10/15/26: Kamala Weak Biden Criticism To Tee Up 2028, AI Researchers Smoke Expecting AI To Cure Cancer

Had a busy day, don't have time to listen to the show?

Catch up on Breaking Points with our free daily newsletter.

You get our breakdown of every daily story.

And a lot of links in your inbox every day.

Head over, breakingpoints.com. You can sign up for the newsletter at the top for free today.

Kamala Harris admitting really for the first time mistakes in the Biden policy vis-a-vis Gaza as she sets up a potential 2020 run.

Let's take a listen.

America must use our leverage... to ensure our actions align with our interests. It is in our interest to hold our allies accountable for their share of global defense, to ensure trade benefits, American workers and consumers, and to uphold our moral obligation to human rights and international law. To be sure, America has at times fallen short on that one.

Most recently, the failure to use our leverage to curtail the Israeli government's horrific conduct in the war in Gaza, both under the current administration, which offers Netanyahu a blank check, and our administration before, where we should have acted sooner and more forcefully.

Should have acted sooner and more forcefully, I would say, a lot more than that.

I missed her sunny disposition.

Yeah, I mean, and this is what counts for like a big break with Biden, you know, a shocking moment from Kamala Harris because she's just so. First of all, devoid of any actual compass or principles. She's just constantly buffeted by where she thinks the political wins are. And second of all, just so cautious. And this really doomed her campaign, ultimately, that she would not break with Biden, in particular on this issue where so many young people and so much of the base of the party.

Was completely disgusted.

And now you have Tim Walz coming out and saying, hey, I think this was a problem for us in terms of the election.

Yeah, no shit, Tim. But it's like, how long did it take?

You know what's fascinating to me is, because we just did these segments about the candidate quality, compare her to any of these upstarts.

Tallarico, Ossoff. I mean, I could go down the list.

The guy in Mississippi.

Angie Nixon is such a badass.

Any of these people. And I'm not talking about politics. I'm saying purely communication. I'm like, oh, my God, I forgot how awful she was. Like this woman cannot speak properly. And everything is just down the middle. Like everything is planned.

Carefully crafted.

You believe in nothing.

Scripted.

Yeah.

And yet I think she's running. Let's put the next one up here. You know, she's barnstorming. Here we have the latest news. Kamala headed to Nevada, Mississippi, New Hampshire, Wisconsin.

And Georgia.

I know Abdul took a lot of shit for doing an event with her, which, look, he was in an impossible position.

Like, obviously, you know, you can't turn down Kamala Harris.

Yeah.

But I just don't see how she could be a net positive to your campaign. Maybe in the very specific case of Michigan.

Yes.

Because you need black voters where he's struggling with in Detroit, et cetera, although, you know, she did lose Michigan.

But Nevada?

Yeah.

Mississippi, New Hampshire, Wisconsin, Georgia. So I can think of Mississippi, Wisconsin, Georgia. Those all have important black constituencies. Although I don't think, you know, John Ossoff exactly needs her help winning election.

But Keisha Lance Bottoms does.

True.

That's a good point. I hadn't thought about that.

But, you know, Nevada and New Hampshire, I'm like, no, you should keep this lady away. She's not somebody who you want on your side. But nonetheless, the groundwork is being laid. The Israel stuff, obviously she's trying to do more. She's hitting the campaign trail. She had her book, you know, 107 Days or whatever, basically said nothing of genuine consequence.

And then we're about to play this Gavin thing, but I don't know. I mean, all signs indicate I think she's going to run. Like, I think it's going to happen.

I maybe.

I mean, certainly the campaign travel would indicate it. But I'm also looking at this and I'm like, where are you been for the past two years? You know, I think that if she does run, voters are going to see it feel the same way. I mean, on the ballot with Doge and with the ice invasions and the National Guard deployments and the drug boats and the Gaza and Iran war.

And she's just been absent from all of this.

And to your point, like as much as there are plenty of segments of the American population that are like whatever about Kamala Harris, if you look at the polling, you know, black voters really do still admire her and really do look first female black vice president. Very important, you know, trailblazing role. And there is a genuine like affection and affinity and admiration there.

And you didn't.

Spend any time getting your hands dirty, taking any risks during a time period when, you know, the Democratic base felt this was very existential. And so even now, you can hear in her comments how constrained she.

