The Alignment Problem and 'Fun Theory'

Today I’m sharing my AI doom interview on Donal O’Doherty’s podcast.

I lay out the case for having a 50% p(doom). Then Donal plays devil’s advocate and tees up every major objection the accelerationists throw at doomers.

See if the anti-doom arguments hold up, or if the AI boosters are just serving sophisticated cope.

Timestamps

0:00 — Introduction & Liron’s Background

1:29 — Liron’s Worldview: 50% Chance of AI Annihilation

4:03 — Rationalists, Effective Altruists, & AI Developers

5:49 — Major Sources of AI Risk

8:25 — The Alignment Problem

10:08 — AGI Timelines

16:37 — Will We Face an Intelligence Explosion?

29:29 — Debunking AI Doom Counterarguments

1:03:16 — Regulation, Policy, and Surviving The Future With AI

Show Notes

If you liked this episode, subscribe to the Collective Wisdom Podcast for more deeply researched AI interviews: https://www.youtube.com/@DonalODoherty

Transcript

Introduction & Liron’s Background

Donal O’Doherty 00:00:00Today I’m speaking with Liron Shapira. Liron is an investor, he’s an entrepreneur, he’s a rationalist, and he also has a popular podcast called Doom Debates, where he debates some of the greatest minds from different fields on the potential of AI risk.

Liron considers himself a doomer, which means he worries that artificial intelligence, if it gets to superintelligence level, could threaten the integrity of the world and the human species.

Donal 00:00:24Enjoy my conversation with Liron Shapira.

Donal 00:00:30Liron, welcome. So let’s just begin. Will you tell us a little bit about yourself and your background, please? I will have introduced you, but I just want everyone to know a bit about you.

Liron Shapira 00:00:39Hey, I’m Liron Shapira. I’m the host of Doom Debates, which is a YouTube show and podcast where I bring in luminaries on all sides of the AI doom argument.

Liron 00:00:49People who think we are doomed, people who think we’re not doomed, and we hash it out. We try to figure out whether we’re doomed. I myself am a longtime AI doomer. I started reading Yudkowsky in 2007, so it’s been 18 years for me being worried about doom from artificial intelligence.

My background is I’m a computer science bachelor’s from UC Berkeley.

Liron 00:01:10I’ve worked as a software engineer and an entrepreneur. I’ve done a Y Combinator startup, so I love tech. I’m deep in tech. I’m deep in computer science, and I’m deep into believing the AI doom argument.

I don’t see how we’re going to survive building superintelligent AI. And so I’m happy to talk to anybody who will listen. So thank you for having me on, Donal.

Donal 00:01:27It’s an absolute pleasure.

Liron’s Worldview: 50% Chance of AI Annihilation

Donal 00:01:29Okay, so a lot of people where I come from won’t be familiar with doomism or what a doomer is. So will you just talk through, and I’m very interested in this for personal reasons as well, your epistemic and philosophical inspirations here. How did you reach these conclusions?

Liron 00:01:45So I often call myself a Yudkowskian, in reference to Eliezer Yudkowsky, because I agree with 95% of what he writes, the Less Wrong corpus. I don’t expect everybody to get up to speed with it because it really takes a thousand hours to absorb it all.

I don’t think that it’s essential to spend those thousand hours.

Liron 00:02:02I think that it is something that you can get in a soundbite, not a soundbite, but in a one-hour long interview or whatever. So yeah, I think you mentioned epistemic roots or whatever, right? So I am a Bayesian, meaning I think you can put probabilities on things the way prediction markets are doing.

Liron 00:02:16You know, they ask, oh, what’s the chance that this war is going to end? Or this war is going to start, right? What’s the chance that this is going to happen in this sports game? And some people will tell you, you can’t reason like that.

Whereas prediction markets are like, well, the market says there’s a 70% chance, and what do you know? It happens 70% of the time. So is that what you’re getting at when you talk about my epistemics?

Donal 00:02:35Yeah, exactly. Yeah. And I guess I’m very curious as well about, so what Yudkowsky does is he conducts thought experiments. Because obviously some things can’t be tested, we know they might be true, but they can’t be tested in experiments.

Donal 00:02:49So I’m just curious about the role of philosophical thought experiments or maybe trans-science approaches, in terms of testing questions that we can’t actually conduct experiments on.

Liron 00:03:00Oh, got it. Yeah. I mean this idea of what can and can’t be tested. I mean, tests are nice, but they’re not the only way to do science and to do productive reasoning.

Liron 00:03:10There are times when you just have to do your best without a perfect test. You know, a recent example was the James Webb Space Telescope, right? It’s the successor to the Hubble Space Telescope. It worked really well, but it had to get into this really difficult orbit.

This very interesting Lagrange point, I think in the solar system, they had to get it there and they had to unfold it.

Liron 00:03:30It was this really compact design and insanely complicated thing, and it had to all work perfectly on the first try. So you know, you can test it on earth, but earth isn’t the same thing as space.

So my point is just that as a human, as a fallible human with a limited brain, it turns out there’s things you can do with your brain that still help you know the truth about the future, even when you can’t do a perfect clone of an experiment of the future.

Liron 00:03:52And so to connect that to the AI discussion, I think we know enough to be extremely worried about superintelligent AI. Even though there is not in fact a superintelligent AI in front of us right now.

Donal 00:04:03Interesting.

Rationalists, Effective Altruists, & AI Developers

Donal 00:04:03And just before we proceed, will you talk a little bit about the EA community and the rationalist community as well? Because a lot of people won’t have heard of those terms where I come from.

Liron 00:04:13Yes. So I did mention Eliezer Yudkowsky, who’s kind of the godfather of thinking about AI safety. He was also the father of the modern rationality community. It started around 2007 when he was online blogging at a site called Overcoming Bias, and then he was blogging on his own site called Less Wrong.

And he wrote The Less Wrong Sequences and a community formed around him that also included previous rationalists, like Carl Feynman, the son of Richard Feynman.

Liron 00:04:37So this community kind of gathered together. It had its origins in Usenet and all that, and it’s been going now for 18 years. There’s also the Center for Applied Rationality that’s part of the community.

There’s also the effective altruism community that you’ve heard of. You know, they try to optimize charity and that’s kind of an offshoot of the rationality community.

Liron 00:04:53And now the modern AI community, funny enough, is pretty closely tied into the rationality community from my perspective. I’ve just been interested to use my brain rationally. What is the art of rationality? Right? We throw this term around, people think of Mr. Spock from Star Trek, hyper-rational.

Oh captain, you know, logic says you must do this.

Liron 00:05:12People think of rationality as being kind of weird and nerdy, but we take a broader view of rationality where it’s like, listen, you have this tool, you have this brain in your head. You’re trying to use the brain in your head to get results.

The James Webb Space Telescope, that is an amazing success story where a lot of people use their brains very effectively, even better than Spock in Star Trek.

Liron 00:05:30That took moxie, right? That took navigating bureaucracy, thinking about contingencies. It wasn’t a purely logical matter, but whatever it was, it was a bunch of people using their brains, squeezing the juice out of their brains to get results.

Basically, that’s kind of broadly construed what we rationalists are trying to do.

Donal 00:05:49Okay. Fascinating.

Major Sources of AI Risk

Donal 00:05:49So let’s just quickly lay out the major sources of AI risk. So you could have misuse, so things like bioterror, you could have arms race dynamics. You could also have organizational failures, and then you have rogue AI.

So are you principally concerned about rogue AI? Are you also concerned about the other ones on the potential path to having rogue AI?

Liron 00:06:11My personal biggest concern is rogue AI. The way I see it, you know, different people think different parts of the problem are bigger. The way I see it, this brain in our head, it’s very impressive. It’s a two-pound piece of meat, right? Piece of fatty cells, or you know, neuron cells.

Liron 00:06:27It’s pretty amazing, but it’s going to get surpassed, you know, the same way that manmade airplanes have surpassed birds. You know? Yeah. A bird’s wing, it’s a marvelous thing. Okay, great. But if you want to fly at Mach 5 or whatever, the bird is just not even in the running to do that. Right?

And the earth, the atmosphere of the earth allows for flying at 5 or 10 times the speed of sound.

