Well, hello Ted. Hey, this is Ted. And Brandon, let's start over. Take two.
Hey, this is Brandon. This is Ted. And we are concerned with AI, or we're concerneding AI. The show is concerning AI.
Is AI concerned with us? Not yet. OK. It may never be.
That remains to be seen. That would be pretty awesome. It would be awesome. But I don't think it would ever be that important.
You know, I signed up for AI service this week. x.ai. x.ai. So you're about to have AI at your finger dance?
Hopefully. I don't know how long I'm going to be on the wait list. But this thing is supposed to help me schedule appointments. I can just CC it on an email when I'm trying to schedule somebody.
And then that takes over the email thread, and it just comes back to me when it's done. Well, you've gotten more important already. I haven't paid for the service. They don't even have a paid version yet.
And I already want to pay for it. I think it'll be cool. Well, having somebody who schedules your appointments, I think it's a real signal or of importance. Well, you know, that's why I wanted it.
It's for, you know, it looks good. Yeah, for sure. So what do we got today? All right.
We've got a really fascinating article. I don't know if it's a article. It's short. It's really short.
It's a post. Let's call it a post. I like it. It's in the library of economics and liberty.
And it's by a guy called Brian Kaplan. I'll put it in the show notes so you guys can read it. It's kind of a libertarian economics professor. Yeah, ish.
March 25th, 2016. You're studying pacifist. Yeah, it's so cool. Interesting guy.
The title is So Far. So Far. Yeah. So he starts talking about social forces like industrialization.
So Far. Industrialization is a good thing. Boom. Big win.
Population growth. So Far. Boom. Appellation growth.
Big win. Computers. So Far. I'm going to stop saying boom.
But big win. Computers have been a big win. So Far. What about nuclear physics?
Probably net positive. Net positive. Immigration. So Far.
Seems like a good thing. Good. Big win. Yeah.
So what he's saying is that the So Far provides Oh sounds ominous. Like, well, we're just about at the end. Like we can imagine all kinds of doomsday scenarios in any of these five areas because the So Far makes it think, make it look like, well, it's just a matter of time before the goodness comes to an end. Well, you think about movies, right?
This is how they always set you up. That's right. The story is always set you up with la la la la la la la la. And you know that it can't be true.
It can't be. Because you know, so Far it's been bad disasters about to fall. Yep. That's just part of all stories have that built into it.
So we're sensitive to we're very sensitive to so far. So let the nightmares begin. He says, you know, we got this so far and therefore we can tell all kinds of doomsday scenarios like industrialization could lead to environmental apocalypse, global totalitarianism, population growth could lead to mass famine. And our favorite going out computers could lead to ultra unfriendly artificial intelligence.
Yeah. So they could. And I've heard all of these. Yeah.
From people who are clearly intelligent. Yeah. And they're doing they scenarios and they seem like you could write a story where they're plausible in all cases. All right.
So he I think he really he punctures all of this just super nicely with this sentence. All right. So that's what in practice most of us pick out a few quote serious nightmares and ideological grounds and dismiss the rest as too silly to entertain. So those ones that are too silly to entertain.
They just help us go to sleep at night because we're like, oof, at least that's not going to happen. Yeah, totally crazy people worrying about that little schon Freud. Yeah. Oh, you poor people worrying about that silly, silly thing.
Oh, wait. But what about this one? Yeah. What about this one?
The one that I'm concerned about? Yeah. No, it's not. No, no, it's serious.
Very serious. So people do think people people like who look people like our friends, people like whoever, people think that this AI concerned that we have might be silly. They do. A lot of people.
Yeah. There's a lot of people. In fact, maybe it is. It's so strange that people could dismiss the one serious concern.
We obviously picked out the right one. We've got the right one. Yeah. Yeah.
So I think it's I think it's interesting what he does to back this up to. He says, Hey, look back over all of history. Yeah. So I think that's a lot of this question.
Have people ever predicted doom before in the past? I think they have many times. Yeah. Pick any, you know, go back to 30,000 years around the campfire.
Someone's going to be telling about how the sky is about to fall or whatever it is. Yeah. We've had stories that were related to the gods or related to nature throughout history. And when we got to advances in technology, we had always these apocalyptic, the luddites were railing against technology and people were worried about the introduction of books because what books?