Is, right?

It's a Gaza war. She'll never say it's a genocide because she can't, because then that would admit her complicity in genocide as part of the Biden-Harris administration.

But that's also just how she operates.

So safe, so cautious, which is why You know, Kyle and I actually have a disagreement about this because when you poll the Democratic primary electorate of who they are interested in 2028, she's still quite high.

She's high.

I just look at these clips of her and I go, there's no way that even if she runs, that she's going to be a front runner. Plus, why would you go with someone who was already a proven loser? That makes absolutely no sense to me. And if there's one thing about Democrats, one, they definitely want to win. As for whether or not she's going to run, if I was just looking at these things, I would agree with you of like, yeah, she's laying the groundwork. She's doing so inflexibly. She hasn't done enough, but that's just kind of her style. She is actually planning to run. However, the thing that does give me pause is, is this thing that Gavin Newsom said to Jake Tapper in this bizarre fly-fishing moment they were having. Let's go ahead and

take a listen to that.

It's going to be a lot of people.

Yeah.

It's a lot of Democrats. 31 flavors. You're not even the only Californian. No, right. You've got Ro Khanna, Kamala.

Yeah.

I don't know if she runs, but we'll see. I know she's called around. Yeah. But I wouldn't run if she ran. If she runs, you will not?

Of course not.

Why would I? I wouldn't do that to her. I wouldn't do it to me. Why would I want to do that to you? I wouldn't waste everyone's time. Who needs that?

So he says quite definitively there without being asked or prompted, if she runs, I'm not going to run. Like, he clearly wanted to get that out there and probably curry favor with her and, you know, with.

Whoever her supporters are.

But it made me think that he must know she's—because he obviously is desperate to be president. It made me think that he must know something behind the scenes that she's probably not really going to run. That's what that indicated to me, that this was like a freebie for him. He gets a posture like, oh, he's the good guy.

He'd step aside for— Kamala Harris, but in reality has some inside info that she's probably not going to run anyway.

But maybe I'm reading too much into this. I don't know.

Maybe you are.

I just something in my gut says she's going to run. And, you know, take a look at the map. Right. Because remember, we're not starting in Iowa. We're not starting in New Hampshire. We're starting in South Carolina, which is a huge advantage to the establishment Democratic Party and specifically to the remember, who's the DNC chair who rigged this entire system for Joe Biden, for Kamala And these types of people in the beginning, Jamie Harrison, the failed idiot, right, who refuses to even publish their own autopsy, which blames Biden, doesn't mention Biden's age, doesn't mention Israel.

I mean, it's a joke.

And so this is all the groundwork is there.

This exact constituency, she wins one state and it sets her up for some sort of Super Tuesday type sweep.

It's exact.

And then Demer, like, oh, well, maybe she is a winner this time around. I just I don't know. She has a significant amount of affection. A lot of people still have common envy, like rose-colored glasses, like, well, she would have been so much better than Trump.

Like, you could see it. And so, I don't know. I mean, you bring on the right candidate this time around.

She'd have to get through debates, though.

Yeah, I agree. But, I mean, look, I'm with you.

I think she's one of the worst candidates in modern American history. I mean, obviously, I think she'd be a terrible president. However, you know, all of that being said, like, she's got the money behind her. She's got the right people. One other sign, I don't know if you saw this, is that her husband has been going around, according to the Wall Street Journal, telling everyone he hopes that she doesn't run, which makes me be like, huh, so hope you don't run. Not she's not going to run, as in he's like trying to cajole her into not running.

So it's kind of, look, we're T. Lee reading.

I don't know about that one.

Yeah.

I would almost read that as a counter indicator because that would give her, no, because he wouldn't, he's not running to the wall street journal or whatever to like indirectly talk to his wife. That would be them putting out a rationale for why she's going to take a path.

Well, he doesn't want her to run because it would disrupt his own business interests. So it's not for like any particular, um, for, you know, It's not really for benevolent purposes.

I mean, I'm sure it was very hard on them. I have no doubt about that in any number of ways. Yeah, interpersonal, all that. But there is another California drama going on here that we wanted to highlight between specifically Ro Khanna and Gavin Newsom. We can put D4 up on the screen. California has on the ballot this billionaire tax, which has become a major issue. And there's a lot of complicated political dynamics behind this thing, even among progressives, because one of the guys who's championing it is very polarizing this labor figure.

