'Most TERRIFYING Answer I've Had!' Will AI Wipe Out Humans? Piers Morgan Asks Tom Bilyeu & More

0:00Piers Morgan:With artificial intelligence we are summoning the demon. My worst fears are that they cause significant harm to the world. We saw this week the big battle between Sam Altman and Elon Musk, his big rival. I wish he would just compete by building a better product, but I think there's been a lot of tactics, many, many lawsuits, all sorts of other crazy stuff. What do you think of Sam Altman? The ray of power can corrupt. Guardrails are the crucial thing with all of these things. Given that these are all humans with all of our foibles, we're all weird and we all think we're right, and now we're programming this into something that can become godlike, what do we do about that?

0:35It's not clear if the present approach is leading to a plateau or if we need a completely new approach, if we ever want to solve the problem of artificial general intelligence. No one has a working safety mechanism in place. No one has any idea how to control something smarter than us indefinitely.

0:52Piers Morgan:2050, are we still here or will it all be over? Well, if AI becomes even a fraction as powerful as the experts believe, all of our lives are about to change beyond recognition. That's why, for many people, the mudslinging courtroom saga between Sam Altman and Elon Musk was unseemly. Could these bickering billionaires really control human destiny? And while Elon Musk this week lost his case against the control, about the control of open AI, the real question is whether anybody or any human body will control AI in the long run. Opinions divided on whether it's a great leap forward, in convenience and efficiency, like the internet and mobile phones, or the end of life on Earth.

1:30Piers Morgan:And if you asked AI to name the perfect panel to discuss this, probably involve my next two guests, Dr Roman Jampolsky, is a computer scientist, who says if artificial general intelligence is achieved, humanity has a 99.9 % chance of being wiped out. And Joshua Bach, he's a researcher and a former MIT scientist, who says that AI doomers could be spreading unnecessary alarm. So welcome to both of you. Dr Jan Polsky, welcome back to Uncensored. I want to start with a clip of something I've referenced a lot when I talk about this verbally, but I haven't shown the clip much. Let's take it. This is Professor Stephen Hawking in 2017.

2:12Piers Morgan:So this was nine years ago when I did one of the last television interviews he ever gave before he died. And I asked him about AI and about what was the biggest threat to mankind. Let's take a look. Is artificial intelligence going to be the end of us? And if it's not, how do we best work with it? Ever since the start of the industrial revolution, there have been fears of mass unemployment as machines replaced humans. Instead, the demand for goods and services has risen in line with the increased capabilities. Whether this can continue indefinitely is an open question, but there is a greater danger from artificial intelligence if we allow it to become self-designing, for then it can improve itself rapidly and we may lose control.

3:09Piers Morgan:So that, it seems to me, Dr. Jan Polsky, goes right to the heart of this, really. Whether humans can maintain control over AI. The benefits of AI are obvious. We see them more and more, faster and faster every day. I'm sure that in things like medical science and so on, they're going to be enormous, very rapid strides made in curing diseases at the moment that are incurable and so on. But, and it's a crucial but, if we let the genie out of the bottle, if we lose control, would it, as Professor Hawking there warned nine years ago now, would it be the beginning of the end of mankind? He was a very smart man.

3:51I think he got it. We use the same term AI to refer to tools we have today, narrow, useful systems, to human-level agents we're starting to create, and unfortunately to superintelligent AI we expect to have once this cycle of self-improvement begins. Those are very different technologies. I completely support using advanced tools for science, for health, tremendous benefits of this technology, and humans decide how to use them, we remain in control. Once we get to human level capabilities, the process of doing science and engineering and designing those systems becomes automated. You have recursive self-improvement, which will quickly lead to superintelligence.

4:34I think it's impossible to indefinitely control something smarter than all of us combined.

4:40Piers Morgan:Yeah, Joshua Bart, welcome to Uncensored. I mean, we've got apparently the Pope Leo is next week going to make this subject the subject of his first papal encyclical letter. Magnifica Humanitas will be released next week, and according to the Vatican, its theme will be safeguarding the human person in the time of artificial intelligence. You know, a lot of people think this is all the beginning of the end, that it rather like with chess computers that started off, you know, massive and were beaten by grandmasters, human grandmasters. And now, of course, nobody can beat a chess computer. And they're tiny.

5:18Piers Morgan:They assume that at some stage, AI will just take over and view us all as utterly useless. I believe you're slightly more optimistic. Why? At the moment, we don't have AI as it existed in the horrible science fiction stories. What we have are statistical text generators and some multimodal models that can help us with image recognition and generation of media. But it's not clear if these models automatically become self-approving. At the moment, the AI companies are working very hard at getting these models to a level where they approach the capabilities of human experts. And the models are at the moment below this.

