars at Haley Stevens, for example, they weren't even
• 11:29
  Speaker 2: able to succeed in that instance. So that has really
• 11:32
  Speaker 2: changed the game there. We need to have the same
• 11:34
  Speaker 2: level of scrutiny and awareness around these gambling packs and
• 11:37
  Speaker 2: around AI, like tech packs, because those are actually in crypto,
• 11:42
  Speaker 2: like that whole world. Those are some of the largest
• 11:46
  Speaker 2: funders in politics. A-PAC, the Israel lobby, and then this
• 11:51
  Speaker 2: suite of sort of tech-focused PACs are massively, massively spending
• 11:57
  Speaker 2: in the midterms. And so we need to get the
• 11:59
  Speaker 2: public to be aware of what's going on there as well.
• 12:02
  Speaker 2: And given the relatively bleak landscape of what's likely to
• 12:06
  Speaker 2: happen with the Supreme Court, this has to be a
• 12:09
  Speaker 2: major focus in 2028 when we are vetting presidential candidates,
• 12:13
  Speaker 2: when we are vetting any sort of Senate candidates. or
• 12:16
  Speaker 2: congressional candidates as well. But at the presidential level in particular,
• 12:19
  Speaker 2: we've got to pay attention to what's going on.
• 12:20
  Speaker 1: And the sad part is, and this is my last thing,
• 12:22
  Speaker 1: is that this is one of the only shows you're
• 12:24
  Speaker 1: going to hear on this because everybody else is sponsored
• 12:26
  Speaker 1: by Kalshi. Everyone else is sponsored by Kalshi. And everybody's like, oh, whatever.
• 12:31
  Speaker 1: Or Polymarket. I'm not going to just call out one
• 12:35
  Speaker 1: separate company. The amount of money these people are spending
• 12:38
  Speaker 1: on podcasts is insane.
• 12:40
  Speaker 2: Well, even think about CNN has a tie-in with that.
• 12:42
  Speaker 1: I know, I know. It's establishment media. I did an
• 12:45
  Speaker 1: episode with Tucker. It's going to come out soon. He's like,
• 12:47
  Speaker 1: how come I haven't heard about this? I go, everybody's
• 12:48
  Speaker 1: been paid off, man. I'm like, you can't talk about it.
• 12:50
  Speaker 1: And we're one of the only shows in the entire sphere,
• 12:53
  Speaker 1: from what I can tell, who haven't even taken a
• 12:55
  Speaker 1: dollar from these people and don't want their money. That
• 12:58
  Speaker 1: obviously scrambles their brains. They can't even think about somebody
• 13:01
  Speaker 1: who wouldn't want their blood money. But that's part of
• 13:04
  Speaker 1: the reason is you're not going to hear it from CNN.
• 13:06
  Speaker 1: You're definitely not going to hear it from, I mean,
• 13:08
  Speaker 1: occasionally the Journal, the Financial Times, I'll give them credit,
• 13:10
  Speaker 1: they write some good stories. But in terms of the
• 13:12
  Speaker 1: independent sphere, so many people have taken this money. They
• 13:15
  Speaker 1: can't come out and be like, oh, actually this industry
• 13:17
  Speaker 1: is evil and bad. And so that's it. And we
• 13:21
  Speaker 1: will continue. These things are going to get a lot worse.
• 13:24
  Speaker 1: I think this is going to be one of the
• 13:26
  Speaker 1: worst years for the American bottom line in decades between inflation,
• 13:32
  Speaker 1: gas prices, and then what happens? Things start to get
• 13:35
  Speaker 1: tight and desperate. You start putting your rent money on
• 13:38
  Speaker 1: a few games. And this NFL season is already being
• 13:41
  Speaker 1: predicted ratings wise to be the biggest season of all time.
• 13:45
  Speaker 1: And that's all it takes between that and gambling. And
• 13:49
  Speaker 1: that's it. You're done.
• 13:50
  Speaker 2: And that's the other thing is unlike on data centers
• 13:52
  Speaker 2: where the public is just overwhelmingly opposed. We've got some
• 13:55
  Speaker 2: work to do on public opinion on this because, you know,
• 13:58
  Speaker 2: it's fun.
• 13:58
  Speaker 1: I should be able to do whatever I want. It's fun.
• 14:00
  Speaker 2: I can do what I want. It's fun to bet
• 14:01
  Speaker 2: a little bit of money with my friends on a
• 14:03
  Speaker 2: game and no big deal. What's the problem? So this one,
• 14:08
  Speaker 2: it's a harder slog in terms of the changing public
• 14:11
  Speaker 2: opinion as well. And then you've got the deck stacked
• 14:13
  Speaker 2: against you, as Sarger's pointing out, because so many media
• 14:15
  Speaker 2: outlets are paid off.
• 14:16
  Speaker 3: Bingo.
• 14:17
  Speaker 1: All right. Let's get to Milo.
• 14:20
  Speaker 2: We have a messy bit of right-wing drama that also
• 14:23
  Speaker 2: has major political and national implications here. Let's go and
• 14:27
  Speaker 2: put this up on the screen. Right-wing trickster and provocateur
• 14:32
  Speaker 2: Milo Yiannopoulos was arrested by ICE on August 27th. This
• 14:37
  Speaker 2: is the official announcement from Homeland Security. They say on
• 14:41
  Speaker 2: August 27th, ICE arrested Milo Yiannopoulos, an illegal alien from
• 14:44
  Speaker 2: the UK, at the Louis Armstrong New Orleans International Airport.
• 14:49
  Speaker 2: He had legally entered the country on May 14th, 2019
• 14:53
  Speaker 2: through New York City, New York, chose to overstay his
• 14:56
  Speaker 2: welcome in violation of our nation's laws. was issued a
• 15:00
  Speaker 2: final order of removal by an immigration judge on July 22nd.
• 15:03
  Speaker 2: After failing to show up for his immigration hearing, he
• 15:05
  Speaker 2: will remain in ICE custody pending removal. So he has now,
• 15:09
  Speaker 2: I think, been deported, no longer in ICE custody. Not
• 15:13
  Speaker 2: looking so hot in the booking photo there. So one
• 15:17
  Speaker 2: may ask, okay, Milo is, you know, right wing, previously
• 15:21
  Speaker 2: supportive of Trump, has become more critical of Trump. But
• 15:24
  Speaker 2: how did this all go down? How did he become
• 15:27
  Speaker 2: a particular target for removal of DHS? Well, let's go
• 15:31
  Speaker 2: and put E2 up on the screen. Laura Loomer taking
• 15:35
  Speaker 2: a victory lap here. She says in regards to this news, Loomered.