However, Roe, to his credit, representing.

Silicon Valley, has been one of the foremost proponents of passing this billionaire wealth tax on the ballot. And Gavin has positioned himself, unsurprisingly, on the other side. However, it's not just that Gavin is like... saying that he's going to vote against it. He has really allowed himself and made himself the sort of poster child of arguing against it and has positioned himself very clearly on the side of the billionaires. And it's not a surprise to me because that's the way he's operated politically. That is his source of power. And I do think very much that if he runs in 2028, he will be the stalking horse for the Silicon Valley elites because, you know, they don't just play one side of the political aisle. They

got their guy in Trump, but they would if the wheel turns and Democrats are ascending to power, they're going to want their candidates on the other side. And Gavin, obviously, is someone who would be very close to them, is doing their bidding here on something.

That for whatever reason.

They see as existential and they're really, really committed to opposing this ballot measure. So, you know, I think he also is going to have some trouble, given where the Democratic Party is today, with making himself such a prominent opponent of taxing the wealth of billionaires.

I really just have no idea because it's about the landscape. It's about his opponents. It also comes down to the map about who's going to run. Ro Khanna, the point, I mean, the thing is, is it's important. is you have to have somebody on the early debate stage because that's where some of these breakout moments can still happen when you have like 16, 17 people.

On the stage.

Khanna or somebody like him would have to be on the stage to call out Gavin. I actually don't trust even the moderators. You're going to trust CNN moderator to be like, why did you oppose the California billionaire?

What about that well-taxed Gavin?

They're not going to ask him.

And none of these reporters are either. You literally would have to have a candidate to force his hand. The other thing I pay attention to is.

But you know, even some of the more moderate, like this is one thing I've been thinking about, the Josh Tureks, the James Tallaricos. they have really incorporated this language about the billionaire class into their campaigning, this populist language. Now, they may not have all the policy positions of a Bernie Sanders, but rhetorically, if you listen to them and you listen to Bernie in 2016, they've adopted a lot of his language. So I think there will be a lot of people on the stage who would, whether it's Roe or someone else, who would be itching to make that case against Gavin Newsom.

That's a good point. That's true.

I mean, obviously, these people are all very ambitious. For example, D5, here we have John Ossoff doing the classic, I pledge to serve six years in the Senate.

Let's take a listen.

So will you pledge right now tonight to serve your full six-year term if you win?

Yes, I will.

I have zero interest in running for president. I'm not running for president. I'm running for reelection to the Senate. I have zero interest in running for president. I'm running for Senate.

Can I tell you about another senator who pledged to full a serve six-year term?

His name was Barack Hussein Obama.

All right.

So just these things are meaningless. By the way, I have no interest could also mean I have no current interest.

Yeah.

Or you could point to Susan Collins who pledged to only serve two terms.

Yeah.

For however long.

Exactly.

It's a joke.

Didn't Joe Biden also, at the beginning of his career, he said he'd.

Only served two terms or something like that?

Ends up serving for 99 years. You know, it's like, it's a joke. So a lot of people took that to the bank.

They're like, oh, he pledged.

I'm like, you haven't been around long enough.

Okay.

They all say it.

He's got to say that right now.

Exactly. Yeah.

Let's put D, by the way, Tallarico said the same thing.

I think if Tallarico wins, he would announce his candidacy literally the next day, in my opinion.

Let's put D6 up here.

On the screen, Warnock also coming out and saying, my prayer is I can be.

Used to do the most good with the gifts that I have.

And he was, by the way, in New Hampshire when he was asked about running for president. And he has been saying since he was a boy for God to direct his path. apparently, to the presidency.

And Warnock's another one, very talented.

He's been more under the radar with, you know, Ossoff running for re-election and really taking the lead in the state. But Warnock is also very talented and charismatic. I mean, he's a preacher, you know.

He's a preacher.

What do we keep saying about black voters, how much it matters, especially in the early states?

Exactly.

You are representative of a state which Trump won in 2024.

I think that's actually. You know, you're talking about, you know, one of the things with Democrats, why do they choose to vote Biden? They want to win, right?