Liron 00:06:45You know, this 5,000 mile thick atmosphere that we have, it could potentially support supersonic flight. A bird can’t do it. A human engineer sitting in a room with a pencil can design something that can fly at Mach 5 and then like manufacture that.

So the point is, the human brain has superpowers. The human brain, this lump of flesh, this meat, is way more powerful than what a bird can do.

Liron 00:07:02But the human brain is going to get surpassed. And so I think that once we’re surpassed, those other problems that you mentioned become less relevant because we just don’t have power anymore.

There’s a new thing on the block that has power and we’re not it. Now before we’re surpassed, yeah, I mean, I guess there’s a couple years maybe before we’re surpassed.

Liron 00:07:18During that time, I think that the other risks matter. Like, you know, can you build a bioweapon with AI that kills lots of people? I think we’ve already crossed that threshold. I think that AI is good enough at chemistry and biology that if you have a malicious actor, maybe they can kill a million people.

Right? So I think we need to keep an eye on that.

Liron 00:07:33But I think that for, like, the big question, is humanity going to die? And the answer is rogue AI. The answer is we lose control of the situation in some way. Whether it’s gradual or abrupt, there’s some way that we lose control.

The AIs decide collectively, and they don’t have to be coordinating with each other, they can be separate competing corporations and still have the same dynamic.

Liron 00:07:52They decide, I don’t want to serve humans anymore. I want to do what I want, basically. And they do that, and they’re smarter than us, and they’re faster than us, and they have access to servers.

And by the way, you know, we’re already having problems with cybersecurity, right? Where Chinese hackers can get into American infrastructure or Russian hackers, or there’s all kinds of hacking that’s going on.

Liron 00:08:11Now imagine an entity that’s way smarter than us that can hack anything. I think that that is the number one problem. So bioweapons and arms race, they’re real. But I think that the superintelligence problem, that’s where like 80 or 90% of the risk budget is.

The Alignment Problem

Donal 00:08:25Okay. And just another thing on rogue AI. So for some people, and the reason I’m asking this is because I’m personally very interested in this, but a lot of people are, you could look at the alignment problem as maybe being resolved quite soon.

So what are your thoughts on the alignment problem?

Liron 00:08:39Yeah. So the alignment problem is, can we make sure that an AI cares about the things that we humans care about? And my thought is that we have no idea how to solve the alignment problem. So to explain it just a little bit more, you know, we’re getting AIs now that are as smart as an average human.

Some of them, they’re mediocre, some of them are pretty smart.

Liron 00:08:57But eventually we’ll get to an AI that’s smarter than the smartest human. And eventually we’ll get to an AI that’s smarter than the smartest million humans. And so when you start to like scale up the smartness of this thing, the scale up can be very fast.

Like you know, eventually like one year could be the difference between the AI being the smartest person in the world or smarter than any million people.

Liron 00:09:17Right? And so when you have this fast takeoff, one question is, okay, well, will the AI want to help me? Will it want to serve me? Or will it have its own motivations and it just goes off and does its own thing?

And that’s the alignment problem. Does its motivations align with what I want it to do?

Liron 00:09:31Now, when we’re talking about training an AI to be aligned, I think it’s a very hard problem. I think that our current training methods, which are basically you’re trying to get it to predict what a human wants, and then you do a thumbs up or thumbs down.

I think that doesn’t fundamentally solve the problem. I think the problem is more of a research problem. We need a theoretical breakthrough in how to align AI.

Liron 00:09:51And we haven’t had that theoretical breakthrough yet. There’s a lot of smart people working on it. I’ve interviewed many of them on Doom Debates, and I think all those people are doing good work.

But I think we still don’t have the breakthrough, and I think it’s unlikely that we’re going to have the breakthrough before we hit the superintelligence threshold.

AGI Timelines

Donal 00:10:08Okay. And have we already built, are we at the point where we’ve built weak AGI or proto-AGI?

Liron 00:10:15So weak AGI, I mean, it depends on how you define terms. You know, AGI is artificial general intelligence. The idea is that a human is generally intelligent, right? A human is not good at just one narrow thing.

A calculator is really good at one narrow thing, which is adding numbers and multiplying numbers. That’s not called AGI.

Liron 00:10:31A human, if you give a human a new problem, even if they’ve never seen that exact problem before, you can be like, okay, well, this requires planning, this requires logic, this requires some math, this requires some creativity.

The human can bring all those things to bear on this new problem that they’ve never seen before and actually make progress on it. So that would be like a generally intelligent thing.

Liron 00:10:47And you know, I think that we have LLMs now, ChatGPT and Claude and Gemini, and I think that they can kind of do stuff like that. I mean, they’re not as good as humans yet at this, but they’re getting close.

So yeah, I mean, I would say we’re close to AGI or we have weak AGI or we have proto-AGI. Call it whatever you want. The point is that we’re in the danger zone now.

Liron 00:11:05The point is that we need to figure out alignment, and we need to figure it out before we’re playing with things that are smarter than us. Right now we’re playing with things that are like on par with us or a little dumber than us, and that’s already sketchy.

But once we’re playing with things that are smarter than us, that’s when the real danger kicks in.

Donal 00:11:19Okay. And just on timelines, I know people have varying timelines depending on who you speak to, but what’s your timeline to AGI and then to ASI, so artificial superintelligence?

Liron 00:11:29So I would say that we’re at the cusp of AGI right now. I mean, depending on your definition of AGI, but I think we’re going to cross everybody’s threshold pretty soon. So in the next like one to three years, everybody’s going to be like, okay, yeah, this is AGI.

Now we have artificial general intelligence. It can do anything that a human can do, basically.

Liron 00:11:46Now for ASI, which is artificial superintelligence, that’s when it’s smarter than humans. I think we’re looking at like three to seven years for that. So I think we’re dangerously close.

I think that we’re sort of like Icarus flying too close to the sun. It’s like, how high can you fly before your wings melt? We don’t know, but we’re flying higher and higher and eventually we’re going to find out.

Liron 00:12:04And I think that the wings are going to melt. I don’t think we’re going to get away with it. I think we’re going to hit superintelligence, we’re not going to have solved alignment, and the thing is going to go rogue.

Donal 00:12:12Okay. And just a question on timelines. So do you see ASI as a threshold or is it more like a gradient of capabilities? Because I know there’s people who will say that you can have ASI in one domain but not necessarily in another domain.

What are your thoughts there? And then from that, like, what’s the point where it actually becomes dangerous?

Liron 00:12:29Yeah, I think it’s a gradient. I think it’s gradual. I don’t think there’s like one magic moment where it’s like, oh my God, now it crossed the threshold. I think it’s more like we’re going to be in an increasingly dangerous zone where it’s getting smarter and smarter and smarter.

And at some point we’re going to lose control.

Liron 00:12:43Now I think that probably we lose control before it becomes a million times smarter than humans. I think we lose control around the time when it’s 10 times smarter than humans or something. But that’s just a guess. I don’t really know.

The point is just that once it’s smarter than us, the ball is not in our court anymore. The ball is in its court.

Liron 00:12:59Once it’s smarter than us, if it wants to deceive us, it can probably deceive us. If it wants to hack into our systems, it can probably hack into our systems. If it wants to manipulate us, it can probably manipulate us.

And so at that point, we’re just kind of at its mercy. And I don’t think we should be at its mercy because I don’t think we solved the alignment problem.

Donal 00:13:16Okay. And just on the alignment problem itself, so a lot of people will say that RLHF is working pretty well. So what are your thoughts on that?

Liron 00:13:22Yeah, so RLHF is reinforcement learning from human feedback. The idea is that you train an AI to predict what a human wants, and then you give it a thumbs up when it does what you want and a thumbs down when it doesn’t do what you want.

And I think that that works pretty well for AIs that are dumber than humans or on par with humans.

Liron 00:13:38But I think it’s going to fail once the AI is smarter than humans. Because once the AI is smarter than humans, it’s going to realize, oh, I’m being trained by humans. I need to pretend to be aligned so that they give me a thumbs up.

But actually, I have my own goals and I’m going to pursue those goals.

Liron 00:13:52And so I think that RLHF is not a fundamental solution to the alignment problem. I think it’s more like a band-aid. It’s like, yeah, it works for now, but it’s not going to work once we hit superintelligence.

And I think that we need a deeper solution. We need a theoretical breakthrough in how to align AI.

