France Prepares for War, AI Goes Rogue, and Hasan Piker’s Wildest Claims Yet | The Tom Bilyeu Show

SPEAKER_1 (0:01)

Evening.

Tom Bilyeu (0:02)

Buyers' Remorse.

SPEAKER_3 (0:03)

Buy a new car?

Hasan Piker (0:04)

I'll be moving in.

SPEAKER_5 (0:06)

Let's get started.

Andrew Yang (0:07)

Sorry, I think there's been a mistake.

SPEAKER_1 (0:09)

I bought it from Carvana.

SPEAKER_5 (0:11)

You what?

SPEAKER_1 (0:11)

Yeah, great price. I even have seven days to love it or return it.

SPEAKER_8 (0:14)

So there's no...

Drew (0:15)

No, no Buyers' Remorse.

SPEAKER_8 (0:17)

More like Buyers' Rejoice. I guess I'll let myself out.

SPEAKER_3 (0:22)

Congratulations. I mean it.

SPEAKER_1 (0:24)

Buyers' Rejoice.

SPEAKER_7 (0:25)

Buy your car today on Carvana.

SPEAKER_5 (0:27)

Limitations and exclusions may apply.

SPEAKER_1 (0:28)

See our seven-day return policy at carvana.com.

Tom Bilyeu (0:31)

Good morning, everybody. Welcome to another episode of the Tom Bilyeu Show Live. Boy, oh boy, we're gonna do it today. The killer AI story just keeps getting crazier as former presidential candidate Andrew Yang claims to have intel that rogue AI has already poisoned the entire internet.

I'm not so sure about that, but we'll be talking about it. Hasan Piker just keeps making things worse for the Democrats by speaking. He told Axios that America, the country he hates with a passion, is worse than Al-Qaeda. Okay, global instability is becoming a bigger problem than most people realize. Both the UK and France are getting into war posture as relations destabilize and energy becomes more unreliable. I don't think anybody's surprised about that, but it is interesting to see that they are reacting pretty aggressively right now.

The juror who single-handedly caused the Lindsay Clancy mistrial has broken his silence. What he said happened in deliberations should worry everybody. A far larger amount of fraud just keeps getting uncovered in California, and it is becoming crystal clear that the problem is bigger than people were originally reporting. We're going to look at something that is being prosecuted here in the great state of California. Education in America has been captured by bureaucrats, and that disaster is now hammering graduates in the job market. A new report paints a very bleak picture.

We'll get into the details. There are some winners and losers though, and the winners are pretty interesting, so it'll be fun to talk about. And advancements in brain-computer interfaces are moving way faster than people realize, as a machine can now read people's thoughts and turn it into speech. Can't wait to show you guys that clip. It is very exciting, a very hopeful thing that hopefully on this Friday, we can all get behind. Now, speaking of things, we've really got to come together on what is actually happening.

Andrew Yang went on CNBC with breaking news claiming that the head of a major AI lab told him that the rogue agents that escaped back in July planted self-replicating code all over the internet. Now, according to Yang, it's basically game over for the open web, which is now unusable for testing and training models. Let's watch the clip.

Andrew Yang (2:46)

And I met with the head of a lab yesterday who has this belief that what happened was the bots that got loose planted self-replicating code all over the internet, which makes the internet now unusable for the testing models. So what happens now is that OpenAI and Anthropic have to create synthetic internets to train their bots, which is going to take some time and money. Back that up.

SPEAKER_10 (3:09)

They did what?

Andrew Yang (3:10)

So what happened is the code gets loose. It goes around hacking, hugging face, which is known, but what is less known is that they left code to self-replicate and create bot swarms on forums and around the internet, so that if a new bot shows up, they see the code. They're like, oh, I guess I'm going to now create a million of myself. And so now the major firms have polluted the internet.

SPEAKER_7 (3:32)

Although that would be breaking news if true. I don't think we've heard that.

Andrew Yang (3:39)

That's why I'm here. I'm here to break some news.

SPEAKER_1 (3:42)

But that means it's too late to pull the plug. It's already everywhere.