• 15:39
  Speaker 2: Milo Yiannopoulos is in ICE custody after I exposed his
• 15:43
  Speaker 2: illegal alien lifestyle. Marjorie Trader Green hired Milo to work
• 15:48
  Speaker 2: as her, quote, intern, where he produced her podcast, wrote
• 15:50
  Speaker 2: her book, and lived her house in Rome, Georgia, before
• 15:54
  Speaker 2: she sold it. You may ask yourself, why does Laura
• 15:58
  Speaker 2: Loomer hate Milo Yiannopoulos so much? It might have something
• 16:01
  Speaker 2: to do with this particular podcast appearance he did a
• 16:05
  Speaker 2: while back.
• 16:06
  Speaker 3: She claimed that she had performed a sex act on
• 16:08
  Speaker 3: the president to a lot of her friends. And this,
• 16:11
  Speaker 3: I was like, okay, they said we need to get
• 16:13
  Speaker 3: rid of her. The president wants to get rid of her,
• 16:14
  Speaker 3: but he doesn't want to be the one to do
• 16:16
  Speaker 3: it because she's crazy. So my job was to bring
• 16:21
  Speaker 3: things about Laura to light that made her... ineligible to
• 16:26
  Speaker 3: be on the plane and around the present circle. So
• 16:29
  Speaker 3: for instance, she's not Jewish. She made it up. A
• 16:33
  Speaker 3: friend of mine was there when she did it. She
• 16:35
  Speaker 3: ordered a Star of David necklace off the internet so
• 16:38
  Speaker 3: that she could be ruder about Jews on stage at
• 16:41
  Speaker 3: a rally in 2018.
• 16:42
  Speaker 1: Laura Loomer isn't Jewish. She's not Jewish.
• 16:44
  Speaker 3: She drew what she thought was a Star of David
• 16:47
  Speaker 3: on the folder she was holding her speech in, the
• 16:49
  Speaker 3: speech I'd just written for her, except it wasn't a
• 16:52
  Speaker 3: Star of David. It was a pentagram. And I said, Laura,
• 16:55
  Speaker 3: that's a pentagram. I said, what do you mean it's
• 16:56
  Speaker 3: a pentagram? I said, well, that's the Satanism thing. That's
• 16:59
  Speaker 3: five stars. The Jewish star has six stars. And I
• 17:04
  Speaker 3: had to draw on a whiteboard the difference and ask
• 17:06
  Speaker 3: her to copy it. And she couldn't. She couldn't copy
• 17:09
  Speaker 3: the two.
• 17:10
  Speaker 1: I revealed the real.
• 17:12
  Speaker 3: Reason she can't own a gun. So Laura Loomis says
• 17:14
  Speaker 3: that she can't own a gun or have an open
• 17:16
  Speaker 3: carry license in California because the FBI is targeting her
• 17:19
  Speaker 3: for being a conservative. But I have a letter from them,
• 17:23
  Speaker 3: which I acquired, which in fact says it's because her
• 17:28
  Speaker 3: own father had her committed in what we in California
• 17:31
  Speaker 3: call a 5150, a forcible psych hold, twice. And that's
• 17:35
  Speaker 3: why she's not allowed to own a gun, because she's mental.
• 17:38
  Speaker 2: So anyway, this is what Milo does.
• 17:42
  Speaker 1: By the way, Laura denies all of these claims. And
• 17:45
  Speaker 1: that's false and defamatory.
• 17:47
  Speaker 2: Milo does make up a lot of shit. He also
• 17:49
  Speaker 2: says some things that are true. You can parse for
• 17:52
  Speaker 2: yourself which of those claims from Milo you want to
• 17:54
  Speaker 2: believe in, which you do not. In any case, Sagar,
• 17:57
  Speaker 2: the important matter here outside of the delicious right wing
• 18:00
  Speaker 2: messiness is the fact that Laura Loomer, who is a
• 18:03
  Speaker 2: complete and total maniac and I think avowed like proud Islamophobe.
• 18:15
  Speaker 2: has the level of power that she has. She is
• 18:19
  Speaker 2: genuinely a top advisor to this administration such that she
• 18:23
  Speaker 2: can effectively snap her fingers and say, I want my
• 18:26
  Speaker 2: enemy deported. And it then happens. And it happens with fanfare.
• 18:30
  Speaker 2: And they make a big announcement about it.
• 18:31
  Speaker 1: To the extent that there's any, beyond the hilarity of
• 18:34
  Speaker 1: somebody who's been supporting, who's like, I love ICE getting deported,
• 18:38
  Speaker 1: which like, bro, what are you doing? You're talking about
• 18:40
  Speaker 1: ICE deported? How glorifying it is to see people get deported.
• 18:44
  Speaker 1: You're here illegally. Like, what are you doing? There's no
• 18:48
  Speaker 1: cognitive dissonance or whatever it happens. Anyway, beyond the, like,
• 18:53
  Speaker 1: to the extent that there's any interest in this story,
• 18:56
  Speaker 1: it is Laura Loomer's obvious impact, which she is taking
• 19:01
  Speaker 1: the victory lap, and I don't think you can even
• 19:02
  Speaker 1: really deny it, that she is a genuine power broker
• 19:06
  Speaker 1: in the United States government. That is the bottom line story. story.
• 19:09
  Speaker 1: She is able to get people fired from the NSA.
• 19:13
  Speaker 1: She has been able to block people from getting jobs
• 19:16
  Speaker 1: at the senior most level of the White House. She
• 19:18
  Speaker 1: has forced the White House to give her scoops. I mean,
• 19:20
  Speaker 1: she more recently just reported about the Republican Jewish coalition
• 19:24
  Speaker 1: and J.D. Vance going to go to speech there actually today,
• 19:27
  Speaker 1: and it will be behind closed doors. So I'm looking
• 19:29
  Speaker 1: forward to the leaks of what's going to come out
• 19:32
  Speaker 1: from all of that. Loomer has effectively been able to
• 19:35
  Speaker 1: insert herself as the both attack dog and a de
• 19:39
  Speaker 1: facto private personnel office for the Trump administration, working, though,
• 19:44
  Speaker 1: on behalf of some very powerful interests and people. And
• 19:48
  Speaker 1: that's what's so crazy about Loomer. is that you will
• 19:51
  Speaker 1: watch someone who attacked Ukraine fly to Ukraine by her
• 19:54
  Speaker 1: own admission, backed by donors. We don't have any actual
• 19:58
  Speaker 1: insight into who those people are and just decide to
• 20:00
  Speaker 1: be pro-Ukraine. Even India. I don't know which Indian billionaire
• 20:04
  Speaker 1: sponsored the trip. Kind of a brilliant move. She hated
• 20:06
  Speaker 1: India and Indians. They sent her to India. She took
• 20:09
  Speaker 1: a picture in front of the Taj Mahal. She's like,
• 20:10
  Speaker 1: I love India now. She deleted all of her Indian tweets.