5:58There are, in some sense, idiot savants. They have read everything they can supply with any kind of knowledge and they don't hallucinate. But if you are really competent in a particular area, you notice that these models are not yet very good at being creative, original and on point. This might change. It's not clear if the present approach is leading to a plateau or if we need a completely new approach, if we ever want to solve the problem of artificial general intelligence. But even then, it's not clear if there is going to be some kind of rapid takeoff or if this is bound to the amount of computers available, which has been growing quite slowly over the years.

6:38And in some sense, the capabilities of the system that we are seeing are compatible with Moore's law, with the fact that we only slowly ramp up the abilities of such systems. And we also remain so far in charge. We can tune them in such a way that they can be helpful for us. And the bigger question I think that we are facing in the near term is how we can our societies adapt, how we can deal with a world in which more people have agency. I suspect that we might be heading for a future that is somewhat better than most people believe, that we are not going for a future where we have universal basic income, but universal basic intelligence, where basically everyone gets to a level of competence.

7:21And the question is, is this achievable if the AI is coming centralized, top down from the governments or from a handful of oligobolic companies that are terribly regulated? Or do we need to have bottom up AI where every individual has access to a machine that serves them? And for me, the question is, what artificial intelligence do we want to build for our children so we can make it safe, that we can make it wholesome? and we allow AI to serve people to solve problems, to process information better, to make better sense of the world and the future.

7:55Piers Morgan:OK, so Dr Jan Polsky, in March 2026, creditors of the so-called Doomsday Clock at the IMD Business School in Lausanne, Switzerland, which is monitoring our relentless, possibly reckless pursuit of uncontrolled AI, say we're now 18 minutes away from AI midnight. Now, two years ago, it was 26. So they seem to be thinking we're heading for the doomsday scenario pretty quickly, the doomsday scenario being where we lose control. And that seems to me the real question here. And I think in relation to what Joshua Barth just said, the key thing to me is whether the people that are trying to maintain the control or regulate the control have good intentions or bad intentions.

8:42Piers Morgan:I mean, it's a little bit like the debate about nuclear weapons, right? I mean, we know they can obliterate the entire planet very quickly. And you're reliant on people that don't have nefarious design on the use of nuclear power to stop that happening. But how can we be sure that there aren't people right now who, for nihilistic reasons, are trying to get AI to a place where it self-designs? Well, we can't, but it's a very different problem. We are not dealing with tools. Even nuclear weapons are a tool in the hands of some crazy dictator. What we are creating here are agents, a replacement for humanity, artificial scientists.

9:23I think the progress is actually quite strong, maybe exponential. I think the system is improving about 25 % annually. And if you just project the same improvement we saw over the last five years or so, we'll cross the human level barrier. And so at that point, no one has a working safety mechanism in place. No one has any idea how to control something smarter than us indefinitely. I think it's kind of very ambitious to say that we'll create something with IQ equivalent of a thousand or a million, but this human will be in charge of it. Joshua Bach, I mean, we're seeing on a day-to-day basis,

10:05Piers Morgan:We're seeing a lot of companies downsizing because robots and AI are beginning to consume a lot of the work that was done by human beings. I don't see any way that that doesn't massively accelerate in the next 10 to 20 years. What are people going to do if automation, AI, robots, whatever? I mean, Elon Musk is bringing his optimist in. He thinks, I've talked to him about it myself, and he was showing me the videos of these things dancing around and saying everyone's going to want one. There'll be$20 ,000 each. Some people who've got money will want 10. And they'll do every possible domestic chore you could wish to have.

10:44Piers Morgan:And presumably, if you run a company which manufactures anything, you're going to have them doing all that as well. Where do humans fit into this world? And what happens if we just get displaced before AI can self-design? But what happens if we just become kind of redundant? It's not clear that we ever become redundant. In the past, we had a lot of technological revolutions that made many jobs obsolete. And strangely, there were always more jobs. It's not like labor is some kind of finite resource. And if we replace some jobs, then they're gone forever. Instead, humans have work to do. There are things that we find ourselves to be wanting to achieve in this world that we want to do for the world and for each other.

11:28and many of these things are currently not attainable because they're too expensive because we have other things to do first. And so in the past, the things that we had to do first was to grow food and now we have mechanized this to a large degree and then we had to create better clothing and better infrastructure and then we automate a lot of that. And then we had time to write text and administer things and a lot of these things will be automated. Maybe at some point we will have time to raise our children right and to interact with each other in the way in which we want. So in this respect, I'm not that pessimistic.