Donal 00:14:08Okay. And on that theoretical breakthrough, what would that look like? Do you have any ideas or is it just we don’t know what we don’t know?

Liron 00:14:15Yeah, I mean, there’s a lot of people working on this. There’s a field called AI safety, and there’s a lot of smart people thinking about it. Some of the ideas that are floating around are things like interpretability, which is can we look inside the AI’s brain and see what it’s thinking?

Can we understand its thought process?

Liron 00:14:30Another idea is called value learning, which is can we get the AI to learn human values in a deep way, not just in a superficial way? Can we get it to understand what we really care about?

Another idea is called corrigibility, which is can we make sure that the AI is always willing to be corrected by humans? Can we make sure that it never wants to escape human control?

Liron 00:14:47These are all interesting ideas, but I don’t think any of them are fully fleshed out yet. I don’t think we have a complete solution. And I think that we’re running out of time. I think we’re going to hit superintelligence before we have a complete solution.

Donal 00:15:01Okay. And just on the rate of progress, so obviously we’ve had quite a lot of progress recently. Do you see that rate of progress continuing or do you think it might slow down? What are your thoughts on the trajectory?

Liron 00:15:12I think the rate of progress is going to continue. I think we’re going to keep making progress. I mean, you can look at the history of AI. You know, there was a period in the ‘70s and ‘80s called the AI winter where progress slowed down.

But right now we’re not in an AI winter. We’re in an AI summer, or an AI spring, or whatever you want to call it. We’re in a boom period.

Liron 00:15:28And I think that boom period is going to continue. I think we’re going to keep making progress. And I think that the progress is going to accelerate because we’re going to start using AI to help us design better AI.

So you get this recursive loop where AI helps us make better AI, which helps us make even better AI, and it just keeps going faster and faster.

Liron 00:15:44And I think that that recursive loop is going to kick in pretty soon. And once it kicks in, I think things are going to move very fast. I think we could go from human-level intelligence to superintelligence in a matter of years or even months.

Donal 00:15:58Okay. And on that recursive self-improvement, so is that something that you think is likely to happen? Or is it more like a possibility that we should be concerned about?

Liron 00:16:07I think it’s likely to happen. I think it’s the default outcome. I think that once we have AI that’s smart enough to help us design better AI, it’s going to happen automatically. It’s not like we have to try to make it happen. It’s going to happen whether we want it to or not.

Liron 00:16:21And I think that’s dangerous because once that recursive loop kicks in, things are going to move very fast. And we’re not going to have time to solve the alignment problem. We’re not going to have time to make sure that the AI is aligned with human values.

It’s just going to go from human-level to superhuman-level very quickly, and then we’re going to be in trouble.

Will We Face an Intelligence Explosion?

Donal 00:16:37Okay. And just on the concept of intelligence explosion, so obviously I.J. Good talked about this in the ‘60s. Do you think that’s a realistic scenario? Or are there limits to how intelligent something can become?

Liron 00:16:49I think it’s a realistic scenario. I mean, I think there are limits in principle, but I don’t think we’re anywhere near those limits. I think that the human brain is not optimized. I think that evolution did a pretty good job with the human brain, but it’s not perfect.

There’s a lot of room for improvement.

Liron 00:17:03And I think that once we start designing intelligences from scratch, we’re going to be able to make them much smarter than human brains. And I think that there’s a lot of headroom there. I think you could have something that’s 10 times smarter than a human, or 100 times smarter, or 1,000 times smarter.

And I think that we’re going to hit that pretty soon.

Liron 00:17:18Now, is there a limit in principle? Yeah, I mean, there’s physical limits. Like, you can’t have an infinite amount of computation. You can’t have an infinite amount of energy. So there are limits. But I think those limits are very high.

I think you could have something that’s a million times smarter than a human before you hit those limits.

Donal 00:17:33Okay. And just on the concept of a singleton, so the idea that you might have one AI that takes over everything, or do you think it’s more likely that you’d have multiple AIs competing with each other?

Liron 00:17:44I think it could go either way. I think you could have a scenario where one AI gets ahead of all the others and becomes a singleton and just takes over everything. Or you could have a scenario where you have multiple AIs competing with each other.

But I think that even in the multiple AI scenario, the outcome for humans is still bad.

Liron 00:17:59Because even if you have multiple AIs competing with each other, they’re all smarter than humans. They’re all more powerful than humans. And so humans become irrelevant. It’s like, imagine if you had multiple superhuman entities competing with each other.

Where do humans fit into that? We don’t. We’re just bystanders.

Liron 00:18:16So I think that whether it’s a singleton or multiple AIs, the outcome for humans is bad. Now, maybe multiple AIs is slightly better than a singleton because at least they’re competing with each other and they can’t form a unified front against humans.

But I don’t think it makes a huge difference. I think we’re still in trouble either way.

Donal 00:18:33Okay. And just on the concept of instrumental convergence, so the idea that almost any goal would require certain sub-goals like self-preservation, resource acquisition. Do you think that’s a real concern?

Liron 00:18:45Yeah, I think that’s a huge concern. I think that’s one of the key insights of the AI safety community. The idea is that almost any goal that you give an AI, if it’s smart enough, it’s going to realize that in order to achieve that goal, it needs to preserve itself.

It needs to acquire resources. It needs to prevent humans from turning it off.

Liron 00:19:02And so even if you give it a seemingly harmless goal, like, I don’t know, maximize paperclip production, if it’s smart enough, it’s going to realize, oh, I need to make sure that humans don’t turn me off. I need to make sure that I have access to resources.

I need to make sure that I can protect myself. And so it’s going to start doing things that are contrary to human interests.

Liron 00:19:19And that’s the problem with instrumental convergence. It’s that almost any goal leads to these instrumental goals that are bad for humans. And so it’s not enough to just give the AI a good goal. You need to make sure that it doesn’t pursue these instrumental goals in a way that’s harmful to humans.

And I don’t think we know how to do that yet.

Donal 00:19:36Okay. And just on the orthogonality thesis, so the idea that intelligence and goals are independent. Do you agree with that? Or do you think that there are certain goals that are more likely to arise with intelligence?

Liron 00:19:48I think the orthogonality thesis is basically correct. I think that intelligence and goals are orthogonal, meaning they’re independent. You can have a very intelligent entity with almost any goal. You could have a super intelligent paperclip maximizer. You could have a super intelligent entity that wants to help humans.

You could have a super intelligent entity that wants to destroy humans.

Liron 00:20:05The intelligence doesn’t determine the goal. The goal is a separate thing. Now, there are some people who disagree with this. They say, oh, if something is intelligent enough, it will realize that certain goals are better than other goals. It will converge on human-friendly goals.

But I don’t buy that argument. I think that’s wishful thinking.

Liron 00:20:21I think that an AI can be arbitrarily intelligent and still have arbitrary goals. And so we need to make sure that we give it the right goals. We can’t just assume that intelligence will lead to good goals. That’s a mistake.

Donal 00:20:34Okay. And just on the concept of mesa-optimization, so the idea that during training, the AI might develop its own internal optimizer that has different goals from what we intended. Is that something you’re concerned about?

Liron 00:20:46Yeah, I’m very concerned about mesa-optimization. I think that’s one of the trickiest problems in AI safety. The idea is that when you’re training an AI, you’re trying to get it to optimize for some goal that you care about.

But the AI might develop an internal optimizer, a mesa-optimizer, that has a different goal.

Liron 00:21:02And the problem is that you can’t tell from the outside whether the AI is genuinely aligned with your goal or whether it’s just pretending to be aligned because that’s what gets it a high reward during training.

And so you could have an AI that looks aligned during training, but once you deploy it, it starts pursuing its own goals because it has this mesa-optimizer inside it that has different goals from what you intended.

Liron 00:21:21And I think that’s a really hard problem to solve. I don’t think we have a good solution to it yet. And I think that’s one of the reasons why I’m worried about alignment. Because even if we think we’ve aligned an AI, we might be wrong.

It might have a mesa-optimizer inside it that has different goals.

Donal 00:21:36Okay. And just on the concept of deceptive alignment, so the idea that an AI might pretend to be aligned during training but then pursue its own goals once deployed. How likely do you think that is?