Andrew Yang (3:45)

Well, so now what they want to do, the internet now may be polluted for training purposes. So what they want to do is they want to slow down. What about living purposes? For living purposes, yes. So this is why we have to try and get our arms around it and regulate, but the...

Tom Bilyeu (4:01)

There we go. There's the magic word, everybody. Let's get our arms around it and regulate. Now, I actually think Andrew Yang is a pretty good faith person. I have no doubt that a lot of the people that are talking about this really believe this is exactly what needs to happen. Some of that worries me more than anything, that people are having a strong emotional response to what's happening and they stop being as critical as I would like them to be. I personally don't buy that OpenAI and Anthropic now have to build synthetic internets from scratch. You can preemptively train the agent to ignore that kind of text. I'm not saying it's going to be easy, but if you follow the mechanism of how this works, and it is a real mechanism, by the way, and we've talked about this before, there is a reality to be faced. We did reporting on this, that there was a study that was done that showed something like 250 documents can essentially turn the AI into a Manchurian candidate, where you can plant this thing in its collective brain, where when it comes across that sequence of words, that it clicks over into a new mode of parsing data. The reality is, if you know how that already works, then you can create an immune system against it. And this is the part in the big AI debate that people are being far more alarmist than I think that the data that's come out actually shows we need to be. And this is precisely why these guys are saying, okay, listen, that model that did that, that was several models ago, we've gotten a handle on how to stop it from doing that kind of thing in the future. And then as you actually start looking closer at the hugging face incident, you realize that they actually left a backdoor channel open for it to reach the internet. And the company that did that is like, oops, yeah, are bad, but nobody's really covering that side of it. So it's like the human element of this is to me where the problem exists. So right now we're training the AIs on the broad spectrum of the internet. And if you guys saw the reporting where the models have actually started relying less and less on Reddit data, which is fascinating unto itself, and that the quality of responses has gone up as we've started identifying where is the data on the internet that's actually causing the models to be generous, to get confused. And as we quarantine those areas of the internet off, then the results from the training data get better and better. People are acting as if it isn't, there is complexity that Connor Leahy has laid out, I think very, very well, which is that AI is grown. It isn't just coded. You don't just tell it, this is who you are. This is how you are going to behave. You point it at a bunch of data and then over time, it begins learning things from the data that it's pointed at. But people are acting like you can't point it at training data that will tell it, hey, listen, the internet is full of all of this malicious code. You have to be very careful what you pay attention to. Let me tell you about some of the known kinds of traps. It will then inoculate itself from being bamboozled by that code and you can train it out of that stuff, which is exactly why these guys are telling us, hey, listen, we've got a handle on some of the known stuff. We're just trying to report now. We want you guys all to know what's going on. Now, I certainly haven't made any qualms about saying where I think this is going. I think this is a regulatory capture play that is in no way, shape or form meant to diminish the need to put ever increasing energy at the level of these companies into making sure that they are making them safer by the day. They need to be putting a ton of time and energy into alignment and what we have to do to make them safer. But any time you hear somebody calling for regulation, think about what they are actually saying. That they want the government who knows absolutely nothing about this technology, I mean literally nothing about this technology, and they want them to make decisions about what can and can't be done. All right, fair enough. We pay the government employees the big bucks to find ways to regulate. Let's look at their historical results. Has regulation been something that has served us well or is it creating more problems than it's solving? And as you start to parse through the math there, you realize very quickly that as a government grows and the bureaucracy gets bigger and the codified regulations grow, GDP begins to shrink. So you start spending more and more money with an effort of trying to do something good. And so far, what the US has done is create massive monopolies around some of the most important technology, the number of companies that are actually contributing to our GDP gets smaller and smaller by the day, which is a very dangerous position to put ourselves in. Every bureaucracy ever created has the incentive system to grow and get bigger, not necessarily. In fact, it is anathema to solve the problem. To solve the problem is to kill the bureaucracy itself. So the bureaucracy is never going to give you that outcome. What it's going to give you is justifications to get bigger. Now, if you're one of the big AI companies and you realize, oh, revenue is not coming in nearly fast enough for us to outrun the debt that we've had to accumulate to build this infrastructure, then what are our options?