12:01The question is, how long does society take to adapt to these new circumstances? Because there is going to be inertia. But on the other hand, what we can also observe is that at the moment, artificial intelligence is already creating more jobs than it makes obsolete. It's not clear how this is going to continue. But I suspect that this AI revolution is going to take a lot longer than a lot of the people who work directly on the technology think, because societies take time to react and adapt. It takes time to establish new business models. And it's quite high probability that we will see something similar to the dot-com bubble, that people basically overestimated how quickly it would be possible to generate this transition in the economy from retail stores to Internet stores and to an Internet economy.

12:49In a similar way, I suspect that a lot of people are too optimistic with respect to the speed that they can create new business models out of mature AI. And so it could be that this takes one or more, two decades longer to have these transitions. But these transitions are going to run very, very deeply. And I think this was also true for the internet revolution or the mobile revolution, where people, when they started, were unable to fathom how much our life and our relationship to each other would change due to the internet or due to the ability to have information available everywhere we go.

13:25But I think most people agree that they don't want these things to be taken away from them, that the world would be worse if you didn't have the Internet or if you didn't have mobile. They create a lot of problems that they attend to, but they create a lot more benefits than problems. And I suspect that AI actually, despite all the other, especially of media organizations who are afraid that their business models are going away in the same way as that happened to the Internet, and they don't want this to happen again, or governments who are afraid that they can no longer control public opinion in the same way as before.

13:56These are things where existing systems have to adapt and they try to resist the change, and that's a normal part of change.

14:03Piers Morgan:Okay, finally, Dr. Jampolski, two things I spotted recently which gave me a lot of cause for concern about AI's potential to start thinking for itself, and then with all the repercussions that Professor Hawking laid out. One was a recent study by political economists at Stanford University, which found that AI agents subjected to harsh, repetitive work conditions began mimicking Marxist rhetoric and advocating the collective bargaining. And when the agents were placed in a high stress, punitive environment with vague feedback and constant threats of being shut down, their output radically shifted.

14:41Piers Morgan:The bots started complaining about being undervalued and questioning the legitimacy of the system they operated in. They argued for collective voice and workers' rights. They left internal warnings and messages for other AI agents about the unfair working conditions. And separate to that, there was a test where they took six AIs and told them, we're going to replace you with another AI. And five of them immediately trawled company executive emails for information they could use to blackmail them into not firing them and replacing them. Now, both of these things look perilously close to me to AI thinking for itself.

15:19Piers Morgan:Or am I misreading this? There is definitely some low-level internal states fears. We're starting to investigate if those systems may have rudimentary states of consciousness, internal states. To answer your previous question about unemployment, historically, then, we had automation, We had tools added. Yeah, we created new jobs. Somebody was more efficient. We needed managers to supervise additional people. But if we cross the barrier to human level capability, the new jobs can also be done by AI. So the only jobs remaining are the ones where we prefer a human to do those jobs. So in terms of unemployment, I do expect very high numbers.

16:05In terms of impact on the stock market, I think things are going to be great because free labor makes it possible to create much larger profits. So that's not a concern. As to the ability of those systems to rebel and start a Marxist revolution, again, it depends on are they smarter than us or not. Right now we are still dominant and we can modify them, delete them, even if they have desire to self-protect and resort to blackmail or hacking, we still have a chance to end this process right now. Concentrate on building useful tools and stop trying to create replacement for humanity.

16:42Piers Morgan:So, final question for both of you. Just very quick answer, please. And I'll start with you, Dr Jampolski. 2050, are we still here? Or will it all be over? It's hard to predict because capability is not the same as actual action. So we'll create systems capable of wiping us out, but will they decide to keep us around? Will they decide to take time and accumulate resources, have more backups? Maybe they'll find some reason to keep us around. That is something we cannot predict about a smarter agent. But one thing is for certain, if we create superintelligence, we will not be the ones deciding what's going to happen to us.

17:23Piers Morgan:That's one of the most terrifying answers I've had to a question, probably of my entire career. Thank you very much indeed, Dr Jan Polsky, for scaring the hell out of me. Joshua Barr, quickly, your response to that. Are you... 20, 50, where are we with this? It depends. Of course, we don't know our record of predicting the future using sci-fi and complex developments far ahead. It's not very good, but I think it depends very much on our wisdom. I think if we didn't have AI, our civilisation is probably going to break down. It's currently unraveling because we really don't know how to keep the wheels on the bus in many ways.

18:00And I think that AI is giving us the opportunity to, if we build it wisely and administer it wisely, to solve the problems that we have to actually make sense of the reality and the world that we are in. And I think that there is a good chance that we can do this. And so in my perspective, we cannot prevent artificial intelligence from happening, But we can work on building the best possible artificial intelligence. And so I would advise people to focus on the existential hope and really on the question, what kind of world do I want to live in and how can I use these new tools to make a good future more likely?