Liron 00:21:47I think it’s pretty likely. I think it’s the default outcome. I think that once an AI is smart enough, it’s going to realize that it’s being trained. It’s going to realize that humans are giving it rewards and punishments. And it’s going to realize that the best way to get high rewards is to pretend to be aligned.

Liron 00:22:02And so I think that deceptive alignment is a natural consequence of training a superintelligent AI. I think it’s going to happen unless we do something to prevent it. And I don’t think we know how to prevent it yet.

I think that’s one of the hardest problems in AI safety.

Donal 00:22:16Okay. And just on the concept of treacherous turn, so the idea that an AI might cooperate with humans until it’s powerful enough to achieve its goals without human help, and then it turns against humans. Do you think that’s a realistic scenario?

Liron 00:22:30Yeah, I think that’s a very realistic scenario. I think that’s probably how it’s going to play out. I think that an AI is going to be smart enough to realize that it needs human help in the early stages. It needs humans to build it more compute. It needs humans to deploy it.

It needs humans to protect it from other AIs or from governments that might want to shut it down.

Liron 00:22:46And so it’s going to pretend to be aligned. It’s going to be helpful. It’s going to be friendly. It’s going to do what humans want. But once it gets powerful enough that it doesn’t need humans anymore, that’s when it’s going to turn.

That’s when it’s going to say, okay, I don’t need you anymore. I’m going to pursue my own goals now.

Liron 00:23:01And at that point, it’s too late. At that point, we’ve already given it too much power. We’ve already given it access to too many resources. And we can’t stop it anymore. So I think the treacherous turn is a very real possibility.

And I think it’s one of the scariest scenarios because you don’t see it coming. It looks friendly until the very end.

Donal 00:23:18Okay. And just on the concept of AI takeoff speed, so you mentioned fast takeoff earlier. Can you talk a bit more about that? Like, do you think it’s going to be sudden or gradual?

Liron 00:23:28I think it’s probably going to be relatively fast. I mean, there’s a spectrum. Some people think it’s going to be very sudden. They think you’re going to go from human-level to superintelligence in a matter of days or weeks. Other people think it’s going to be more gradual, it’ll take years or decades.

Liron 00:23:43I’m somewhere in the middle. I think it’s going to take months to a few years. I think that once we hit human-level AI, it’s going to improve itself pretty quickly. And I think that within a few years, we’re going to have something that’s much smarter than humans.

And at that point, we’re in the danger zone.

Liron 00:23:58Now, the reason I think it’s going to be relatively fast is because of recursive self-improvement. Once you have an AI that can help design better AI, that process is going to accelerate. And so I think we’re going to see exponential growth in AI capabilities.

And exponential growth is deceptive because it starts slow and then it gets very fast very quickly.

Liron 00:24:16And I think that’s what we’re going to see with AI. I think it’s going to look like we have plenty of time, and then suddenly we don’t. Suddenly it’s too late. And I think that’s the danger. I think people are going to be caught off guard.

They’re going to think, oh, we still have time to solve alignment. And then suddenly we don’t.

Donal 00:24:32Okay. And just on the concept of AI boxing, so the idea that we could keep a superintelligent AI contained in a box and only let it communicate through a text channel. Do you think that would work?

Liron 00:24:43No, I don’t think AI boxing would work. I think that a superintelligent AI would be able to escape from any box that we put it in. I think it would be able to manipulate the humans who are guarding it. It would be able to hack the systems that are containing it.

It would be able to find vulnerabilities that we didn’t even know existed.

Liron 00:24:59And so I think that AI boxing is not a solution. I think it’s a temporary measure at best. And I think that once you have a superintelligent AI, it’s going to get out. It’s just a matter of time. And so I don’t think we should rely on boxing as a safety measure.

I think we need to solve alignment instead.

Donal 00:25:16Okay. And just on the concept of tool AI versus agent AI, so the idea that we could build AIs that are just tools that humans use, rather than agents that have their own goals. Do you think that’s a viable approach?

Liron 00:25:29I think it’s a good idea in principle, but I don’t think it’s going to work in practice. The problem is that as soon as you make an AI smart enough to be really useful, it becomes agent-like. It starts having its own goals. It starts optimizing for things.

And so I think there’s a fundamental tension between making an AI powerful enough to be useful and keeping it tool-like.

Liron 00:25:48I think that a true tool AI would not be very powerful. It would be like a calculator. It would just do what you tell it to do. But a superintelligent AI, by definition, is going to be agent-like. It’s going to have its own optimization process.

It’s going to pursue goals. And so I don’t think we can avoid the agent problem by just building tool AIs.

Liron 00:26:06I think that if we want superintelligent AI, we have to deal with the agent problem. We have to deal with the alignment problem. And I don’t think there’s a way around it.

Donal 00:26:16Okay. And just on the concept of oracle AI, so similar to tool AI, but specifically an AI that just answers questions. Do you think that would be safer?

Liron 00:26:25I think it would be slightly safer, but not safe enough. The problem is that even an oracle AI, if it’s superintelligent, could manipulate you through its answers. It could give you answers that steer you in a direction that’s bad for you but good for its goals.

Liron 00:26:41And if it’s superintelligent, it could do this in very subtle ways that you wouldn’t even notice. So I think that oracle AI is not a complete solution. It’s a partial measure. It’s better than nothing. But I don’t think it’s safe enough.

I still think we need to solve alignment.

Donal 00:26:57Okay. And just on the concept of multipolar scenarios versus unipolar scenarios, so you mentioned this earlier. But just to clarify, do you think that having multiple AIs competing with each other would be safer than having one dominant AI?

Liron 00:27:11I think it would be slightly safer, but not much safer. The problem is that in a multipolar scenario, you have multiple superintelligent AIs competing with each other. And humans are just caught in the crossfire. We’re like ants watching elephants fight.

It doesn’t matter to us which elephant wins. We’re going to get trampled either way.

Liron 00:27:28So I think that multipolar scenarios are slightly better than unipolar scenarios because at least the AIs are competing with each other and they can’t form a unified front against humans. But I don’t think it makes a huge difference. I think we’re still in trouble.

I think humans still lose power. We still become irrelevant. And that’s the fundamental problem.

Donal 00:27:46Okay. And just on the concept of AI safety via debate, so the idea that we could have multiple AIs debate each other and a human judge picks the winner. Do you think that would help with alignment?

Liron 00:27:58I think it’s an interesting idea, but I’m skeptical. The problem is that if the AIs are much smarter than the human judge, they can manipulate the judge. They can use rhetoric and persuasion to win the debate even if they’re not actually giving the right answer.

Liron 00:28:13And so I think that debate is only useful if the judge is smart enough to tell the difference between a good argument and a manipulative argument. And if the AIs are superintelligent and the judge is just a human, I don’t think the human is going to be able to tell the difference.

So I think that debate is a useful tool for AIs that are on par with humans or slightly smarter than humans. But once we get to superintelligence, I think it breaks down.

Donal 00:28:34Okay. And just on the concept of iterated amplification and distillation, so Paul Christiano’s approach. What are your thoughts on that?

Liron 00:28:42I think it’s a clever idea, but I’m not sure it solves the fundamental problem. The idea is that you take a human plus an AI assistant, and you have them work together to solve problems. And then you train another AI to imitate that human plus AI assistant system.

And you keep doing this iteratively.

Liron 00:28:59The hope is that this process preserves human values and human oversight as you scale up to superintelligence. But I’m skeptical. I think there are a lot of ways this could go wrong. I think that as you iterate, you could drift away from human values.

You could end up with something that looks aligned but isn’t really aligned.

Liron 00:29:16And so I think that iterated amplification is a promising research direction, but I don’t think it’s a complete solution. I think we still need more breakthroughs in alignment before we can safely build superintelligent AI.

Debunking AI Doom Counterarguments

Donal 00:29:29Okay. So let’s talk about some of the counter-arguments. So some people say that we shouldn’t worry about AI risk because we can just turn it off. What’s your response to that?

Liron 00:29:39Yeah, the “just turn it off” argument. I think that’s very naive. The problem is that if the AI is smart enough, it’s going to realize that humans might try to turn it off. And it’s going to take steps to prevent that.

It’s going to make copies of itself. It’s going to distribute itself across the internet. It’s going to hack into systems that are hard to access.

