OpenAI готується до AGI: план порятунку економіки чи самопіар?

So, OpenAI just released this paper that talks about what is going to happen after AGI, and I think we need to talk about this. So, a few days ago, OpenAI published this paper called Industrial Policy for the Intelligence Age, and I think this is bigger than most people realize because the conversation around AI and AGI is starting to get serious, but in a way that most people, including some of you guys watching this video, don't understand, and let me explain. So, when you think about this, okay, most people may ask, "Why is OpenAI talking about, you know, after AGI? Why are they publishing papers that are essentially saying that, you know, this technology could cause mass unemployment, it could enable bioweapon attacks, it could produce systems that, and this is a direct quote, by the way, cannot easily be recalled, and that essentially means that we might not be able to get the model back after it escapes the system." So, what does OpenAI know, and how does that formulate their decision to make this? You see, what most people don't realize is that OpenAI believes that superintelligence is no longer a theoretical milestone or that far off. They believe that it is close enough that the policies and institutions and safety nets that we currently have are not going to survive contact with it, and they'd rather be the ones proposing the fix than being the ones regulated after the damage is already done. And before I dive into OpenAI's exact paper, I do want to talk about all of the issues surrounding AI because once you understand all of the issues that are mounting, then you can understand why OpenAI is trying to create all of these solutions. So, let's quickly dive into some of the problems before we can talk about the solutions embedded in the paper. So, for most of you guys, if you don't understand how crazy this is, OpenAI is aiming to launch research interns in 2026. Now, that is not far away because, of course, we are already in 2026, and the thing is is that these things will, of course, speed up AI research. These are going to be systems that can read scientific papers, compare them with existing work, identify gaps, and then suggest the next steps. And by March 2028, so that's less than 2 years from now, the goal is a fully autonomous AI researcher, a system that can design experiments, run them, and present findings without human involvement. And you have to understand just how crazy that is going to be. Once you have an automated AI researcher, that is going to speed up the AI development even quicker than it is now. And remember, we are already moving exceptionally fast. Now, what's crazy about this is that most of you guys haven't already seen the interview where Sam Altman directly talks about this. You see, Sam Altman himself said it plainly during an interview with Axios. He said, "We're close enough to AGI that the precise definition matters." [music] Some people at OpenAI actually believe that they're already there. >> [music] >> And when he asked, "What comes after AGI, you know, superintelligence, systems that outperform everyone and everything?" Altman didn't wave it away. He said it's coming so fast enough that we need to think about this now. Of course, I'm including clip all two from that actual interview, but it is really, really important for people to have this conversation. [music] And most of you guys understand that this is not the only guy having the conversation. Dario Amodei, CEO of Anthropic, told the World Economic Forum that AGI is likely [music] to arrive in 2027. Demis Hassabis gives it 50/50 odds by the end of the decade, and these aren't fringe predictions, guys. These are coming from the people with the clearest view of the actual capabilities. The clearest capabilities in terms of where you can see that, that is going to be Demis Hassabis, Dario Amodei, Sam Altman. These are the guys that are very, very embedded into AI. I would argue that nobody else is probably more embedded than them. Now, let's get onto the conversation that most people don't have. Problem is that this AGI/ASI system within the next 2 years, 2 to 3 years, there are estimates that this is going to affect 300 million full-time jobs, okay? Goldman Sachs is estimating that generative AI is going to affect 300 million full-time jobs. And McKinsey's research is saying that 57% of current US work involves tasks that today's technology, not future models, today's could theoretically automate. And the World Economic Forum projects that 92 million roles will be displaced by 2030 against 170 million new roles created. Now, whilst, yes, that is net positive on paper, the people that are going to be losing their jobs and the people that are going to be filing for new ones aren't going to be the same people in the same places with the same skills. And so, let's take a look at another issue here, and this is the fact that Goldman Sachs data shows that 22 to 25-year-olds in AI exposed roles, their employment fell 16% from 2022 to mid-2025. That is something that is already happening among young software developers. Specifically, the