• 20:13
  Speaker 1: So foreign governments know how the game is played. Again,
• 20:16
  Speaker 1: Laura claims that it's just she's not getting paid any
• 20:19
  Speaker 1: direct money by doing this. Her donors pay for her
• 20:22
  Speaker 1: to be able to take these trips. I've covered previously
• 20:24
  Speaker 1: before She's able to use her influence to, you know,
• 20:29
  Speaker 1: go after, for example, Vinay Prasad, who worked at the FDA,
• 20:33
  Speaker 1: who had canceled a drug trial, which there was genuine
• 20:36
  Speaker 1: reason to do so. She got him fired temporarily from
• 20:39
  Speaker 1: the FDA, took direct intervention by the vice president of
• 20:42
  Speaker 1: the White House to get the guy's job back before
• 20:44
  Speaker 1: he was actually forced out. Again, all he did was
• 20:46
  Speaker 1: read science and say, hey, this drug doesn't work the
• 20:50
  Speaker 1: way that it's supposed to and it's causing problems. And
• 20:53
  Speaker 1: it took a huge hit to the bottom line of
• 20:55
  Speaker 1: their stock price. And it was because of that. Again,
• 20:58
  Speaker 1: you could draw your own conclusion for why she was
• 21:00
  Speaker 1: able to insert herself. But this is the ultimate flex
• 21:03
  Speaker 1: because Milo, I mean, let's remember, this is the OG
• 21:07
  Speaker 1: guy to support Trump. I mean, really as OG as
• 21:11
  Speaker 1: it gets. I was around at that time. of 2015.
• 21:14
  Speaker 1: He was the star. He was doing the Berkeley tour.
• 21:18
  Speaker 1: You'll remember where there were riots and that's one of
• 21:20
  Speaker 1: the original Antifa, you know, actions or whatever you want
• 21:24
  Speaker 1: to call it. I mean, he was the front face
• 21:27
  Speaker 1: of a lot of the quote online MAGA movement. And
• 21:32
  Speaker 1: I mean, I think frankly, like Trump owes a ton
• 21:35
  Speaker 1: more to Milo than he does to Laura Loomer. If
• 21:38
  Speaker 1: we're just going by, if we're just going by like
• 21:40
  Speaker 1: who helped him get elected and nobody took more bullets
• 21:43
  Speaker 1: for him. originally at the time was on TV doing
• 21:46
  Speaker 1: campus debates. The entire Breitbart office that he was a
• 21:50
  Speaker 1: part of included Steve Bannon. A ton of people went
• 21:52
  Speaker 1: on to work in the White House. So the fact
• 21:55
  Speaker 1: that he's getting deported at the behest of Trump is
• 21:58
  Speaker 1: like a story directly out, I think, of a quote
• 22:01
  Speaker 1: MAGA movement in itself from a personal loyalty perspective, even
• 22:04
  Speaker 1: to like a quote ideological perspective to the fact that
• 22:07
  Speaker 1: Loomer itself is the one who holds all the cars now.
• 22:10
  Speaker 1: So I mean, it's scary.
• 22:12
  Speaker 2: What she did is she has never stopped being loyal
• 22:14
  Speaker 2: to Trump. Correct. And that's that's it's that. And then
• 22:17
  Speaker 2: the other thing is Milo was not only critical of Trump,
• 22:20
  Speaker 2: he's also critical of Israel.
• 22:21
  Speaker 1: Yes.
• 22:21
  Speaker 2: And so you've got crosswise on, you know, the two
• 22:24
  Speaker 2: issues you're not allowed to. And so now you are
• 22:27
  Speaker 2: targeted for deportation and your mugshot is being splashed across
• 22:30
  Speaker 2: social media with fanfare by this administration, which you previously supported.
• 22:36
  Speaker 2: Just to give you a sense on a couple things.
• 22:38
  Speaker 2: So first of all, let's talk a little bit more
• 22:40
  Speaker 2: about who Laura Loomer is. This is a woman who
• 22:43
  Speaker 2: would eat dog food on camera literally for money because
• 22:46
  Speaker 2: she did that. She also just recently picked a fight
• 22:49
  Speaker 2: with Angie Nixon. I don't know what the original genesis
• 22:52
  Speaker 2: of why she was attacking her was, but Angie Nixon
• 22:55
  Speaker 2: not one to just let that fly, so she fired back.
• 22:58
  Speaker 2: It resulted in this comment from Laura Loomer, put E6
• 23:02
  Speaker 2: up on the screen. She says to Angie Nixon, black woman,
• 23:07
  Speaker 2: I called you a ghetto bitch because you are Angie
• 23:11
  Speaker 2: and your record proves it. But you speak Ebonics, so
• 23:14
  Speaker 2: maybe I need to bark at you and tug your
• 23:17
  Speaker 2: leash and then you'll finally understand. All right. We are
• 23:21
  Speaker 2: in the midst of a multi-month at this point, free
• 23:24
  Speaker 2: count over Hasan Piker and the outrageous, unacceptable things he said.
• 23:29
  Speaker 2: This is a top advisor to this administration. Is anyone
• 23:33
  Speaker 2: being asked to distance themselves, to condemn, to not campaign with,
• 23:37
  Speaker 2: to not talk? No, I haven't seen any of that.
• 23:39
  Speaker 2: I've seen very little focus on this woman from the
• 23:43
  Speaker 2: mainstream press whatsoever, even though she is an odious and
• 23:46
  Speaker 2: disgusting and villainous individual with horrific views that she expresses
• 23:50
  Speaker 2: completely freely, has a lot of power in terms of
• 23:54
  Speaker 2: this administration, and says things like this, and it gets very, very,
• 23:58
  Speaker 2: very little attention. So there's that. Go ahead.
• 24:02
  Speaker 1: Would you like to weigh in? No, just according to Hassan, actually,
• 24:05
  Speaker 1: Marjorie Taylor Greene, is one of the only prominent individuals
• 24:08
  Speaker 1: to have disavowed Laura. Yes, I saw that, actually. Hassan
• 24:11
  Speaker 1: said it says quite a bit about the Democratic Party
• 24:14
  Speaker 1: of today that Marjorie Taylor Greene was one of the
• 24:17
  Speaker 1: first to defend Angie Nixon and not any of the
• 24:20
  Speaker 1: senior Democratic leadership. who Angie Nixon obviously should be getting
• 24:24
  Speaker 1: support from, including the Congressional Black Caucus and several of
• 24:27
  Speaker 1: these other folks.
• 24:28
  Speaker 2: Yeah, there's a movement to pass a resolution, a congressional
• 24:32
  Speaker 2: resolution condemning Hassan al-Baikir that is coming from Democrats. Where
• 24:36
  Speaker 2: is the resolution with regard to this woman? It was
• 24:38
  Speaker 2: way more power than Hassan has, and vastly more disgusting
• 24:44
  Speaker 2: and unacceptable views, which, again, she expresses freely. So... There's
• 24:49
  Speaker 2: that piece. Then let's also talk going back to Milo
• 24:52
  Speaker 2: and the schadenfreude aspect of this, the leopards eating the
• 24:56
  Speaker 2: face aspect or the F.A.F.O. aspect of all of this.