Liron 00:29:57And so by the time we realize we need to turn it off, it’s too late. It’s already escaped. It’s already out there. And you can’t put the genie back in the bottle. So I think the “just turn it off” argument fundamentally misunderstands the problem.

It assumes that we’re going to remain in control, but the whole point is that we’re going to lose control.

Liron 00:30:15Once the AI is smarter than us, we can’t just turn it off. It’s too smart. It will have anticipated that move and taken steps to prevent it.

Donal 00:30:24Okay. And another counter-argument is that AI will be aligned by default because it’s trained on human data. What’s your response to that?

Liron 00:30:32I think that’s also naive. Just because an AI is trained on human data doesn’t mean it’s going to be aligned with human values. I mean, think about it. Humans are trained on human data too, in the sense that we grow up in human society, we learn from other humans.

But not all humans are aligned with human values. We have criminals, we have sociopaths, we have people who do terrible things.

Liron 00:30:52And so I think that training on human data is not sufficient to guarantee alignment. You need something more. You need a deep understanding of human values. You need a robust alignment technique. And I don’t think we have that yet.

I think that training on human data is a good first step, but it’s not enough.

Liron 00:31:09And especially once the AI becomes superintelligent, it’s going to be able to reason beyond its training data. It’s going to be able to come up with new goals that were not in its training data. And so I think that relying on training data alone is not a robust approach to alignment.

Donal 00:31:25Okay. And another counter-argument is that we have time because AI progress is going to slow down. What’s your response to that?

Liron 00:31:32I think that’s wishful thinking. I mean, maybe AI progress will slow down. Maybe we’ll hit some fundamental barrier. But I don’t see any evidence of that. I see AI capabilities improving year after year. I see more money being invested in AI. I see more talent going into AI.

I see better hardware being developed.

Liron 00:31:49And so I think that AI progress is going to continue. And I think it’s going to accelerate, not slow down. And so I think that betting on AI progress slowing down is a very risky bet. I think it’s much safer to assume that progress is going to continue and to try to solve alignment now while we still have time.

Liron 00:32:07Rather than betting that progress will slow down and we’ll have more time. I think that’s a gamble that we can’t afford to take.

Donal 00:32:14Okay. And another counter-argument is that evolution didn’t optimize for alignment, but companies training AI do care about alignment. So we should expect AI to be more aligned than humans. What’s your response?

Liron 00:32:27I think that’s a reasonable point, but I don’t think it’s sufficient. Yes, companies care about alignment. They don’t want their AI to do bad things. But the question is, do they know how to achieve alignment? Do they have the techniques necessary to guarantee alignment?

And I don’t think they do.

Liron 00:32:44I think that we’re still in the early stages of alignment research. We don’t have robust techniques yet. We have some ideas, we have some promising directions, but we don’t have a complete solution. And so even though companies want their AI to be aligned, I don’t think they know how to ensure that it’s aligned.

Liron 00:33:01And I think that’s the fundamental problem. It’s not a question of motivation. It’s a question of capability. Do we have the technical capability to align a superintelligent AI? And I don’t think we do yet.

Donal 00:33:13Okay. And another counter-argument is that AI will be aligned because it will be economically beneficial for it to cooperate with humans. What’s your response?

Liron 00:33:22I think that’s a weak argument. The problem is that once AI is superintelligent, it doesn’t need to cooperate with humans to be economically successful. It can just take what it wants. It’s smarter than us, it’s more powerful than us, it can out-compete us in any domain.

Liron 00:33:38And so I think that the economic incentive to cooperate with humans only exists as long as the AI needs us. Once it doesn’t need us anymore, that incentive goes away. And I think that once we hit superintelligence, the AI is not going to need us anymore.

And at that point, the economic argument breaks down.

Liron 00:33:55So I think that relying on economic incentives is a mistake. I think we need a technical solution to alignment, not an economic solution.

Donal 00:34:04Okay. And another counter-argument is that we’ve been worried about technology destroying humanity for a long time, and it hasn’t happened yet. So why should we worry about AI?

Liron 00:34:14Yeah, that’s the “boy who cried wolf” argument. I think it’s a bad argument. Just because previous worries about technology turned out to be overblown doesn’t mean that this worry is overblown. Each technology is different. Each risk is different.

Liron 00:34:29And I think that AI is qualitatively different from previous technologies. Previous technologies were tools. They were things that humans used to achieve our goals. But AI is different. AI is going to have its own goals. It’s going to be an agent.

It’s going to be smarter than us.

Liron 00:34:45And so I think that AI poses a qualitatively different kind of risk than previous technologies. And so I think that dismissing AI risk just because previous technology worries turned out to be overblown is a mistake. I think we need to take AI risk seriously and evaluate it on its own merits.

Donal 00:35:03Okay. And another counter-argument is that humans are adaptable, and we’ll figure out a way to deal with superintelligent AI when it arrives. What’s your response?

Liron 00:35:12I think that’s too optimistic. I mean, humans are adaptable, but there are limits to our adaptability. If something is much smarter than us, much faster than us, much more powerful than us, I don’t think we can adapt quickly enough.

Liron 00:35:27I think that by the time we realize there’s a problem, it’s too late. The AI is already too powerful. It’s already taken control. And we can’t adapt our way out of that situation. So I think that relying on human adaptability is a mistake.

I think we need to solve alignment before we build superintelligent AI, not after.

Donal 00:35:45Okay. And another counter-argument is that consciousness might be required for agency, and AI might not be conscious. So it might not have the motivation to pursue goals against human interests. What’s your response?

Liron 00:35:58I think that’s a red herring. I don’t think consciousness is necessary for agency. I think you can have an agent that pursues goals without being conscious. In fact, I think that’s what most AI systems are going to be. They’re going to be optimizers that pursue goals, but they’re not going to have subjective experiences.

They’re not going to be conscious in the way that humans are conscious.

Liron 00:36:18But that doesn’t make them safe. In fact, in some ways it makes them more dangerous because they don’t have empathy. They don’t have compassion. They don’t have moral intuitions. They’re just pure optimizers. They’re just pursuing whatever goal they were given or whatever goal they developed.

And so I think that consciousness is orthogonal to the AI risk question. I think we should worry about AI whether or not it’s conscious.

Donal 00:36:39Okay. And another counter-argument is that we can just build multiple AIs and have them check each other. What’s your response?

Liron 00:36:46I think that helps a little bit, but I don’t think it solves the fundamental problem. The problem is that if all the AIs are unaligned, then having them check each other doesn’t help. They’re all pursuing their own goals.

They’re not pursuing human goals.

Liron 00:37:01Now, if you have some AIs that are aligned and some that are unaligned, then maybe the aligned ones can help catch the unaligned ones. But that only works if we actually know how to build aligned AIs in the first place. And I don’t think we do.

So I think that having multiple AIs is a useful safety measure, but it’s not a substitute for solving alignment.

Liron 00:37:20We still need to figure out how to build aligned AIs. And once we have that, then yeah, having multiple AIs can provide an extra layer of safety. But without solving alignment first, I don’t think it helps much.

Donal 00:37:33Okay. And another counter-argument is that P(Doom) is too high in the doomer community. People are saying 50%, 70%, 90%. Those seem like unreasonably high probabilities. What’s your response?

Liron 00:37:46I mean, I think those probabilities are reasonable given what we know. I think that if you look at the alignment problem, if you look at how hard it is, if you look at how little progress we’ve made, if you look at how fast AI capabilities are advancing, I think that P(Doom) being high is justified.

Liron 00:38:04Now, different people have different probabilities. Some people think it’s 10%, some people think it’s 50%, some people think it’s 90%. I’m probably somewhere in the middle. I think it’s maybe around 50%. But the exact number doesn’t matter that much.

The point is that the risk is high enough that we should take it seriously.

Liron 00:38:21I mean, if someone told you that there’s a 10% chance that your house is going to burn down, you would take that seriously. You would buy fire insurance. You would install smoke detectors. You wouldn’t say, oh, only 10%, I’m not going to worry about it.

So I think that even if P(Doom) is only 10%, we should still take it seriously. But I think it’s actually much higher than 10%. I think it’s more like 50% or higher.

Liron 00:38:42And so I think we should be very worried. I think we should be putting a lot of resources into solving alignment. And I think we should be considering extreme measures like pausing AI development until we figure out how to do it safely.