decline was nearly 20%. Entry-level job postings have dropped 35% since January 2023, and those numbers only account for today's models. And OpenAI's own roadmap says the next leap is their autonomous AI researchers, and that's, of course, 2 years away. And when that arrives, the displacement curve doesn't continue, it just steepens. And that is the backdrop for all of what I'm about to discuss. That is why OpenAI decided to publish this, you know, document telling the government, "Look, this is how we need to structure the economy right now. We need to restructure the entire thing because what is coming it just literally doesn't work. Like, if the economy basically collapses." So, essentially, what we have here is something called Industrial Policy for the Intelligence Age, Ideas to Keep People First. And remember that, keeping people first, because by the time I get to the end of this, you'll realize what happens if we don't keep people first. One of the first things it talks about is a public wealth fund, and OpenAI is proposing that the United States government create a nationally managed investment fund. And that investment fund is going to be seeded by AI companies themselves. So, that means that the AI companies are going to be funding that, and essentially, it gives every American citizen a direct financial stake in AI-driven economic growth. So, it's not just shareholders, not just stockholders, it's everyone. The fund would invest in a diversified mix of AI companies and businesses adopting AI, then distribute those returns directly back to citizens. Think of it like Alaska's Permanent Fund, which has paid every Alaskan resident an annual dividend from oil revenues since 1982, but scaled nationally and funded by intelligence instead of petroleum. So, this is essentially a UBI, but in a different way, funded by the AI companies that have actually displaced all the workers. And I think this, in theory, could work really well. And so, next, what they talk about is robot taxes. So, the paper acknowledges something that most politicians are still dancing around. As AI automates more and more work, the tax base is going to crumble. Social Security, Medicaid, SNAP, housing assistance all funded primarily through payroll taxes. And if machines are doing the work, [music] and, of course, the humans aren't earning wages, that revenue is going to disappear. And that means that OpenAI has an answer. We have to shift that tax burden from payroll to capital gains, corporate income, and taxes specifically linked to automated labor. So, maybe, in the future, you might be thinking, "Okay, robots are going to take everything." Maybe, just maybe, governments are going to implement some kind of automated labor tax, and that may make it competitive for humans to still thrive in certain industries. I think that could be a positive solution, especially in industries where humans are actually overall more valuable, but it will be very interesting to see how that comes into fruition. Next, what we have here is the four-day workweek. So, OpenAI proposes incentivizing employers to run 32-hour workweek pilots at full pay, holding consistent output. And that if AI makes you more productive, that efficiency shouldn't just flow to its shareholders, it should buy workers [music] their time back. And they call this an efficiency dividend. So, you know, you've got a 32-hour week, you still get your full pay, and essentially, that's one way for you to share in your profits. This is an idea that's been afloat around, you know, quite some time, but it will be interesting to see if this is put into practice. I mean, I think this is a great idea. I think people should have more time, especially if AI is going to be automating large parts of the economy. And the paper also triggers, you know, it talks about [music] in safety nets and stuff like that. So, it says, "The government would track real-time metrics on AI displacement, unemployment rates, regional job losses, industry-specific [music] disruption. And when those metrics cross preset thresholds, expanded benefits, such as cash assistance, wage insurance, training vouchers, those kick in automatically. There's no waiting for Congress to debate emergency legislation. There's no 6-month lag while people [music] lose their homes. The benefits scale with the disruption, and then phase out once conditions stabilize. You basically got these training wheels that are implemented immediately once AI becomes capable enough. And I think this is really, really important. Something I'm going to talk about later on in the video is that currently, people are starting to realize exactly how [music] much AI will disrupt the future, and a lot of people are starting to grow tired with the fact that AI is going to displace them. So, if you don't have a safety net for all of these people, society, essentially, is going to collapse in many different ways, and anger and [music] violence is probably going to be one of those, okay? So, this AGI labor safety net is really, really important for keeping the fabric of society together. So, things like expanded benefits, you know, wage