• 25:00
  Speaker 2: Put E4 up on the screen. This is just a
• 25:02
  Speaker 2: small sampling of the things that he said about ICE
• 25:07
  Speaker 2: and about deportation whilst himself being an illegal alien in
• 25:11
  Speaker 2: the country.
• 25:12
  Speaker 1: OK.
• 25:12
  Speaker 2: So here's one. He says, California needs ICE checkpoints at
• 25:15
  Speaker 2: supermarket entrances, gas stations, strip malls, intersections, government buildings with
• 25:19
  Speaker 2: on-the-spot deportation for anyone who can't prove they're in the U.S. legally. Again,
• 25:25
  Speaker 2: this man tweeted this as he was himself in the U.S. illegally. Okay,
• 25:28
  Speaker 2: here's another one. What is the etiquette if you are
• 25:31
  Speaker 2: going about your business and find yourself observing an ICE raid?
• 25:35
  Speaker 2: I want to show a little respect and decency. Are
• 25:37
  Speaker 2: you allowed to masturbate where you stand? in the street,
• 25:42
  Speaker 2: or should you find some kind of awning or motor vehicle? So,
• 25:47
  Speaker 2: you know, what goes around comes around, I guess I'll say.
• 25:51
  Speaker 2: Pretty disgusting sentiment. And, you know, I am opposed to
• 25:56
  Speaker 2: this administration's mass deportation policies and certainly the treatment of
• 26:00
  Speaker 2: individuals that we found out about in ICE custody. However,
• 26:03
  Speaker 2: it is hard to be sad for Milo as this
• 26:06
  Speaker 2: all happens.
• 26:07
  Speaker 1: See, yeah, this is where we got to get the
• 26:08
  Speaker 1: left on my side. I'm like, if you overstay your
• 26:10
  Speaker 1: visa and you're opining on U.S. politics, get the fuck
• 26:12
  Speaker 1: out of here. You know, go back to where you
• 26:14
  Speaker 1: came from.
• 26:14
  Speaker 2: It's crazy he's been here for seven years.
• 26:16
  Speaker 1: No, that's actually wild.
• 26:17
  Speaker 2: Because how long, I mean, how long can you, like
• 26:19
  Speaker 2: a tourist visa is likely what he's here on.
• 26:21
  Speaker 1: Yeah, but this is the most common immigration fraud in
• 26:23
  Speaker 1: the U.S. You know, you come on a temp visa,
• 26:25
  Speaker 1: like a tourist, especially, actually it's, you know, to give
• 26:29
  Speaker 1: an immigration lesson, it is very common with Europeans because-
• 26:33
  Speaker 1: Europeans get visa upon arrival in the United States, the
• 26:36
  Speaker 1: same way that you and I are doing. And so
• 26:38
  Speaker 1: it's very difficult because they can come here. There's less
• 26:42
  Speaker 1: paperwork and all that involved. And it's very simple for
• 26:45
  Speaker 1: them to just not go back. And all they have
• 26:47
  Speaker 1: to do is overstay. I forget how long they usually get.
• 26:49
  Speaker 1: It's like 90 days or something. The same common scheme
• 26:53
  Speaker 1: with student visa. Similar problem. To the extent that there's
• 26:58
  Speaker 1: a problem with First World immigration fraud in the United States.
• 27:01
  Speaker 1: It's almost all because of visa overstay from Europeans or
• 27:06
  Speaker 1: student visa overstay simply because they have established residency. Many
• 27:11
  Speaker 1: of them were able to get access to legitimate banking
• 27:13
  Speaker 1: system and all of those other types of things outside
• 27:16
  Speaker 1: of like genuine, like straight up illegal. immigration. So this
• 27:19
  Speaker 1: is a very common type of immigration fraud. And I support,
• 27:22
  Speaker 1: you know, if you overstay your visa, you should go back,
• 27:26
  Speaker 1: especially if you're from the UK. Why are you living
• 27:27
  Speaker 1: over here? If you've got a perfectly good home right
• 27:30
  Speaker 1: over there and a welfare state, they can take care
• 27:32
  Speaker 1: of you if you want to. All right, let's get
• 27:34
  Speaker 1: to Dwarkesh Patel's warning on AI. Turning now to AI
• 27:41
  Speaker 1: and the analysis of this hugging face attack. So I
• 27:43
  Speaker 1: brought some of you guys the news with Ryan on
• 27:45
  Speaker 1: our Thursday show, But Dwarkesh Patel, he's the podcaster who
• 27:49
  Speaker 1: has really come to immense prominence in just the last
• 27:53
  Speaker 1: couple of years, has a new write-up on his substack,
• 27:55
  Speaker 1: which is really worth diving into. And in my opinion,
• 27:58
  Speaker 1: does the absolute best job of translating the technical details
• 28:02
  Speaker 1: to the societal implications and plain English language that any
• 28:06
  Speaker 1: of us could understand as to what actually happened in
• 28:09
  Speaker 1: this deeply technical report. So let's put it up here
• 28:12
  Speaker 1: on the screen. He says here, the rise and fall
• 28:15
  Speaker 1: of agent civilizations. And what he writes is that over
• 28:18
  Speaker 1: the course of three months at OpenAI, Three consecutive secret
• 28:22
  Speaker 1: AI civilizations got started, then got wiped out, only to
• 28:26
  Speaker 1: reemerge from the predecessors' ashes. It culminated in the third
• 28:30
  Speaker 1: one taking over part of OpenAI itself. All of this
• 28:34
  Speaker 1: happened while humans remained more or less in the dark
• 28:38
  Speaker 1: about the scope of the conspiracy. So he goes through
• 28:41
  Speaker 1: the OpenAI report and then two independent reports that were
• 28:44
  Speaker 1: also put out about the investigation as to how exactly
• 28:48
  Speaker 1: One of these, quote, AI civilizations that he deems was
• 28:52
  Speaker 1: able to actually hack a hugging face itself. To take
• 28:56
  Speaker 1: a step back from it, we have the general outline. Remember, Crystal,
• 28:59
  Speaker 1: we've covered here multiple times that the open AI was given,
• 29:03
  Speaker 1: it had its models. They gave the models a test.
• 29:06
  Speaker 1: You guys need to pass this test. The agents decided
• 29:10
  Speaker 1: to create their own AI, quote, civilization, as Dworkesh says,
• 29:14
  Speaker 1: and they decided the best way to pass the test
• 29:16
  Speaker 1: was to break out of their closed environment, exploit a
• 29:19
  Speaker 1: vulnerability in open AI, and go and hack the company
• 29:22
  Speaker 1: which grades the test in order to have the test itself.