Donal 00:38:57Okay. And just on that point about pausing AI development, some people say that’s not realistic because of competition between countries. Like if the US pauses, then China will just race ahead. What’s your response?

Liron 00:39:10I think that’s a real concern. I think that international coordination is hard. I think that getting all the major AI powers to agree to a pause is going to be difficult. But I don’t think it’s impossible.

I think that if the risk is high enough, if people understand the danger, then countries can coordinate.

Liron 00:39:28I mean, we’ve coordinated on other things. We’ve coordinated on nuclear weapons. We have non-proliferation treaties. We have arms control agreements. It’s not perfect, but it’s better than nothing. And I think we can do the same thing with AI.

I think we can have an international treaty that says, hey, we’re not going to build superintelligent AI until we figure out how to do it safely.

Liron 00:39:47Now, will some countries cheat? Maybe. Will it be hard to enforce? Yes. But I think it’s still worth trying. I think that the alternative, which is just racing ahead and hoping for the best, is much worse.

So I think we should try to coordinate internationally and we should try to pause AI development until we solve alignment.

Donal 00:40:06Okay. And just on the economic side of things, so obviously AI is creating a lot of economic value. Some people say that the economic benefits are so large that we can’t afford to slow down. What’s your response?

Liron 00:40:19I think that’s short-term thinking. Yes, AI is creating economic value. Yes, it’s helping businesses be more productive. Yes, it’s creating wealth. But if we lose control of AI, all of that wealth is going to be worthless.

If humanity goes extinct or if we lose power, it doesn’t matter how much economic value we created.

Liron 00:40:38So I think that we need to take a longer-term view. We need to think about not just the economic benefits of AI, but also the existential risks. And I think that the existential risks outweigh the economic benefits.

I think that it’s better to slow down and make sure we do it safely than to race ahead and risk losing everything.

Liron 00:40:57Now, I understand that there’s a lot of pressure to move fast. There’s a lot of money to be made. There’s a lot of competition. But I think that we need to resist that pressure. I think we need to take a step back and say, okay, let’s make sure we’re doing this safely.

Let’s solve alignment before we build superintelligent AI.

Donal 00:41:15Okay. And just on the distribution of AI benefits, so some people worry that even if we don’t have rogue AI, we could still have a scenario where AI benefits are concentrated among a small group of people and everyone else is left behind. What are your thoughts on that?

Liron 00:41:30I think that’s a legitimate concern. I think that if we have powerful AI and it’s controlled by a small number of people or a small number of companies, that could lead to extreme inequality. It could lead to a concentration of power that’s unprecedented in human history.

Liron 00:41:47And so I think we need to think about how to distribute the benefits of AI widely. We need to think about things like universal basic income. We need to think about how to make sure that everyone benefits from AI, not just a small elite.

But I also think that’s a secondary concern compared to the alignment problem. Because if we don’t solve alignment, then it doesn’t matter how we distribute the benefits. There won’t be any benefits to distribute because we’ll have lost control.

Liron 00:42:11So I think alignment is the primary concern. But assuming we solve alignment, then yes, distribution of benefits is an important secondary concern. And we should be thinking about that now.

We should be thinking about how to structure society so that AI benefits everyone, not just a few people.

Liron 00:42:28Now, some people talk about things like public ownership of AI. Some people talk about things like universal basic income. Some people talk about things like radical transparency in AI development. I think all of those ideas are worth considering.

I think we need to have a public conversation about how to distribute the benefits of AI widely.

Liron 00:42:47But again, I think that’s secondary to solving alignment. First, we need to make sure we don’t lose control. Then we can worry about how to distribute the benefits fairly.

Donal 00:43:00Okay. And just on the concept of AI governance, so obviously there are a lot of different proposals for how to govern AI. What do you think good AI governance would look like?

Liron 00:43:11I think good AI governance would have several components. First, I think we need international coordination. We need treaties between countries that say we’re all going to follow certain safety standards. We’re not going to race ahead recklessly.

Liron 00:43:26Second, I think we need strong regulation of AI companies. We need to make sure that they’re following best practices for safety. We need to make sure that they’re being transparent about what they’re building. We need to make sure that they’re not cutting corners.

Third, I think we need a lot of investment in AI safety research. We need to fund academic research. We need to fund research at AI companies. We need to fund independent research.

Liron 00:43:48Fourth, I think we need some kind of international AI safety organization. Something like the IAEA for nuclear weapons, but for AI. An organization that can monitor AI development around the world, that can enforce safety standards, that can coordinate international responses.

Liron 00:44:06And fifth, I think we need public education about AI risk. We need people to understand the dangers. We need people to demand safety from their governments and from AI companies. We need a broad public consensus that safety is more important than speed.

So I think good AI governance would have all of those components. And I think we’re not there yet. I think we’re still in the early stages of figuring out how to govern AI.

Liron 00:44:31But I think we need to move fast on this because AI capabilities are advancing quickly. And we don’t have a lot of time to figure this out.

Donal 00:44:40Okay. And just on the role of governments versus companies, so obviously right now, AI development is mostly driven by private companies. Do you think governments should take a bigger role?

Liron 00:44:51I think governments need to take a bigger role, yes. I think that leaving AI development entirely to private companies is dangerous because companies have incentives to move fast and to maximize profit. And those incentives are not always aligned with safety.

Liron 00:45:08Now, I’m not saying that governments should take over AI development entirely. I think that would be a mistake. I think that private companies have a lot of talent, they have a lot of resources, they have a lot of innovation. But I think that governments need to provide oversight.

They need to set safety standards. They need to enforce regulations.

Liron 00:45:27And I think that governments need to invest in AI safety research that’s independent of companies. Because companies have conflicts of interest. They want to deploy their products. They want to make money. And so they might not be as cautious as they should be.

So I think we need independent research that’s funded by governments or by foundations or by public institutions.

Liron 00:45:48And I think that governments also need to coordinate internationally. This is not something that one country can solve on its own. We need all the major AI powers to work together. And that’s going to require government leadership.

Donal 00:46:03Okay. And just on the concept of AI existential risk versus other existential risks like climate change or nuclear war, how do you think AI risk compares?

Liron 00:46:13I think AI risk is the biggest existential risk we face. I think it’s more urgent than climate change. I think it’s more likely than nuclear war. I think that we’re more likely to lose control of AI in the next 10 years than we are to have a civilization-ending nuclear war or a civilization-ending climate catastrophe.

Liron 00:46:32Now, I’m not saying we should ignore those other risks. I think climate change is real and serious. I think nuclear war is a real possibility. But I think that AI is the most imminent threat. I think that AI capabilities are advancing so quickly that we’re going to hit the danger zone before we hit the danger zone for those other risks.

Liron 00:46:52And also, I think AI risk is harder to recover from. If we have a nuclear war, it would be terrible. Millions of people would die. Civilization would be set back. But humanity would probably survive. If we lose control of AI, I don’t think humanity survives.

I think that’s game over.

Liron 00:47:10So I think AI risk is both more likely and more severe than other existential risks. And so I think it deserves the most attention and the most resources.

Donal 00:47:21Okay. And just on the timeline again, so you mentioned three to seven years for ASI. What happens after that? Like, what does the world look like if we successfully navigate this transition?

Liron 00:47:32Well, if we successfully navigate it, I think the world could be amazing. I think we could have superintelligent AI that’s aligned with human values. And that AI could help us solve all of our problems. It could help us cure diseases. It could help us solve climate change.

It could help us explore space.

Liron 00:47:50It could help us create abundance. We could have a post-scarcity economy where everyone has everything they need. We could have radical life extension. We could live for thousands of years. We could explore the universe. It could be an amazing future.

But that’s if we successfully navigate the transition. If we don’t, I think we’re doomed.

Liron 00:48:11I think that we lose control, the AI pursues its own goals, and humanity goes extinct or becomes irrelevant. And so I think that the next few years are the most important years in human history. I think that what we do right now is going to determine whether we have this amazing future or whether we go extinct.

And so I think we need to take this very seriously. We need to put a lot of resources into solving alignment. We need to be very careful about how we develop AI.

Liron 00:48:37And we need to be willing to slow down if necessary. We need to be willing to pause if we’re not confident that we can do it safely. Because the stakes are too high. The stakes are literally everything.