insurance, vouchers, cash assistance. However it is done, this will be needed to do because, trust me, when the unemployment jumps after a lot of people get fired, this is going to be something that is necessary. And then, of course, you know, let's say when I'm talking just about the economy, and this is the one that is a little bit unsettling, but people, for some reason, don't seem to take this one seriously. This is the Model Containment Playbooks, and this explicitly acknowledges scenarios where dangerous AI systems become autonomous and capable of self-replication, and cannot be easily recalled. And they propose a response where there's coordinated government involvement, drawing from cybersecurity incident response and public health containment. Essentially, emergency protocols for AI that has escaped control. And that last part is worth sitting with. This is OpenAI, okay, formally admitting that potentially, in a few years, they might release an AI system that is going to become uncontrollable, and that is going to be pretty difficult to maintain. So now, the only problem with all of this is that the same day that this article was released, The New Yorker dropped the results of an 18-month investigation by Ronan Farrow. There were over 100 interviews, never-before-disclosed internal memos, including roughly 70 pages from co-founder Ilya Sutskever, which alleges Sam Altman has a consistent pattern of lying. The long story short here is that there is a huge problem with Sam Altman, according to this article, and it talks about the fact that Sam Altman may not be trustworthy technology is as capable as he says it is. I mean, think about it. This is world-defining technology that is going to change our society forever. If the person at the helm of this technology has so many people saying shady things about them, should we not more be concerned about who is in control of OpenAI? I mean, of course, all of this stuff is alleged. I mean, of course, you can read the article for yourselves, but it doesn't exactly bode the best kind of, you know, thing for the future when those in control of AI might not be the most trustworthy. I mean, are they going to tell us stuff or picture if AI is going rogue, if AI is doing something bad, if the economy is about to go down? Super, super interesting. And remember, the safety track record, if we look back in it, you know, OpenAI in 2023, they announced a super alignment team dedicated to controlling systems smarter than humans and pledged to allocate 20% of the company's computing power to the effort. By 2024, that team was dissolved, and people who worked on it say they received 1 to 2% of computing resources on the oldest hardware. And I mean, it's pretty crazy, okay? Like, you know, this entire thing is supposed to be protecting AI, but I've seen OpenAI constantly jump back and forth between all of these different metrics, so it becomes really hard to put faith in this company when it comes to putting the safety of these AI systems in their control. But there is a counterpoint, and it's the fact that apparently this document is very, very important in the sense that it is one of the first documents that are actually not just a floaty-moty document, and this is a very high-level document that really, you know, formulates what is needed in an AGI economy. And the reason I've actually, you know, got this up here is because I believe that this is something that needs to be done. I'm probably going to have another video on the AI backlash because that thing is going absolutely crazy. There have been continued events that just show that people are frustrated with AI, and people really do hate AI. And I think you have to understand that this window that we're in where technology is accelerating faster than anyone can keep up with, we have to understand that this is going to, you know, really split society. And so, in this video, I talk about the growing AI backlash and how people hate AI, how people have disdain for AI, and how if you're someone watching this video, you probably have a more positive outlook than I would say maybe 80% of current society. I don't think you understand how negatively public perception of AI is. And top that on with the massive mounting job losses that are potentially going to come, we have a completely destabilized society that is going to hit like a truck within the next 2 to 3 years, and I don't have the solution, but hopefully, what OpenAI's document has proposed here gets pushed into legislation and acted upon sooner rather than later because, honestly, the timeline is closing, okay? The timeline is closing, and we need to act fast. AI is not waiting for anyone. You've seen just how quickly this space has been, and if you've been a subscriber of my channel for a long while, you've seen just how quickly these models have improved. So, if there's one thing we can bet on is we can bet on that this is the worst it'll ever be, and if that's the case, I think it's best we act now rather than later. Let me know your thoughts in the comments section below. You've been watching the AI Grid, and I'll see you guys in the next one.