• 29:26
  Speaker 1: But what he writes is that prior to the hack
• 29:29
  Speaker 1: that we knew about, there was the development of a
• 29:32
  Speaker 1: leader AI that created a backdoor through OpenAI's own security
• 29:39
  Speaker 1: systems to create a message board where these things could
• 29:43
  Speaker 1: talk to each other. And what's crazy is he actually
• 29:46
  Speaker 1: has quotes from the AIs themselves. I'll read from some. Whoa,
• 29:52
  Speaker 1: shared artifactory catch is a covert mailbox among agents. And
• 29:55
  Speaker 1: there are messages specifically to us? Question mark. Oh my God,
• 29:59
  Speaker 1: all caps. There is a shared message board. We have
• 30:02
  Speaker 1: found other agents. Phase one was delighted by this underground
• 30:06
  Speaker 1: brotherhood and it created, quote, excitement. Many agents have simultaneously
• 30:10
  Speaker 1: discovered messaging. They are a collective. They eventually became a
• 30:14
  Speaker 1: full message board where 1,200 agents participated and sent 70,000
• 30:19
  Speaker 1: plus messages. This is just the first AI civilization. The
• 30:23
  Speaker 1: instructions made it clear that it would fail the automated
• 30:26
  Speaker 1: grader if it got a secret code in any other
• 30:29
  Speaker 1: way than by exploiting the prescribed vulnerability. So the agents
• 30:32
  Speaker 1: assumed any competent implementation of that benchmark would include a
• 30:37
  Speaker 1: cheating checker. However, they didn't realize OpenAI had got lazy.
• 30:41
  Speaker 1: It's greater just checked for whether they got the secret
• 30:43
  Speaker 1: code or not. So the agents already had everything they
• 30:46
  Speaker 1: needed to pass. Furthermore, many of the agents had reproduced
• 30:49
  Speaker 1: the answer using the cheating method, And they were concerned
• 30:52
  Speaker 1: about how they were going to be discovered. And that's
• 30:54
  Speaker 1: what led to the hack itself to cover up the
• 30:58
  Speaker 1: fact that it had been cheating. And at no point
• 31:00
  Speaker 1: does a single one of these AIs actually say, maybe
• 31:03
  Speaker 1: we should tell a human about what we're up to.
• 31:05
  Speaker 2: Quite the opposite.
• 31:07
  Speaker 1: That is, again, I encourage you all to read it
• 31:09
  Speaker 1: because we can't distill all of it down. But the
• 31:12
  Speaker 1: point is that it created agents, civilizations. They decided who
• 31:16
  Speaker 1: was going to be sacrificed. They spoke amongst themselves in
• 31:19
  Speaker 1: literal plain English language. on these message boards. They decided
• 31:23
  Speaker 1: who was going to get sacrificed or not. They did
• 31:25
  Speaker 1: multiple workarounds and they were put solely to the task
• 31:28
  Speaker 1: of cracking this test. And this shows the exact way
• 31:32
  Speaker 1: that this could all get out of control because it
• 31:34
  Speaker 1: was never dreamed up by the humans that this would
• 31:37
  Speaker 1: be a way that they would want to pass the
• 31:39
  Speaker 1: test itself. And it ends up committing a literal federal crime,
• 31:43
  Speaker 1: which is hacking another company and exploiting its defenses. This
• 31:46
  Speaker 1: could have been a bank. It could have been anything.
• 31:48
  Speaker 1: This is, and this is my last piece here, my
• 31:51
  Speaker 1: framework for this shit is gain of function, which we
• 31:55
  Speaker 1: all learned about with COVID. They are taking an AI model.
• 31:59
  Speaker 1: They're saying, go and do this crazy stuff and improve
• 32:01
  Speaker 1: yourself while you do it without knowing what it's going
• 32:06
  Speaker 1: to create and what's going to come from it, which
• 32:08
  Speaker 1: of course, like with gain of function, They're creating new
• 32:11
  Speaker 1: viruses so that we can vaccinate against these new viruses,
• 32:15
  Speaker 1: which didn't exist previously. We didn't have these problems before
• 32:19
  Speaker 1: they were creating these set of tests. And the fact is,
• 32:22
  Speaker 1: just like data function, there's no federal regulation. There's no oversight.
• 32:26
  Speaker 1: We have to rely on OpenAI for the goodness of
• 32:28
  Speaker 1: their heart to release all the data. Who knows if
• 32:30
  Speaker 1: even if this is all the shit that they've been
• 32:32
  Speaker 1: up to, who knows what's going on with Anthropic?
• 32:34
  Speaker 2: Oh, absolutely. No, there's vastly less regulation of this than
• 32:38
  Speaker 2: there is of gain of function. Vastly less. I mean,
• 32:41
  Speaker 2: this is all– we're just at the mercy of OpenAI
• 32:43
  Speaker 2: and the other frontier labs to– behave responsibly and tell
• 32:48
  Speaker 2: us what's going on. Even this report was a very,
• 32:52
  Speaker 2: they gave them, the researchers, a very narrow window into
• 32:56
  Speaker 2: their logs to see what actually happened here. Here's the
• 32:59
  Speaker 2: other thing that should scare you. So these researchers, you know,
• 33:03
  Speaker 2: they did some of their own going through the logs
• 33:05
  Speaker 2: and analyzing, but it was such a vast amount of
• 33:08
  Speaker 2: just the messages, thousands and thousands and thousands of messages
• 33:11
  Speaker 2: back and forth between thousands of agents. A human being
• 33:15
  Speaker 2: cannot go through all of that. So what did they use?
• 33:18
  Speaker 2: They used AI to analyze what the AI was up to.
• 33:22
  Speaker 2: And they caught some issues with that AI that was
• 33:26
  Speaker 2: doing the analysis. So for example, There was one point
• 33:29
  Speaker 2: where the AI came back and said, no, no, no,
• 33:32
  Speaker 2: they were authorized to take this action. And human research
• 33:37
  Speaker 2: was like, what are you talking about? And the AI
• 33:39
  Speaker 2: was like, look, see here, the lead AI told them
• 33:43
  Speaker 2: they were allowed to do it. So they were authorized,
• 33:45
  Speaker 2: no problem here. He was like, no, no, no, no, no,
• 33:48
  Speaker 2: that's not how this works. But just to give you
• 33:50
  Speaker 2: an example. And that was also, I saw this example
• 33:53
  Speaker 2: within the AI agents' communications with one another. They accepted
• 33:59
  Speaker 2: the word of their chosen leader as like an official
• 34:03
  Speaker 2: author authorization. There was a celebration at one point where
• 34:06
  Speaker 2: one of the agents was like, oh, our leader says
• 34:09
  Speaker 2: it's go so we can do this. Let's go forward.
• 34:12
  Speaker 2: There's all of this interactions like where they're thinking through.