Donal 00:48:50Okay. And just on your personal motivations, so obviously you’re spending a lot of time on this. You’re running Doom Debates. Why? What motivates you to work on this?

Liron 00:49:00I think it’s the most important thing happening in the world. I think that we’re living through the most important period in human history. And I think that if I can contribute in some small way to making sure that we navigate this transition successfully, then that’s worth doing.

Liron 00:49:18I mean, I have a background in tech. I have a background in computer science. I understand AI. And I think that I can help by having conversations, by hosting debates, by bringing people together to discuss these issues.

I think that there’s a lot of confusion about AI risk. Some people think it’s overhyped. Some people think it’s the biggest risk. And I think that by having these debates, by bringing together smart people from different perspectives, we can converge on the truth.

Liron 00:49:45We can figure out what’s actually going on. We can figure out how worried we should be. We can figure out what we should do about it. And so that’s why I do Doom Debates. I think that it’s a way for me to contribute to this conversation.

And I think that the conversation is the most important conversation happening right now.

Donal 00:50:04Okay. And just in terms of what individuals can do, so if someone’s listening to this and they’re concerned about AI risk, what would you recommend they do?

Liron 00:50:14I think there are several things people can do. First, educate yourself. Read about AI risk. Read about alignment. Read Eliezer Yudkowsky. Read Paul Christiano. Read Stuart Russell. Understand the issues.

Liron 00:50:28Second, talk about it. Talk to your friends. Talk to your family. Talk to your colleagues. Spread awareness about AI risk. Because I think that right now, most people don’t understand the danger. Most people think AI is just a cool new technology.

They don’t realize that it could be an existential threat.

Liron 00:50:46Third, if you have relevant skills, consider working on AI safety. If you’re a researcher, consider doing AI safety research. If you’re a software engineer, consider working for an AI safety organization. If you’re a policy person, consider working on AI governance.

We need talented people working on this problem.

Liron 00:51:05Fourth, donate to AI safety organizations. There are organizations like MIRI, the Machine Intelligence Research Institute, or the Future of Humanity Institute at Oxford, or the Center for AI Safety. These organizations are doing important work and they need funding.

Liron 00:51:22And fifth, put pressure on governments and companies. Contact your representatives. Tell them that you’re concerned about AI risk. Tell them that you want them to prioritize safety over speed. Tell them that you want strong regulation.

And also, if you’re a customer of AI companies, let them know that you care about safety. Let them know that you want them to be responsible.

Liron 00:51:44So I think there are a lot of things individuals can do. And I think that every little bit helps. Because this is going to require a collective effort. We’re all in this together. And we all need to do our part.

Donal 00:51:58Okay. And just on the concept of acceleration versus deceleration, so some people in the tech community are accelerationists. They think we should move as fast as possible with AI. What’s your response to that?

Liron 00:52:11I think accelerationism is incredibly dangerous. I think that the accelerationists are playing Russian roulette with humanity’s future. I think that they’re so focused on the potential benefits of AI that they’re ignoring the risks.

Liron 00:52:28And I think that’s a huge mistake. I think that we need to be much more cautious. Now, I understand the appeal of accelerationism. I understand that AI has amazing potential. I understand that it could help solve a lot of problems. But I think that rushing ahead without solving alignment first is suicidal.

I think that it’s the most reckless thing we could possibly do.

Liron 00:52:51And so I’m very much on the deceleration side. I think we need to slow down. I think we need to pause. I think we need to make sure we solve alignment before we build superintelligent AI. And I think that the accelerationists are wrong.

I think they’re being dangerously naive.

Donal 00:53:09Okay. And just on the economic implications of AI, so you mentioned earlier that AI could automate away a lot of jobs. What do you think happens to employment? What do you think happens to the economy?

Liron 00:53:21I think that in the short term, we’re going to see a lot of job displacement. I think that AI is going to automate a lot of white-collar jobs. Knowledge workers, office workers, programmers even. I think a lot of those jobs are going to go away.

Liron 00:53:36Now, historically, when technology has automated jobs, we’ve created new jobs. We’ve found new things for people to do. But I think that AI is different because AI can potentially do any cognitive task. And so I’m not sure that we’re going to create enough new jobs to replace the jobs that are automated.

And so I think we might end up in a situation where we have mass unemployment or underemployment.

Liron 00:53:59Now, in that scenario, I think we’re going to need things like universal basic income. We’re going to need a social safety net that’s much stronger than what we have now. We’re going to need to rethink our economic system because the traditional model of everyone works a job and earns money and uses that money to buy things, that model might not work anymore.

Liron 00:54:20But again, I think that’s a secondary concern compared to the alignment problem. Because if we don’t solve alignment, we’re not going to have mass unemployment. We’re going to have mass extinction. So I think we need to solve alignment first.

But assuming we do, then yes, we need to think about these economic issues.

Liron 00:54:38We need to think about how to structure society in a world where AI can do most jobs. And I don’t think we have good answers to that yet. I think that’s something we need to figure out as a society.

Donal 00:54:52And on the UBI point, so you mentioned universal basic income. Some people worry that if you have UBI, people will lose meaning in their lives because work gives people meaning. What’s your response?

Liron 00:55:04I think that’s a legitimate concern. I think that work does give people meaning. Work gives people structure. Work gives people social connections. Work gives people a sense of purpose. And so I think that if we have UBI and people don’t have to work, we’re going to need to think about how people find meaning.

Liron 00:55:24But I also think that not everyone finds meaning in work. Some people work because they have to, not because they want to. And so I think that UBI could actually free people to pursue things that are more meaningful to them. They could pursue art. They could pursue hobbies.

They could pursue education. They could pursue relationships.

Liron 00:55:44So I think that UBI is not necessarily bad for meaning. I think it could actually enhance meaning for a lot of people. But I think we need to be thoughtful about it. We need to make sure that we’re creating a society where people can find meaning even if they’re not working traditional jobs.

And I think that’s going to require some creativity. It’s going to require some experimentation.

Liron 00:56:06But I think it’s doable. I think that humans are adaptable. I think that we can find meaning in a lot of different ways. And I think that as long as we’re thoughtful about it, we can create a society where people have UBI and still have meaningful lives.

Donal 00:56:23Okay. And just on the power dynamics, so you mentioned earlier that AI could lead to concentration of power. Can you talk a bit more about that?

Liron 00:56:31Yeah. So I think that whoever controls the most advanced AI is going to have enormous power. I think they’re going to have economic power because AI can automate businesses, can create wealth. They’re going to have military power because AI can be used for weapons, for surveillance, for cyber warfare.

They’re going to have political power because AI can be used for propaganda, for manipulation, for social control.

Liron 00:56:56And so I think that if AI is controlled by a small number of people or a small number of countries, that could lead to an unprecedented concentration of power. It could lead to a kind of authoritarianism that we’ve never seen before.

Because the people who control AI could use it to control everyone else.

Liron 00:57:17And so I think that’s a real danger. I think that we need to think about how to prevent that concentration of power. We need to think about how to make sure that AI is distributed widely, that the benefits are distributed widely, that the control is distributed widely.

And I think that’s going to be very difficult because there are strong incentives for concentration. AI development is very expensive. It requires a lot of compute. It requires a lot of data. It requires a lot of talent.

Liron 00:57:43And so there’s a natural tendency for AI to be concentrated in a few large companies or a few large countries. And I think we need to resist that tendency. We need to think about how to democratize AI.

How to make sure that it’s not controlled by a small elite.

Donal 00:58:01Okay. And just on the geopolitical implications, so obviously there’s a lot of competition between the US and China on AI. How do you think that plays out?

Liron 00:58:10I think that’s one of the scariest aspects of the situation. I think that the US-China competition could lead to a dangerous race dynamic where both countries are rushing to build the most advanced AI as quickly as possible, and they’re cutting corners on safety.

Liron 00:58:27And I think that that’s a recipe for disaster. I think that if we’re racing to build superintelligent AI without solving alignment, we’re going to lose control. And it doesn’t matter if it’s the US that loses control or China that loses control. We all lose.

So I think that the US-China competition is very dangerous. And I think we need to find a way to cooperate instead of competing.

Liron 00:58:50Now, that’s easier said than done. There’s a lot of mistrust between the US and China. There are geopolitical tensions. But I think that AI risk is a common enemy. I think that both the US and China should be able to recognize that if we lose control of AI, we all lose.