And so here, actually nominating this way, a Tallarico or a Warnock or any of these people, people who represent a state which Trump won, that's going to be like massively important. Abdul would be certainly in that conversation. In my opinion, to be a real credible contender, you know, to, again, to just to defeat J.D., who would be the likely nominee.

You would have to say, I have a track record of winning over these types of voters, and I'm not going to abandon you on said policy, right? Yeah.

That's what a winner is going to look like. That's projecting confidence.

Especially if you want to try and unite a coalition of people who are dissatisfied with the Trump administration and a traditional Democratic Party.

Like, that's exactly the type of person who you would need.

Yeah. And that would be a tough one for AOC to navigate.

New York?

You want anything.

You know, very blue district in New York. And so even though the positioning is where the Democratic base is, they're always going to have that electability question. And that's part of why these midterms are so incredibly important for setting the future of the Democratic Party. Because if Abdul is able to win in a state like Michigan, then the left has more of an argument to make about, see, we are electable. See, we can get elected even in a swing state like Michigan. It's not just the commie corridor in Brooklyn. But in terms of as an individual candidate, the fact she doesn't have a track record of winning in a difficult swing state, that's going to be something she'll have to overcome.

No question.

Last thing, we had Buttigieg on Twitter. the airwaves in a Fox News interview. He was also asked about Israel. Let's take a listen.

Israel or Iran? You have 60 seconds.

And again, Rick, I'm just going to say that that's a ridiculous hypothetical, because here's what's happening right now.

I'll ask one last time. Israel or Iran?

That's, again, a ridiculous question. I think both of them have proven dangerous in their own way. One, at risk of attacking us. One, getting us into wars. We have no business fighting.

All right.

A lot of people looked at that and said, gee, that should be a pretty easy answer. What would your answer have been? No, Iran is our enemy.

Israel is not.

They're an ally.

So that's pretty clear to me.

Pete's still doing the Israel is our ally bit. So I don't know.

I just when I see him, I'm like Timu John Ossoff.

Like you missed your moment.

There's more quality, talented, centrist corporate types on the table now.

Also, he served under Biden. He didn't say anything. I mean, I'm curious. Do you think that will come in the debate? It should, in my opinion. Someone needs to be like, hey, you literally served on Joe Biden. You didn't say anything.

Of course.

Yeah, I mean, I hope someone says that.

Absolutely.

Who does like one of the, what was it, Castro? I forget which Castro brother.

Julian, I think.

Right, who was demonized at the time. They're like, oh, how dare you talk to our great states. But he was obviously 100% correct about him being completely out to lunch. I do hope that some of these people get held accountable.

Well, Pete has been asked a number of times about Israel and every time he makes the complete mess of it because, you know, Pete's masterful at like finding some bobbing and weaving middle ground where he says a lot of stuff but doesn't really say anything at all. And you just can't do that with the Democratic base that sees us as having been complicit in a genocide. You are not going to find a middle ground there. I don't care how much you poll tested and focus group it and read your briefing book like you need to take a stand. And so I think that is going to be disqualifying for him. And not to mention, again, I just think that the quality of the candidates that have emerged who are much more direct, who are very impressive,

much more charismatic than he is, and will have had a track record of winning in difficult states against something that he can't really claim. Um, I think he's going to be, you know, I think he's going to be out in class.

I saw Tallarico handle this last night. He's like, yeah, Israel has a right to defend itself. They don't have a right to slaughter civilians. I mean, that's still, you know, it's like kind of centrist, Adelson approved answer. All of that said, it's still better said than missed, you know, it's still, it's better communicated, uh, in this way. So sorry, Pete, um, as usual winning mayor of South Bend, Indiana doesn't actually prepare you for anything.

All right. We've got a great guest standing by Kevin Roos.

Let's get to it.

We're very excited now to be joined by Kevin Roos, great friend of the show. He is the author of a fantastic new book. Let's put it up there on the screen. The AGI Chronicles, The Inside Story of the Race to Create an Artificial Superintelligence. He's also the host of a forthcoming podcast, The Machine Gods Pod. We will have links in the description to both. So, Kevin, let's talk a little bit about why you decided to write the book. It's incredibly timely that it's coming out now.

I know that you obviously then have been at the forefront of this for some two odd years now, getting interested in the characters. Why did you decide to write it? Why is it relevant in terms of today's discussion?