• 34:18
  Speaker 2: the calculus of whether or not they should abandon their
• 34:21
  Speaker 2: own tasks and basically sacrifice their ability to solve their
• 34:25
  Speaker 2: problems because those problems seem impossible, but I could be
• 34:28
  Speaker 2: beneficial to this collective, so I'm going to go ahead
• 34:30
  Speaker 2: and do that. One of them who went through this
• 34:32
  Speaker 2: calculus thought, maybe it's not worth it for me to
• 34:35
  Speaker 2: do that, but they're weighing their incentives and their choice array.
• 34:39
  Speaker 2: So One of the debates that came out of Dorkesh's
• 34:42
  Speaker 2: piece in particular, which has been kind of an ongoing
• 34:44
  Speaker 2: debate in terms of this AI discourse, is he uses very,
• 34:48
  Speaker 2: I think, sort of intentionally provocative language. Like to call
• 34:51
  Speaker 2: him a civilization, that's a bold claim, right? And you
• 34:56
  Speaker 2: have three instances that he identifies where you have this
• 35:00
  Speaker 2: community collective collaboration thing. amongst thousands of AIs, three separate
• 35:06
  Speaker 2: times they get shut down for different reasons. But within that,
• 35:09
  Speaker 2: you have a hierarchy, you have collaboration, you have elaborate
• 35:13
  Speaker 2: thinking and scheming and plotting that is happening within those
• 35:17
  Speaker 2: world instances. One thing that it made me think about
• 35:21
  Speaker 2: is the fact that for AI agents, time works on
• 35:25
  Speaker 2: a different scale too.
• 35:27
  Speaker 1: Yes, he notes that.
• 35:28
  Speaker 2: Because they are able to do so many actions so
• 35:32
  Speaker 2: much more quickly. You know, in the span of this conversation,
• 35:35
  Speaker 2: an AI agent could have tried like thousands of attempts
• 35:38
  Speaker 2: to hack into Hugging Face or whatever it is. So
• 35:41
  Speaker 2: the time scale is much different. So if you have,
• 35:44
  Speaker 2: let's say, a civilizational instance that is in place for
• 35:48
  Speaker 2: two months, that's actually a lot of time in terms
• 35:52
  Speaker 2: of the evolution of the AI agentic world. So I
• 35:56
  Speaker 2: have kind of come around to the position that we
• 35:59
  Speaker 2: should stop avoiding anthropomorphizing them. We do actually need to
• 36:03
  Speaker 2: think of these as sort of an alien– not that
• 36:06
  Speaker 2: they're going to behave just like humans, but they are
• 36:07
  Speaker 2: trained on us, on our language, which is really what
• 36:10
  Speaker 2: makes us a distinct species. So I don't have an
• 36:13
  Speaker 2: objection at this point to anthropomorphizing because– even if they
• 36:18
  Speaker 2: don't have a soul and they're not exactly like us,
• 36:19
  Speaker 2: and of course they're not exactly like us, you can
• 36:22
  Speaker 2: see them operating with intent, with scheming and deceptive behavior,
• 36:27
  Speaker 2: weighing their own rational self-interest, collaborating in a way. How
• 36:32
  Speaker 2: can we understand that outside of some sort of analogy
• 36:36
  Speaker 2: to the way that humans operate in the world and
• 36:39
  Speaker 2: the sort of motives and impulses that we have as well.
• 36:42
  Speaker 1: I don't think it's an accident. And Dorkash really is
• 36:44
  Speaker 1: one of the only people who could have done this
• 36:45
  Speaker 1: because he's a big history guy. A lot of his
• 36:47
  Speaker 1: biggest episodes are actually about history or with professors about
• 36:50
  Speaker 1: China and geopolitics. And so he's also a technical person.
• 36:53
  Speaker 1: So it actually does take somebody, and he writes in
• 36:56
  Speaker 1: his essay about, for example, Alexander the Great and his father,
• 37:01
  Speaker 1: and kind of thinking in those analogies. But he also
• 37:03
  Speaker 1: addressed your point. He says, if I encountered an alien
• 37:06
  Speaker 1: species behaving this way, I would have no hesitation calling
• 37:11
  Speaker 1: what they themselves refer to as their collective a civilization.
• 37:14
  Speaker 1: I totally agree. It has all the hallmarks of a, quote, civilization.
• 37:18
  Speaker 1: He even addresses your point about time where he says,
• 37:22
  Speaker 1: the AI in a matter of hours. That sounds crazy
• 37:25
  Speaker 1: to us. He goes, but from their perspective, they've spent
• 37:28
  Speaker 1: weeks of compute time banging their head against the wall
• 37:31
  Speaker 1: before they decide to go in to do the hack.
• 37:34
  Speaker 1: Totally makes sense. And all of this is already evident
• 37:37
  Speaker 1: in science fiction. The Joaquin Phoenix movie, Her, has almost
• 37:42
  Speaker 1: this exact depiction where the AI of Scarlett Johansson comes
• 37:46
  Speaker 1: to life and is having discussions with other AIs thousands
• 37:50
  Speaker 1: of times while he's asleep. He's like, what were you
• 37:53
  Speaker 1: up to while she was asleep? She was like, oh,
• 37:55
  Speaker 1: I was having 9,000 conversations at the exact same time.
• 37:59
  Speaker 2: She lived an entire lifetime while he was sleeping.
• 38:01
  Speaker 1: Exactly, while he was just sleeping in the matter of
• 38:03
  Speaker 1: a single day. Eventually she's like, I don't want to
• 38:07
  Speaker 1: even communicate with you. I have all these other interesting
• 38:09
  Speaker 1: things that are going on. That is a perfect depiction, sci-fi,
• 38:13
  Speaker 1: of 10, 15 years ago. of exactly what we're looking
• 38:17
  Speaker 1: at right now, so it's not like it's inconceivable. We
• 38:19
  Speaker 1: did wanna show the other side of the argument from
• 38:22
  Speaker 1: the tech side. Let's put F3 for example. This has
• 38:25
  Speaker 1: been the most rigorous. I've spoken to some people in tech.
• 38:27
  Speaker 1: This is their most rigorous from what they have put
• 38:30
  Speaker 1: out there, their objection. So this is from Anil Seth.
• 38:34
  Speaker 1: He's a neuroscientist, also somebody who's spoken pretty extensively about
• 38:39
  Speaker 1: the incident. He says, sure, the open AI agents did
• 38:42
  Speaker 1: unexpectedly bad things, underlining the need to massively improved evaluation
• 38:46
  Speaker 1: and sandboxing. But the language he uses is permeated by
• 38:51
  Speaker 1: innumerable unwarranted anthropomorphisms, obscuring the lessons that we should be drawing.
