ExclusiveAI’s Most Dangerous Turning Point

If the CIA or NSA detects that some of these companies, like OpenAI, XAI, or Anthropic, are about to cross to the self-improvement threshold, they're saying the result would be that the Marines and the FBI will secure the buildings, will physically confiscate the GPUs, the NVIDIA GPUs, and the employees will be detained under a state secrets act. So that's how serious this is. It's pretty crazy, I was not aware of it. Welcome to Silicon Valley Front Row, where we bring you the front row seat in Silicon Valley. I'm David Lam and this is Steve Ispas. So Steve, we're gonna be talking about different levels of artificial intelligence and where we're at. So what do you know about this? Well, I had to do quite a bit of research. We talked to some experts and somehow I got a feeling that this is such a big and important thing that maybe we'll try our best to educate our audience as much as we know, as much as we can. And they can, of course, do their own research, but it seems like it's a really major thing. It's something that, especially talking about the last level of AI, it's something bigger than anything that ever happened on this planet in so many ways. Yeah, like reaching the ultimate form or phase of the AI. And what's interesting, it seems like it's easy to tell when we got there. That's the thing I was trying to understand, like how do we know that we got to this phase? So maybe let's look at the different artificial intelligence. Yeah, so the three types of artificial intelligence are artificial narrow intelligence, ANI, artificial general intelligence, AGI, that's stage two, and stage three is artificial super intelligence, ASI. So it seems like we're kind of in the middle of the middle right now, in the sense the top one, I think maybe a better diagram is here where it shows this pyramid. And then we have kind of Siri, Alexa, and Cortana. These tools are the narrow ANI. And then when you talk to AGI, we're moving towards things like JGPT, Grok, Gemini, and Copilot. And they're still kind of in the middle of it. And then we get to ASI, artificial super intelligence. So that's the part I wanted to talk about most because I think that's the most fascinating and scary, mostly scary. Yeah, something that supposedly could- Shut down the whole earth, pretty much. Yeah, and be better than humans. So I think the first question we ask is like, what can happen if ASI is reached or close to be reached? So when ASI is here, I think the world would just be a different place. But other than that, I'm not sure exactly how it happened, but let's see what Grok says. So Grok says, once any government becomes convinced rightly or wrongly that a lab is less than three to 12 months away from ASI, declaring a national or global emergency becomes one of the highest probability events. And that's far more likely than a quiet, peaceful transition. That's kind of interesting that they would take such drastic measures to address this. Maybe I'll read the next one. Here, we address what happens if it's detected that a model is about to move to the self-improvement threshold. So the intelligence reaches a limit where it can pretty much solve any problems. It can overnight develop or get Nobel prizes. It can develop that kind of science or in medicine, in all fields, can be like the top echelon and solve the world's problems. And at the same time, it can get into this self-improvement loop. So what it means is that it starts to rewrite itself. Like maybe there's a piece of code that this exists in and it rewrites itself to make it better. But it does it faster and faster. So it can take, I don't know how long, maybe hours or minutes or days that it rewrites itself. So every time it rewrites, it becomes that much more efficient or that much smarter. So I think what Grok is saying in this case is that if the CIA or NSA detects that some of these companies like OpenAI, XAI, Anthropic are about to cross to the self-improvement threshold when basically nothing will stop them anymore. They'll just improve at an unbelievable speed. They're saying the result would be that the Marines and the FBI will secure the buildings, will physically confiscate the GPUs, the Nvidia GPUs, the chip that actually runs this. And the employees will be detained under a state secret act. So that's how serious this is. It's pretty crazy. I was not aware of this. And this is coming from Grok and AI itself, right? It's kind of predicting these things. Yeah, if the FBI and CIA has to come in to confiscate the servers or things, that is like another level of national security. And then we asked Grok another question. How do we know that ASI is here? According to experts, there are several pins or several signs that we know it's here. One, it'd be huge layoffs. Now again, that takes time, but it could happen within days that there's like huge layoffs because people are no longer needed. So like one ASI, can it just take over like an entire factory of computer scientists? I mean, not factory, but a firm? Probably, probably. And do the work of all those people and more. Wow. So imagine like the entire coding department of Meta. Yeah. And then it just resorts down to just one computer. But I think it's not just the computer department. I think it would be pretty much robots as well doing the work