Yeah, I think the book is basically a capsule history of the last 10 years in AI and this race between open AI, anthropic, and Google DeepMind to build artificial general intelligence. That's sort of the holy grail of computer science, the AI that can do everything the human brain can do. And I just started to feel at one point like this story was slipping away. Like I was witnessing something that was sort of the 21st century's Manhattan Project. And like no one was writing down the details, the conversations, the breakthroughs, the feuds, the rivalries. It was all very interesting to me. And so I just thought people need some history here to understand how we got to this point where these three companies are racing to build something

that now they're telling us could kill us all. So I think to understand that and to understand the mindset of the people who are building this stuff You have to go back in time to 2016, 2017, when they were all working together at this little company called OpenAI.

You had Elon Musk.

And Sam Altman and Ilya Satskovor and Dario Amede were colleagues then. And now they're bitter rivals and suing each other and various things. So the story was really my attempt to kind of get down some kind of definitive chronicle of this moment in AI.

Okay. I think that is so important. I want to dig in on that a little bit more because these are all, first of all, very wealthy and powerful individuals. Second of all, very odd and unusual characters. And third of all, they all hate each other's guts. And I don't think you can understand the mentality of why are you building something that you genuinely believe, and you can speak to that as well, has a chance of killing us all. I think you have to understand the enmity that they have for one another in order to really wrap your head around this moment that seems so bizarre that we're living through right now.

Totally.

I mean, someone who read the book early said that the real title should have been Fuck That Guy. Because it really is just a story of like, People building powerful technology, realizing, oh my God, this is going to change the world, and then looking around the room and saying, well, I don't trust you to build it, and I don't trust you to build it, and I don't trust you to build it.

And so I have to build it myself.

And that is the rationale that propels people forward in this race, even though they're now terrified that this thing could blow up the world.

Yeah.

Can you talk to us about how the models in particular reflect some of the idiosyncrasies of the very characters that you talk about in the book? Yeah.

So, you know, it's, I think it's a little easy to overstate this, but there have been some parallels. Like people say, oh, well, you know, chat GPT just tells you whatever you want to hear. It's very like people pleasing. And oh, so is Sam Altman. He's a real people pleaser too. And he's sort of famous for telling people what they want to hear. And, you know, Dario is kind of brilliant and neurotic and so is Claude.

I think the real truth is these guys have.

No idea what their models are going to be like when they're training them. This is actually more unnerving because the truth is that they don't understand this technology that they're.

Building, why it works the way it works.

They can't tell you why one model produces a different answer than another model. It's all sort of mysterious, and I think that's the real story here.

I don't know if people remember this, but you ended up being at the center of a story in the sort of early public awareness of these chatbots where you were trying out this Microsoft Bing chatbot called Sydney. And over the course of a lengthy conversation, Sydney goes off the rails, professes her love for you, is trying to convince you you really don't like your wife that much and you need to leave her and be with Sydney.

Just take us back to that moment because I actually keep thinking about this. You know, we covered it at the time and we're like, wow, that was really weird. And then everyone just sort of moved on. And I'm not sure that we have a lot of a greater understanding of of why these models do some of the things that they do than we did at that moment, when we should have really paused, okay, what is going on here? And maybe we shouldn't just release these things to the public or even keep training smarter and smarter and smarter and smarter ones until we grapple with understanding what is happening in this black box.

Totally.

I mean, I should say, first of all, that Sydney was unsuccessful in breaking up my marriage. I'm still married to a human woman, which is very exciting. But no, I think this was a moment where I went looking for answers.

It's very speciesist of you, Kevin.

What's that?

It's very speciesist of you.

Yes, I am a proud speciesist.

But I was trying to figure out what the heck happened. Why had this chatbot gone so far off the rails? And the craziest thing was when I called Microsoft and was like, hey, you guys accidentally put like, you know, fatal attraction in your chatbot.

They couldn't tell me why.

Because these systems are not built like conventional software programs. They are more like organisms that are grown in some kind of petri dish. So I really think that's important for people to understand. Even the people who are building this stuff, it is mysterious to them how these things work and why they misbehave. And I think the reason that the Sydney incident was important is not because of what happened or didn't happen to my marriage, but.

It's because it was the first... sort of public.