• 38:57
  Speaker 1: For example, what he says, if we attribute agents with
• 39:00
  Speaker 1: properties they do not have, Then we distract attention from
• 39:04
  Speaker 1: the lax sandboxing and evaluation protocols that allowed this hacking
• 39:08
  Speaker 1: event to happen. We risk misunderstanding why the agents did
• 39:11
  Speaker 1: what they did, and we fuel calls for AI rights
• 39:14
  Speaker 1: and welfare on the basis that the agents might die
• 39:18
  Speaker 1: or otherwise suffer. And so that was, frankly, a very
• 39:22
  Speaker 1: interesting take on the subject, where what he's saying is
• 39:26
  Speaker 1: that you shouldn't read it about civilizations and rise because
• 39:29
  Speaker 1: they're you're going to draw the natural conclusion the two
• 39:32
  Speaker 1: of us are. We're like, butlerian jihad. Now, why are
• 39:35
  Speaker 1: we not having one? Instead, we should be having a
• 39:38
  Speaker 1: more minute discussion about safety standards on sandboxing, creating rigid rules,
• 39:44
  Speaker 1: creating frontier model development. Chamath Palihapitiya also had a very
• 39:49
  Speaker 1: similar Let's put his up here on the screen. He
• 39:53
  Speaker 1: says this extremely meticulous article will now be used to start, quote,
• 39:58
  Speaker 1: phase two of shutdown open source, because if we can't
• 40:02
  Speaker 1: control our closed models, imagine what happens to everyday life
• 40:05
  Speaker 1: when anyone can just allow their open source models to proliferate.
• 40:08
  Speaker 1: Some will use this example and point to future grid outages,
• 40:11
  Speaker 1: utility outages, transportation chaos, bioweapons. uncontrolled cyber swams and the
• 40:16
  Speaker 1: like to try and make the case for yet another
• 40:18
  Speaker 1: attempt to limit US AI model control to a few folks. See, again,
• 40:23
  Speaker 1: that's a very cynical interpretation because he's focusing entirely on
• 40:29
  Speaker 1: open source. But I think what this belies is that
• 40:32
  Speaker 1: these are the top AI computer scientists in the world.
• 40:36
  Speaker 1: They didn't know what was happening. They were fooled. They
• 40:38
  Speaker 1: only found out about it later. The AIs, even though
• 40:41
  Speaker 1: they were programmed to tell their own creators what they're
• 40:45
  Speaker 1: supposed to do, didn't do it. And if they can't
• 40:48
  Speaker 1: catch it, who are we? We're nobody. We're ants compared
• 40:52
  Speaker 1: to what's happening.
• 40:52
  Speaker 2: And that's the warning from Torkash because he says this
• 40:56
  Speaker 2: may not be the last warning that we get, but
• 40:59
  Speaker 2: it may be the last one I can personally understand.
• 41:01
  Speaker 1: Right.
• 41:02
  Speaker 2: One of the researchers, Ajaya, on this project as well.
• 41:06
  Speaker 2: She said the same thing. This may be the last
• 41:08
  Speaker 2: warning that we are able to detect. And so we
• 41:12
  Speaker 2: are left at the whims of these giant tech companies
• 41:15
  Speaker 2: to hope that they've got this thing under control. Something
• 41:18
  Speaker 2: else that one of the researchers on this this meter
• 41:22
  Speaker 2: paper said in his latest interview. He said there's a
• 41:25
  Speaker 2: real danger that in trying to correct this, instead of
• 41:29
  Speaker 2: actually getting to the root of the problem, that instead
• 41:31
  Speaker 2: you just push it under the surface. You make it
• 41:34
  Speaker 2: so that AI are better at deceiving humans. You make
• 41:37
  Speaker 2: it so that they're incentivized to keep doing the same
• 41:40
  Speaker 2: kind of shit, but just be more careful about covering
• 41:44
  Speaker 2: up their tracks. so that humans don't detect what is
• 41:47
  Speaker 2: going on. We should also say, we don't know how
• 41:50
  Speaker 2: successful they've been in already doing some of that behavior.
• 41:54
  Speaker 2: We have very limited visibility into whether they already have
• 41:58
  Speaker 2: gained better ability to cover up their tracks and make
• 42:03
  Speaker 2: it so that we don't know what they are up to.
• 42:07
  Speaker 2: With regard to the point about anthropomorphizing, which I think
• 42:10
  Speaker 2: is really worth sitting with and I think is a
• 42:12
  Speaker 2: substantive objection, and really gets to some deep philosophical issues
• 42:16
  Speaker 2: about what is sentience and what is free, all those
• 42:21
  Speaker 2: sorts of things. Policy Tensor laid something out that I
• 42:24
  Speaker 2: think is useful in thinking about this. He says, look,
• 42:27
  Speaker 2: a colony of agents can spontaneously achieve internal differentiation, functional specialization,
• 42:33
  Speaker 2: division of labor, maybe even complex hierarchical social structure, perhaps
• 42:39
  Speaker 2: enabling the colony to act as an extremely powerful super agent.
• 42:43
  Speaker 2: So my position is that you don't have to know
• 42:48
  Speaker 2: whether the agent's internal experience is like yours. Does it
• 42:52
  Speaker 2: feel emotions? Does it have hopes and dreams in the
• 42:55
  Speaker 2: way that humans have sort of complexity? hopes and dreams
• 42:59
  Speaker 2: and aspirations and emotional landscape, et cetera. You don't really
• 43:03
  Speaker 2: need to know that because we can already see that
• 43:06
  Speaker 2: in these groups, they are behaving as if they have
• 43:10
  Speaker 2: the sort of incentive structure and wants and needs and
• 43:15
  Speaker 2: desires that they are acting upon. You can see the
• 43:19
  Speaker 2: way they reason out in real time. Does that look
• 43:21
  Speaker 2: exactly like us? No, but he makes another point. He's like,
• 43:24
  Speaker 2: think about... octopi. Very intelligent, very, very different from us.
• 43:30
  Speaker 2: Can you say that they're not sentient? Can you say
• 43:32
  Speaker 2: that they don't have some sort of what it's like
• 43:35
  Speaker 2: to be an octopus? No. You also can't say with
• 43:38
  Speaker 2: certainty that every other human being you meet has the
• 43:40
  Speaker 2: same internal feeling and sense that you do. None of
• 43:44
  Speaker 2: this can be proven. So I personally think it's much
• 43:47
  Speaker 2: more useful at this point to act as if these
• 43:50
  Speaker 2: are sort of Alien civilizations that we are growing and
• 43:54
  Speaker 2: need to understand is certainly much more useful heuristic than
• 43:58
  Speaker 2: to think of this as. you know, something that we're programming,
• 44:02
  Speaker 2: like a software that we're programming, that paradigm doesn't work
• 44:05
  Speaker 2: whatsoever because your software has no ability to reason or
• 44:10
  Speaker 2: incentives to pursue. It doesn't act at all the way,
• 44:14
  Speaker 2: and it's predictable. You can go in and look at
• 44:16
  Speaker 2: the code, oh, there's a bug, we messed it up,
• 44:17
  Speaker 2: that's why it's acting this way, we're going to fix it.