Wow. Yeah. They weren't. But in a good way.
Right. I think there's been doom scenarios all along, whether that's from the suit sayers, from the holy people or from whoever. Just anybody that is a thinking person can clearly find something. All right.
So Brian Kaplan is pointing out that we're still here. Are we? We are. We are.
We're set aside the simulations right now. But I mean, assuming that we are still here, what he's saying then is therefore it's very, very unlikely that we've gotten lucky a million times in a row. Right. I think that's from his.
That's what he said. That's right. I think that that makes sense to me. It completely makes sense to me too.
Therefore, the dooms that are being forecasted right now are probably not really likely. They're probably in the same category. Yeah. Yeah.
It's actually a pretty good argument, I think. I think it is too. We're not talking about, if we're talking about like one in a million, then they can't be a 1% likelihood of any of these doom scenarios or one of them will already happen. Right.
Right. And they can't even be one in a thousand. They can't even be a tenth of a percent. Right.
I don't know how many of these doom scenarios there have been, but it seems like each one, the probability of each one would have to be very small infotestimal in order for us still to be alive today. Yeah. And with the benefit of hindsight, it's easy to laugh at what people use to think. Yeah.
Probably probably almost from any vantage point. Is that true? Oh, I think that's true. It's easy to look back and I think so.
I think that's probably true. Even if your belief is no less silly than the one that came before, from where you sit in the wisdom of the future, it's easy to look back and laugh. Yeah. I don't think that you and I can sit here and laugh at our concern about AI, because it's a concern that we have.
Yeah. We may look back at it in laugh in 20 years. Or even, you know, people 100 years from now may find this moment, a historically interesting moment because of how many people bought into this particular doomsday scenario and then they talk about, you know, all the things that went into it and of course how silly they are or how silly they were. Right.
Yeah. That could be cool. I like that. If this were true, then, you know, we don't need to show anymore.
You know, we can just call it quits because we don't need to be concerned. You know, I'm kind of living that way, even though I'm still doing the show. Yeah. You know, I mean, I'm like going to work.
I'm doing all the time horizon on something like this. We're really short. Seems like I should be doing something different other than just the show. But I'm not.
So there must be a big part of me that actually kind of does what we do. Yeah. Well, maybe, maybe it's also, I mean, it's such a big problem that it's hard to tell what we should be doing, you know, and maybe just having the conversations that we've been having on the show and with people online about the show, maybe that's what we should be doing right now. There are lots of people talking about say, I say the author in the world anyway, you know, not lots, but you know, enough.
There are people out there talking about it. So it's unclear what we could be doing. But if this were really true, if we really believed this argument, and this argument, I feel is pretty strong in the sense that we are still around and industrialization. One thing he points out here, industrialization, for example, if all of the industrialization do Zoom stories only started about 10 years ago, then maybe we got to take them really seriously and think this is something different.
There's a different kind of a thing happening here because we never forecast a Doom because industrialization until now. And therefore now we need to start paying attention. But the industrialization Doom story started in 1750. Therefore, you know, it's been a continuous stream of those things scenarios.
So therefore we don't need to be concerned today. All right. Let's read one more quote out of this. All right.
All right. And this sort of, I think sums it up. Well, I guess we already completely summarized it. So well, let's just read that last paragraph.
If you're more of it across the board, I've got nothing more to say, although I'm eager to bet you. If you're selectively morbid, though, which I think we're both selected. I think we are. Yeah.
All right. He's talking to us. He is. He's talking to us.
Yeah. Thanks, Ryan. I think probably had us in mind. I'm sure he listens to the show for sure.
If you're selectively morbid, though, I'd like to know why the nightmares that keep you up at night are so much more compelling than the nightmares that put you to sleep. It's a good writer. Yeah. I mean, I have to be honest, I am not worried about the ocean current shutting down because of global warming and sudden like freezing of the earth into snow wall.
Earth again. Yeah. I mean, you see, but it's got to be a non-zero probability, you know, expectation that that could happen. Right.
So why is that? I mean, that would suck. It would. And if there's any chance of that, we shouldn't be worried about that just as much as we're worried about this.