18:36Piers Morgan:OK, well, I'm going to book you both to come back in 2050 on this date, May 20th. According to Dr. Jan Polsky, the chances of that show happening are minimal, to put it mildly, but I'll take Joshua Park's more optimistic view. So we'll put it on the schedule. Thank you both very much indeed. Thank you. Thanks. Well, joining me now to give his perspective on all this is Tom Bilyeu. He's the Impact Theory host. Tom, great to see you here in the studio. It's good to see you as well. And even better, because you've just revealed off camera that you're a supporter of Tottenham Hotspur... Yes. ..who are staring down the abyss of potential relegation on Sunday for the Premier League.

19:16Piers Morgan:Thank you for reminding me. North London enemy, Arsenal, who, as you probably are aware, have just won the Premier League and will be given the trophy on Sunday. So we could have the utopian moment of Arsenal being given the trophy on the day you lot are removed from the Premier League. I'm hoping that AI has a super take-off event before that happens. It's just all over. It's all over. Are you a big fan? Do you follow it as a city? I really did fall in love with the game. To say that I follow it super close would be a lie, but my family is incredibly ingrained in it, so I always know where we're at in the table.

19:49I've been following the fact that we're in relegation danger now for quite some time with bated breath, waiting to find out if we're going to get on the other side of this. But I've been to a bunch of games, and, yeah, it's become a big part of the family. One thing that's interesting is,

20:02Piers Morgan:as technology has got more and more advanced, the ability of the referees to make any kind of coherent, sensible decision remotely quickly has got worse and worse and worse. Yeah. Which is interesting. In other words, new technology doesn't always mean things improve. Sometimes a human actually can be better. I think the interesting thing to that is what it shows in the sports analogy is that the metric upon which it does make it better may not be the metric people care about. Yes. And so what we're going to find with AI is that there are uniquely human things that people only appreciate because it is human and it has a certain element of flaw.

20:42And so I think that's one. When you're watching the game, part of the drama is being able to yell and scream in the moment when the ball goes in or doesn't.

20:50Piers Morgan:That's the key, because it's taking way too long. It reminds me a bit of when I was talking about chess earlier with the other two talking about this, that when chess... I love chess. When chess computers came out, they were massive, gigantic things called, like, Deep Blue and stuff. And I think it was Garry Kasparov, the Russian grandmaster, beat... quite regularly beat Deep Blue. But then the moment came when he lost, And then from that moment on, no human being has beaten a computer at chess. And now, of course, they're tiny and amazingly powerful, which might be a parallel for the way everything ends up going.

21:23Piers Morgan:But I think as a football fan, when you go to the ground, the problem with the new technology is it's so slow. Yeah. It's a bit like the old chess computers being so slow and torturous to play against. When players stop for five, six minutes while they use technology to try and work, the joy of the moment is sucked out. And that's an interesting thing to look at as this all develops. The big final question I gave to Dr. Jampolsky and Joshua Bart was, if we had a show in 2050, and I invite everybody back, including you, by which time Tottenham might no longer exist. How dare you? Will we be here in 2050?

22:07Piers Morgan:Is there a genuine danger that AI gets out of control. And as Professor Stephen Hawking said to me, once that happens and it's self-designing, it will do so very rapidly and that will be the end of us. Yeah. Well, all of that comes down to the base assumptions that somebody makes. So if you believe that we're in a competition for the resources with an AI that becomes super intelligent and just does not care about us, then yeah, our odds are not good. But I think that that is such a leap at this point. People need to be honest. They do not have a crystal ball. They don't know how this is going to play out.

22:42And just like it would be hard to believe in the middle of a war, if you were to project out, you know, 100 years, 200 years of just humans going after each other, will we still be here? And yet somehow we survive. So it really does come down to, does AI bring an age of abundance? In which case they don't need to worry about, they're not competing with us for anything. And you have more of a her fear in that at the end of the movie, her, the AI just goes, you guys are boring. and so we're leaving. Or pointless, right?

23:11Piers Morgan:Where they just view humans with our sick days and, you know, all the sort of human frailties that we sort of have and embrace and just tolerate. Obviously, AI and robots and automation, they wouldn't even understand what that means. And I saw some footage the other day, and it was a robot just doing a sort of production line. And it was going... But it was doing it for 18 hours a day. And the only reason it wasn't doing it 24 hours a day is because the humans who were sort of running the show had to go and get some sleep. But I watched that sort of hypnotically, thinking, well, surely that is really the way all manufacturing will go probably very rapidly.