that people do. And again, if the AI reaches the stage where it's self-improving, I think they are able to build robots. And as it self-improves even more, to build them very fast. So it could be, you know, they can have all the factors automatically start doing things and create these robots in a matter of hours or days. Yeah, that's really interesting because I was only thinking from a virtual perspective, but physically, if it can manifest in Android form, that is pretty scary, very interesting. So another factor is an existential and cultural hysteria. So what that means is Grok says that there would be public panic over losing our monopoly on intelligence. We could think about viral memes that we've seen about AI taking over human souls. Philosophy podcasts could be exploding in the realms of religions. They could update their doctrines to include machine souls. That's what Grok says here. This overall idea of an existential crisis as people realize that they're becoming less capable. There would probably be debates on X Reddit about sparks of AGI. Newspaper articles or op-eds could be warning about an era of unlike any before. Another step would be what they call government regulation. And this one, we talked about already how governments would declare a national emergency to deal with it. And this probably would be at all levels. There would be international treaties on AI like nuclear non-proliferation. Let's see what else they talk about. So on the positive side, if things don't get destroyed before then, it would be a huge acceleration in scientific and medical breakthrough. Yeah, that could be a good thing. I think the crucial issue on all of this is that as long as the AI function is fully aligned with humans, then it's okay. But even if it's just a little bit off, then there could be major problems. To me, I think that's probably the biggest challenge. How do you keep that fully aligned? Because even if you create rules, like again in the iRobot movie, or you create rules that robots should never harm humans, if it starts rewriting the code itself, at that point, they can decide, well, we're gonna change this rule. Why do we need this rule exactly? Well, there's a lot more to this, but we just wanted to give our audience a sense of how serious it is. Now, another question we ask is like, how close are we to this? And then maybe let's just give a few examples. The short answer is between a few months and a few years, but there are a few people that give specific numbers. And Elon Musk thinks that 2026 to 2028 would be the timeframe. That's close. That's very close. It's like next year. Dario Amadei, who is the Anthropic CEO, thinks 2027 to 2029, so he's like a year off. And Sam Altman is probably, in that sense, more conservative, he thinks around 2030. So that gives a little longer survival rate for humanity. And Dennis Hassabis from DeepMind Google, he's talking about 2030 to 2035, so five to 10 years out. So again, and there's other more aggressive ones we're talking about a few months. So, well, 2026 is less than a few months. It's a few days away. Yeah, and you know, recently, the Trump administration also is kind of pushing to lower regulations in the AI race. So this, coupled in with what we've heard about OpenAI, NVIDIA, making investments and trying to pull in more money to build infrastructure, this seems like it could be heading towards this direction. Well, there's a reason for that, David, and I think the reason is, again, from a military strategy perspective, whoever gets there first, and I'll read it to you what Grok says again, any country or group that controls the ASI instantly becomes the permanent sole superpower. Everyone else surrenders or is neutralized in hours. So when you have that kind of motivation, to me, it's no coincidence that these companies are working so hard, and the administration is working so hard to enable this. And I'm sure China is doing the same. We just don't know about it. And it explains why people like Mark Zuckerberg would spend $100 million salary just for one engineer, to the ones who are able to develop this, because whoever gets it within hours will be the sole. It's pretty mind-blowing. Like you said, if the U.S. is trying to get there before China, that it's like a national security and military advantage. And it explains further why that would be the case. You can win any war before it starts, because whatever this ASI can do is design unhackable encryption, unstoppable hypersonic weapons, bioweapons that target only one person on Earth, perfect cyber offense and cyber defense, because everything is now so much controlled by computers. Once you have control of that, again, you can pretty much do everything if you take it over. So hopefully this is not gonna happen for a while, or it's gonna be positive, but it doesn't look good. Well, we'll just have to see what happens, maybe even next year, 2026, next year. So we'll see if we'll have more interesting episodes. We'll do a follow-up on this topic once we know more information. For audience, please tell us about your thoughts about ASI or artificial superintelligence. Where do you think it's taking us? How far it is from the current moment? Yeah, let us know in the comments. And I'm David Lam, and this is Steve Vispas. We'll see you next time on Silicon Valley Front Row.