I don't know. I guess it depends on if we think any part of humanity can survive that kind of ecological disaster. In my mind, it is a difference in kind. It is probably a little slower, even if it was fast, fast might be okay.
Right. How long would it take to indicate earlier? Right. I mean, the earth is hard to the earth is a yeah, it's hard to.
Okay. So it's the best of the environment or the, I guess, worst of the environmental scenarios. But none of them keep me up. None of the environmental catastrophes feel like they could happen at the snap of a finger.
They all feel like they are slow and they might grind us down, but they're not going to like, suck us out. Right. And even if they wipe out 90% of the human population or 90% of the species on earth, it still feels like biology can survive, you know, like life. Life on earth will be, we can survive.
We can survive. I don't want any of us to survive. And we have gone through a couple of snowballers. We sure have.
So I mean, that seems about as bad as it can get to me. Yeah. And I think about this one, why does this one keep me up at night? And it actually doesn't, for the same reasons that it doesn't, you know, for you.
Right. So it's not actually keeping you up because I'm not sure what to do about it. Yeah. But this is the one that I'm concerned about of all the doomsday scenarios.
And it's because, well, we even have a word for it. It's the singularity. You know, the mathematical singularity is the point which cannot be computed. And when we talk about the singularity of artificial intelligence, it's the point beyond which we can't predict because it's this, it's what, it's a natural algorithm, an exponential curve.
All right. So, so let's trace his conversation with Elie Azir a little bit further. I don't know. I think that's the point of view of compelling.
He's maybe a high priest of our fears, for us. He's our doomsair. He sure is. Okay.
So he's our doomsair. Yeah. And he says, which of these things lays out a list of, that's basically the case for doomsday AI. Right.
We'll put that in the show notes as well. And, you know, Brian is, is kind of short in his reply. He doesn't really go into a huge amount of detail here. But I think if I were to try to boil down this so far, my response post of Brian's, it seems like the place that he diverges from Elie Azir is a disagreement that rapid capability gain and large capability differences are that plausible.
Like, he just doesn't see that actually happening. He seems to be thinking linearly. He seems to be looking at the progress of the last year and thinking the next year will be about that much progress. He thinks the last 10 years, let's extrapolate that forward.
The last, the next 10 years will be about the same amount of progress that we've had in the next 10 years. And therefore, he says, you know, it's possible that, you know, I'll be grateful if we have self-driving cars before my younger son is old enough to drive 10 years from now. My kids could get their drive license in nine years. I don't think they're ever going to need one.
What's the difference between me and him? Why do we think so differently? I think it's because I've started to train myself to think exponentially. Yeah.
Well, you know, I think Y2K, for those of you who are younger, you may not remember Y2K was the fear that when we turned over to 2000, suddenly all our dates that were two-digit dates were going to, you know, all our software systems were going to cause all kinds of problems and, you know, there could be financial meltdown. And who knows, maybe even airliners might plummet from the sky because of getting confused about the date. And we had to get to the point where we had two-digit dates because bits were expensive. We just never updated.
Yeah. Well, we just kept using the value-producing software that had the two-digit dates in it. And I think it's safe to say that was a fizzle, not that a huge amount of money didn't go into trying to fix all the problems that people could see. I mean, there was a lot of work to fix that.
So that was a huge hype. A huge hype. It was a big hype. That's the point.
It wasn't a correct hype or was it an over-hype? I don't know the answer. I'm happy to say that it was an over-hype. Okay, maybe it wasn't.
But if it was a correct amount of hype, and the reason that we didn't see the disasters is because people took it seriously and they took the actions that needed to be taken, then this would be a success. This would be a case where the hype was appropriate and therefore the actions were taken and solved the problem. Now, maybe that didn't happen. Maybe it was really over-hype.
Maybe we had 10 times more talk about it than we needed to. And maybe we even spent more money than we needed to fix it. But the point is we did fix it. We fixed the problem.
If we had it there was a problem. Well, there were some problems. I think that we had to change our systems from two-digit to four-digit. And so we had to do it at some point.
Whether or not there would have been a disaster had it turned to zero. You know, I don't know. I don't know. Which of our systems would really have been a disaster?