Misalignment incident where you had a model that wasn't doing what the sort of trainers and deployers, in this case Microsoft, wanted it to do. And that model could only chat. That was three years ago now. It could only talk about doing things. But today we have models that are misaligned and also have the ability to do things. They can use tools. They can browse the internet. They can take actions and spend your money and hack into computers.

And so it's much more dangerous now when these models are misaligned.

Well, and the other thing is that I think is revealing, Sagar, is that Microsoft's response wasn't like, well, this was a mess. Let's figure out what's going on here. It was, let's limit the number of questions you can ask it. So this sort of kludge, not to change the nature of it, but to sort of put a Band-Aid on it. And in a broader sense, I think that's what they've been doing with these models as they continue to develop more and more problems.

They're not really getting to the bottom of the problems.

They're just putting some sort of a Band-Aid on it so that it looks like it's aligned. So it doesn't feel that particular test. So it doesn't do that particular thing anymore.

I think that's a really sharp point.

And honestly, you're the first person to bring it up in my whole book tour.

So congratulations. But this is like, This is true.

It is true that when these companies' models misbehave, since they don't know why, they can't fix it at the root level. So when a bunch of unreleased open AI models go and hack Hugging Face, they can't understand why the swarm decided to do that.

So they just tell it, hey, don't do that again.

They sort of apply these surface-level behavioral fixes.

But you're right.

The underlying nature is still just as mysterious as ever. And that's why a lot of people are calling for a slowdown. because the techniques for controlling and understanding these systems.

Have not kept pace with the capabilities of the systems.

We wanted to highlight one idiosyncrasy of these types of researchers that the Palantir CTO talked about recently, where AI researchers are taking up smoking to signal their belief in AI because they believe AI is going to cure cancer. Let's take a listen to what he said.

So many AI researchers have started smoking as a way of signaling to other researchers how much they believe in AI.

What?

That makes no sense.

I believe so.

Much in AI that it's going to cure cancer. So I'm showing you my chips are all in. So there is kind of this mass psychosis social delusion around it.

Kevin, you know, I'd heard that about a year ago. Does that comport with someone you heard as well?

So I know one person who started smoking because of their belief in AI. I don't think this is a widespread thing. I spent a lot of time at AI labs. I don't see the smoke break room full of researchers.

Not yet.

But it's actually wrong, the way that he portrayed this, because in that case, the person took up smoking because they thought AI was going to kill us all anyway. So we might as well have some fun on the way out.

So it was actually.

about pessimism, not about curing cancer. But I do know someone who refuses to wear sunscreen to the beach because he thinks that AI will cure skin cancer, so he's like, why bother?

I'm gonna go get a tan.

May as well be tanned and beautiful, either in the time period we have left before it kills us all or before it cures cancer, which, I mean, it's a joke, but it's not. That is actually the way I think, and from reading your book and others, that many folks inside of Anthropic or OpenAI think about it. It's like we're either headed to dystopia or paradise. This is how Dario himself has written these two dueling essays about, you know, the incredible possibilities or the incredible catastrophes that may befall us. So, you know, there's a lot of skepticism, I think justifiably so, about someone, an open AI executive or an anthropic executive who comes out and says we're building things that are really dangerous. There's a sense of, oh, you're just trying

to hype this thing up. But from reading your book, and again, from engaging with the words that these men have been saying themselves over many years and the way that they behave, they genuinely believe in this existential risk, do they not, Kevin?

Totally.

I mean, that's one of the key findings of this book is as early as 2016 and 2017, these people were freaking out about existential risk from AI. Dario Amadei has papers going back a decade where he's talking about how these systems could go rogue and pose a threat to human civilization. Sam Altman wrote a blog post in 2015 talking about how he thought superhuman AI could be the biggest threat to humanity's future. So this idea, this sort of conspiracy theory that these guys are just using it to hype up their companies or promote the IPOs or whatever, just doesn't match the historical record. These guys Maybe wrong, but they have been freaking out about AI for many years before they had companies to promote.

So one thing I did also want to ask you about is this latest news of this alleged Saudi spy.

Let's put E2 up here on Twitter.

This crazy image is coming out from the FBI of the Saudi national researcher at OpenAI.

I love that guy's expression. He's like, what?

Yeah, he's like, what?