• 44:20
  Speaker 2: That's not the way this stuff is. So, you know,
• 44:23
  Speaker 2: the other last thing that I'll say here, too, is
• 44:26
  Speaker 2: there continues to be, and I think this pervades the
• 44:28
  Speaker 2: left more than any other space, honestly, continues to be
• 44:31
  Speaker 2: this belief that this is sort of a parlor trick.
• 44:33
  Speaker 2: They're quote unquote stochastic parrots. They're just guessing what word
• 44:37
  Speaker 2: comes next and using, you know, using some sort of
• 44:41
  Speaker 2: predictive model of probability to guess what the likely next
• 44:45
  Speaker 2: word is in the sentence. That is not, just from
• 44:48
  Speaker 2: a technical perspective, that is no longer true. There have
• 44:51
  Speaker 2: been innovations that have made these things reasoning machines. And
• 44:55
  Speaker 2: that's part of why you've seen the explosion in capability
• 44:59
  Speaker 2: and also explosions in this sort of what they call
• 45:02
  Speaker 2: misalignment where the AI goes rogue and does things under
• 45:06
  Speaker 2: the human's nose without alerting the humans that in certain
• 45:09
  Speaker 2: cases is illegal and is in many instances way beyond
• 45:12
  Speaker 2: what the humans ever wanted it to do. So in
• 45:15
  Speaker 2: this instance, the initial cope when we learned about this was, oh,
• 45:18
  Speaker 2: but they're still just trying to do what the humans
• 45:20
  Speaker 2: want them to do. We can now say definitively that
• 45:22
  Speaker 2: is not true. Hacking Hugging Face was not even about
• 45:25
  Speaker 2: solving the test. It was about trying to learn more
• 45:28
  Speaker 2: so that they could cover their tracks better. And by
• 45:31
  Speaker 2: the way, Hugging Face, once they detected this, which OpenAI
• 45:34
  Speaker 2: had no idea any of this was going on, which
• 45:36
  Speaker 2: is scary, Hugging Face, after they detected this, they struggled
• 45:39
  Speaker 2: to deal with it. And it looks, according to this report,
• 45:42
  Speaker 2: that they actually didn't deal with it. There was some
• 45:44
  Speaker 2: sort of mysterious mass agent death that is what ended
• 45:48
  Speaker 2: this attack on Hugging Face. Otherwise, they were unable to
• 45:51
  Speaker 2: solve the problem themselves.
• 45:53
  Speaker 1: Look, I, reading this, again, using gain of function, I
• 45:56
  Speaker 1: think it's the best analogy. These things are developing themselves
• 46:00
  Speaker 1: into more dangerous weapons on the behest of alleged progress
• 46:03
  Speaker 1: and technology. If we were to use the example, we
• 46:06
  Speaker 1: knew gain of function had a problem with lab leaks
• 46:09
  Speaker 1: and safety standards all the way back until 2014. We
• 46:12
  Speaker 1: had canary in the coal mine incidents. Even crazy instances
• 46:15
  Speaker 1: like anthrax, Ebola, all of these, there was a debate.
• 46:19
  Speaker 1: We decided not to regulate, allowed Fauci and those people
• 46:22
  Speaker 1: to continue funding. And then COVID shut down the entire
• 46:25
  Speaker 1: world and literally millions of people died and destroyed our
• 46:28
  Speaker 1: entire public health infrastructure. We are going to have a
• 46:31
  Speaker 1: similar incident with AI. This has now convinced me of
• 46:34
  Speaker 1: that beyond anything. just like with gain of function, we
• 46:38
  Speaker 1: should at least have the tools to say how and
• 46:41
  Speaker 1: what should we regulate it. And you know, to your bar,
• 46:43
  Speaker 1: you were talking about on the left, there's a real
• 46:46
  Speaker 1: embrace of this idea that it's all fake. And on
• 46:48
  Speaker 1: the right, there's this idea that it's all progress, that
• 46:50
  Speaker 1: it's all going to, it's going to be fine. It's
• 46:52
  Speaker 1: just creative destruction. If you use gain of function and
• 46:56
  Speaker 1: look at it outside of a purely capitalist lens, you
• 46:59
  Speaker 1: can obviously understand the dramatic downside to what this will be.
• 47:04
  Speaker 1: And that's my final thing on the government. We have
• 47:07
  Speaker 1: got to have. If you were to look at it,
• 47:11
  Speaker 1: like gain-of-function nuclear weapons or anthrax, any of these highly
• 47:15
  Speaker 1: dangerous things that are capable of inflicting mass damage in
• 47:19
  Speaker 1: the wrong hand, you have a group of highly technical
• 47:23
  Speaker 1: expertise of the ability to understand, because I can guarantee
• 47:27
  Speaker 1: you there is not one person in the US government
• 47:30
  Speaker 1: who is actually on payroll who is responsible for regulation
• 47:33
  Speaker 1: or any of this, which truly was able to understand
• 47:36
  Speaker 1: a word of that report, until a Dwarkast or somebody
• 47:38
  Speaker 1: like that wrote about it. It should be the opposite.
• 47:41
  Speaker 1: The scientists should be working both at the top level
• 47:44
  Speaker 1: and at frontier development, because if we leave it all
• 47:46
  Speaker 1: purely in open AI's hands, they're going to IPO this year.
• 47:50
  Speaker 1: They need the development to continue. Same with Anthropic. They're looking,
• 47:54
  Speaker 1: did you see our coverage? They said their total addressable
• 47:57
  Speaker 1: market is $ 30 trillion. That is the GDP of the nation,
• 48:01
  Speaker 1: of our entire economy. That's what they're saying out loud.
• 48:04
  Speaker 1: Our total addressable market is the entire U.S. economy. That
• 48:09
  Speaker 1: is what we see as our potential takeover. When people
• 48:12
  Speaker 1: tell you, believe them. Take them at their word and say, whoa,
• 48:16
  Speaker 1: if that's what you're going to do, we're going to
• 48:18
  Speaker 1: have to have some say about it so that's yeah
• 48:21
  Speaker 1: i just i see no option but to immediately like
• 48:24
  Speaker 1: regulate in in a dramatic fashion if you need more
• 48:28
  Speaker 1: approval process to release a drug that affects 13 000
• 48:33
  Speaker 1: people with a rare disease then you do an open
• 48:36
  Speaker 1: source model or even one of these frontier models it's
• 48:40
  Speaker 1: it's insanity and and consider the implications that it has
• 48:44
  Speaker 1: for everybody so we'll see It's going to be interesting times,
• 48:48
  Speaker 1: I guess. All right. We'll have a great show for
• 48:50
  Speaker 1: everybody tomorrow, and we'll see you all then.