I still believe that it was a lot of hype and not worth probably. Alright, so let's say it wasn't over-hype and it was kind of a fizzling. Even if it wasn't, let's say it was. And what can we learn from it?
If it weren't. Right. So what came with it? This is another doomsday scenario in our past.
Yeah, that has to do with an acute transition. Yep. Right? That turned out to not be that acute.
Yep. So let's say that the... Okay, so the history of predicting AI has always been... It's 20 years out.
Yep. And I would say we're still saying that. And why is that? Because they keep going up by a year every year.
They'll link it to the predictions. Not for everybody, but for a lot of people. So why is that? Well, maybe a part of the reason for that is that we keep finding the brain is a lot trickier of a thing than we think it is.
And what we're actually doing is a lot harder than we think it is. And we keep pushing how hard that is back as we move along. The brain's actually doing more than we thought it's doing. It's got more computational resources than we thought.
Yep. It's a problem as exponential growth though because it doesn't really matter how much the brain is doing. If it's doing 10 times as far as much as we think, that's only like three doublings. Right.
Right. So the fact that it's always 20 years out is a very imprecise thing to say. And it could be that we were just silly. We were just stupid as humans looking into the future at some point.
And now we're seeing the exponential. Like are we at the elbow or the knee of the exponential now where things actually really change? Yeah. Is this a special one?
Are we right? Are we the only people, not that there are other people, is our camp that's worried about AI, the only people who are right to worry about their duvetus aesunaria. What do you think? It just feels irrecoverable.
If we have a runaway paperclip maximizer, there's just no way around it. If that comes about, there's no Terminator scenario, there's no movie story, there's no band of, there's no band of misfits trying to fight back, it's just over. It's over for humanity. Yeah, it's a very boring movie.
It's a movie that the paperclip maximizer is not going to want to watch. And we won't be around to watch it. Yeah. I don't know.
This still feels like about the best argument that we've come across so far. I mean, it feels pretty good. Yeah, I agree. So, you know, I'm going to go with him and say that I think that the likelihood of it happening are vanishingly small.
Is vanishingly small. What does that, let's say we go with it for a second. Okay, let's just try it. Okay, I got it.
All right. So, the likelihood of runaway, of an intelligence explosion, I think that's the crux of his argument. Are we saying that the likelihood of an intelligence explosion or an intelligence explosion that results in the destruction of all biological life on Earth? Well, or does his argument depend on the form of an intelligence result in life?
I felt like his argument hinged on it. He didn't really care about AI's having a non-trivial chance of being dangerously unnice. He's like, yeah, yeah, maybe whatever, but there's lots of stuff that's unnice. I mean, I think rolled up into that is the previous assertion that he made that it just seems un-possible that we're going to have the explosion.
I didn't tell you. Yeah, and we've been talking about AlphaGo the last couple episodes and we could say that, you know, AlphaGo gets a lot of hype and it's this big thing and my life isn't any different because of AlphaGo. What's the big deal? It's just more hype.
I mean, is that the crux of this argument that, look, we hype things all the time. They're just not that big a deal. Life goes on. Life goes on.
We always think that the time we're living in a special humanity has always thought that. Yeah, right. But life goes on. It's gone on all these years and we'll just keep going on.
Let's just keep on keeping on. Well, should we try to embrace that for a week and see how it's, let's leave it, let's leave it sit and yeah, let's see how we feel next week. It's hard. I'm a little unsettled.
I don't know if I can do it for a whole week. Yeah. All right, I'll try. Yeah.
Well, we said we would, right? When we kind of beat up on those ideas that, you know, that then offered us several shows ago and then we agreed that we were going to really try to embrace all ideas because if we're just ideological ourselves and this show isn't really worth much for anybody, you know, we, oh, it's still worth a lot for me to be. It's just a rah, rah, you know, rah, rah, rah. Are we great because we're so concerned?
Yeah, that's not what we're about. We're about trying to find out what's really going on. Okay. We'll try it for a week.
Yeah. We invite you to try it for a week to and let us know what you think. Yeah. So we'll see you online, concerning.ai or Facebook, the group is called concerning AI.
Yep. And we'll see you on the next show. Super. Bye.
Bye. Bye.