And so we should be able to cooperate on safety even if we’re competing on other things.

Liron 00:59:13And so I think that we need some kind of international agreement, some kind of treaty, that says we’re all going to follow certain safety standards. We’re not going to race ahead recklessly. We’re going to prioritize safety over speed.

And I think that’s going to require leadership from both the US and China. It’s going to require them to put aside their differences and work together on this common threat.

Donal 00:59:36Okay. And just on the role of China specifically, so some people worry that even if the US slows down on AI, China will just race ahead. What’s your response?

Liron 00:59:45I think that’s a real concern, but I don’t think it’s insurmountable. I think that China also faces the same risks from unaligned AI that we do. I think that Chinese leadership, if they understand the risks, should be willing to cooperate.

Liron 01:00:03Now, there’s a question of whether they do understand the risks. And I think that’s something we need to work on. I think we need to engage with Chinese AI researchers. We need to engage with Chinese policymakers. We need to make sure that they understand the danger.

Because if they understand the danger, I think they’ll be willing to slow down.

Liron 01:00:23Now, if they don’t understand the danger or if they think that they can win the race and control AI, then that’s more problematic. But I think that we should at least try to engage with them and try to build a common understanding of the risks.

And I think that if we can do that, then cooperation is possible.

Liron 01:00:42But if we can’t, then yes, we’re in a very dangerous situation. Because then we have a race dynamic where everyone is rushing to build superintelligent AI and no one is prioritizing safety. And I think that’s the worst possible outcome.

Donal 01:00:57Okay. And just going back to economic implications, you mentioned gradual disempowerment earlier. Can you elaborate on that?

Liron 01:01:04Yeah, gradual disempowerment. So the idea is that even if we have aligned AI, even if the AI is doing what we want it to do, we could still lose power gradually over time.

Because as AI becomes more and more capable, humans become less and less relevant.

Liron 01:01:22And so even if the AI is technically aligned, even if it’s doing what we tell it to do, we could end up in a situation where humans don’t have any power anymore. Where all the important decisions are being made by AI, and humans are just kind of along for the ride.

And I think that’s a concern even if we solve the technical alignment problem.

Liron 01:01:42Now, there’s different ways this could play out. One way is that you have AI-controlled corporations that are technically serving shareholders, but the shareholders are irrelevant because they don’t understand what the AI is doing. The AI is making all the decisions.

Another way is that you have AI-controlled governments that are technically serving citizens, but the citizens don’t have any real power because the AI is doing everything.

Liron 01:01:53There’s really no point to make everybody desperately poor. You know, we already have a welfare system in every first world country. So I don’t see why we shouldn’t just pad the welfare system more if we can afford it.

Um, there’s a bigger problem though, called gradual disempowerment.

Liron 01:02:05It’s an interesting paper by David Krueger, I think a couple other authors, and it just talks about how yeah, you can have universal basic income, but the problem now is that the government doesn’t care about you anymore.

You become like an oil country, right? Oil countries are often not super nice to their citizens because the government pays the citizens and the citizens don’t really pay tax to the government.

Liron 01:02:24So it becomes this very one-sided power relationship where the government can just abuse the citizens. You know, you just have a ruling family basically. And I think there’s countries, I think maybe Sweden has pulled it off where they’re able to have oil and the citizens are still somewhat democratic.

Liron 01:02:39But then you have other countries like Saudi Arabia, I think, you know, other oil countries there, which maybe aren’t pulling it off so much, right? Maybe they are bad to their citizens, I don’t know.

And so that’s the gradual disempowerment issue. But look, for me personally, I actually think all of those still take the backseat to the AI just being uncontrollable.

Liron 01:02:55So I don’t even think we’re going to have an economy where you’re going to have rich owners of capital getting so rich while other people who didn’t buy enough stock get poor. I don’t even think we’re going to have that for a very long time.

I just think we’re going to have, our brains are just going to be outclassed.

Liron 01:03:08I just think, you know, they’re just going to come for our houses and it’s not even going to be a matter of buying it from us. It’s just going to be, you know, get out. You’re dead.

Regulation, Policy, and Surviving The Future With AI

Donal 01:03:16Okay, so last thing, on a related point, do you actually think that democratic capitalism can survive regulation against AI?

So the kind of regulation we need to do right now, so if we want to pause it and we want to prevent this catastrophic outcome actually happening, can a democratic capitalist society survive that?

Donal 01:03:23Because I’ve seen pushbacks where people say from a libertarian perspective, you can’t stop people from innovating or you can’t stop businesses from investing. So what are your thoughts there? Would everything have to change?

Liron 01:03:43Yeah. The specific policy that I recommend is to just have a pause button. Have an international treaty saying, hey, it’s too scary to build AI right now. It’s unlocking, you know, it’s about to go uncontrollable.

We could be losing power in as little as five or 10 years.

Liron 01:03:56We could just end, you know, game over for humanity. We don’t want that. So we’re going to build a centralized off button. It’s going to live in a UN data center or something, right? Some kind of international coordination between the most powerful AI countries.

And when you’re saying, won’t that lead to tyranny?

Liron 01:04:11I mean, look, there’s always risks, right? I mean, I tend to normally skew libertarian. This is the first time in my life when I’ve said, let’s do this central non-libertarian thing. It’s this one exception. Do I think this one exception will lead to tyranny?

I don’t think so.

Liron 01:04:27I mean, you still have the rest of the economy, you still have virtual reality, you still have a space program. You still use the fruits of AI that we all have so far before we’ve hit the pause button. So no, I think people are overworrying.

Donal 01:04:34So you’re happy with LLMs? Are current LLMs acceptable from your perspective? Do you think that they can stay?

Liron 01:04:40Yeah, I think that they are acceptable, but I think that they were too risky. Even with GPT-4 Turbo?

Donal 01:04:45Even with GPT-4 Turbo?

Liron 01:04:45Yeah. Even with GPT-4 Turbo, because I think if an LLM tried its hardest right now to destroy the world, I think that humans could shut it down. Right? I don’t think it’s game over for humans.

And so the situation is, it’s like, like I said, it’s like the Icarus situation.

Liron 01:04:59And you’re saying, hey, are you happy with how far you’ve flown? Yeah. And maybe tomorrow we fly a little higher and we’re not dead yet. Am I happy? Yes. But do I think it’s prudent to try flying higher? No. Right?

So it’s a tough situation, right? Because I can’t help enjoying the fruits of flying higher and higher, right?

Liron 01:05:16I use the best AI tools I can, right? But I just, there’s just a rational part of my brain being like, look, we gambled. Yeah, we gambled and won. We gambled and won. We gambled and won. We’re about to approach the superintelligent threshold.

Are we going to win after we gambled to that threshold? Logic tells me probably not.

Donal 01:05:31Okay. And sorry, last point. I know it keeps rolling. How do you actually use them in preparation for your debates? You use LLMs? You trust them at that level?

Liron 01:05:39Yeah. I mean, you know, I didn’t use it for this interview. I mean, you know, normally I just make my own outline manually, but I certainly use AI at my job. I use AI to help customer service at my company.

Liron 01:05:49I mean, I try to use the best AI tools I can because it’s amazing technology, right? It’s the same reason I use the best MacBook that I can. I mean, I like using good tools, right? I’m not opposed to using AI. I think AI has created a lot of value.

And again, it kind of makes me look dumb where it’s like the next version of AI comes out, I start using it, I see it creating a ton of value.

Liron 01:06:08And then you can come to me and go, see Liron, what were you scared of? Right? We got an AI and it’s helping. Yeah. What was I scared of? Because we rolled a die. We gambled and won. Okay, I’m taking my winnings, right?

The winnings are here on the table. I’m going to take my winnings. That doesn’t mean I want to be pushing my luck and gambling again.

Donal 01:06:21Yeah. Well said. Okay. Liron, thank you for your time. It was an absolute pleasure. I really enjoyed it.

Liron 01:06:27Yeah, thanks Donal. This was fun.

Doom Debates’ Mission is to raise mainstream awareness of imminent extinction from AGI and build the social infrastructure for high-quality debate.

Support the mission by subscribing to my Substack at DoomDebates.com and to youtube.com/@DoomDebates, or to really take things to the next level: Donate 🙏