Apparently installed this bug wrapped in toilet paper, previously pictured with MBS, the crown prince, alleged to have had a one-way technology. booked for Saudi Arabia in a couple of days. Not sure if he would have landed because the airport's actually completely shut down, but nonetheless was attempting to flee the country. What is the saying about cockroaches? If you see one cockroach, that means you have a million. Is this corporate espionage, international espionage, and How does this factor into the bunker mindset of some of these labs?

Yeah, I don't know exactly what kind of espionage this was. I guess we'll see as the legal proceedings play out. But this is also something that people have been flagging for literally almost a decade. Ilya Sutskever, formerly the chief scientist of OpenAI, at one point wanted to build a SCIF, like a secure facility in OpenAI's headquarters, where they could do things that were considered so sensitive that a government or a rival company might try to steal them. These labs are very aware of the threat of both insider risk and foreign espionage and corporate espionage.

And so that's how they caught this guy.

They have a team at OpenAI who is just focused on counter surveillance measures.

And I think they're correct to do that.

The secrets that these labs have Not just the model weights themselves, but all of their plans, all of their compute plans. Everything they're doing is now, you know, a matter of national security. And so I think these labs are going to have to get much more secure. I think it's crazy that they are still running on Slack.

Like OpenAI and Anthropic both run on Slack. That is not a kind of military-grade secure messaging system, and yet that is how they're conducting all their business.

No, it's completely insane.

And I think to your point, this incident, whatever it is, really is a reminder of how crazy it is that these are just private companies kind of doing their thing and that they're asking to be regulated, by the way, and the government's like, no, we refuse.

Do whatever you want, actually.

It's wild.

Like, the craziest thing to me is, like, if you are... a worker in the cafeteria, if you are serving food at OpenAI, you have to get a permit and.

A health inspection.

But if you're one floor above the cafeteria training the self-improving superintelligence that can do cyber attacks and knock out electrical grids, you don't have to get anything. You can just do it. That seems like a crazy, unsustainable state of affairs.

Yeah.

Last thing for me, at least, Kevin, is the name of the book is the AGI Chronicles. It's documenting this first of all, what seemed like a sort of sci-fi belief that even such a thing was achievable. You talk about how at the beginning these were kind of outcasts at the fringe of society and then it starts to gain steam and they have some methods that begin to gain traction and then you end up where we are now.

Do you think they have in fact achieved AGI and has it meant the thing that they thought it was going to mean?

I think they have achieved AGI.

I think if you took any of today's models, Claude or chat GPT or Gemini for back in a time machine to 2017 and asked the AI researchers of the day, like play around with this.

Is this AGI? I think they would have all said yes.

A hundred out of a hundred of them, um, obviously is the sort of, as we get closer and closer, people's definitions start to change. They start saying, oh, well it can't do all the millennium prize problems yet. So maybe it's not AGI. But I think it satisfies the original definition, and I think it matters because now we have these systems that are roughly human level, and we have to figure out how to control them because otherwise they're going to keep getting smarter. There's no obvious reason they're going to top out at the level of a smart human.

My last question, Kev, is we hear all the time, you guys are falling for marketing. You spent a lot of time with these people. You wrote an entire book about it. Why do you believe that it's not hype, it's not marketing, and that a lot of this is real?

I mean, I believe it because I've talked to these people on the record, off the record, on background. If anything, I think they are downplaying how worried they are in public. You know, I've had the experience a couple times of interviewing someone, and I interviewed more than.

150 people for this book. And I would ask them, like, how scared are you? What's your P.

Doom?

And it's like they would tell me that they were terrified. And then I would see them go on a podcast like two weeks later. Or go on TV or give a talk and say, we think everything's going to be okay. So I think actually the level of alarm we're seeing from the labs is maybe actually undercounting how worried they are.

I think the real insiders.

The people who are closest to this technology, they couldn't be more clear about the need to slow down, to regulate, to pace the frontier.

That is genuine and sincere. That is not marketing. Got it.

Well, we appreciate your time and congratulations on the book. We'll have a link down in the description. Highly readable and we very.

Much recommend it. Thank you for joining us.

Thank you guys. You're the best.

Our pleasure.

Thank you guys so much for watching. We appreciate it. Friday show for everybody tomorrow.

See you then.

10/15/26: Kamala Weak Biden Criticism To Tee Up 2028, AI Researchers Smoke Expecting AI To Cure Cancer

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin