Tim Urban: The Artificial Intelligence Revolution | Oredev 2015

All right. Well, thanks for having me here. Sweden is cool. I was here once before, but this is more fun in Malmo. Other than Tuesday night, they showed me Swedish hospitality by giving me old fish. What's your issue? What's the deal with that? And so when I said, you know, I'm here, I'm going to do it, and, you know, a couple of the... locals brought me over and said, you know, you have to try it, so I think Ben was one of them. And I, you know, I took a deep breath and I had it, and I said, great, I said, you know, I did it, I managed, and I said, okay, your turn on this. And everyone said, no, we don't have to, we're Swedish. So that was, that was annoying. But, but it's been, it's been great other than that, and now I get to have a story for people when I get home. So, as Ben mentioned, I have this blog, Wait But Why, I started with it a couple years ago. And that's the... That's the thing there. That's the little logo. And I write about a lot of different things. It's kind of a work in progress. It's not really any specific topic. It's kind of, I write about whatever I would normally procrastinate by reading about. And then, of course, once that's the topic, then, of course, I don't want to do it anymore and I procrastinate on something else. And, you know, it goes on from there. But, you know, like, the kind of thing I would do in my spare time is I would try to figure out, okay, if I have two parents and four grandparents and eight... It starts to get bigger than the world population at some point, which then gets very confusing, but that's because there's this thing where the cousins marry each other, and anyway, it's a little off topic. By the way, a little fun fact I learned is that 80% of all human births have come from second cousins or closer. So that's something for you to nibble on while we're going through this. So then sometimes I just, I would say to myself, okay, so... The oceans are, you know, there's so much water in the oceans, but also, there's also only like five miles deep, or kilometers, eight kilometers deep, and if you were, if you just had the earth as like a meter-sized ball, that would be like less than a millimeter, so I would try to figure out how much water is there, and then I would do something like this, and that's how, if you stripped it all, it would be something like that size. Or sometimes I would say, how much gold? has ever been mined and I would look up that and then I would end up saying okay well that's how much gold's been mined and it's you know it'd be like a 20 meter cube and then I would start to say okay well actually But how much wealth is there in the whole world? And I would find some articles that have figured out, you know, that it's $241 trillion is the equivalent of all the net worth of every person in the world, so how big would that cube be? And it gets even bigger. And this is the kind of thing that I spend my days doing while everyone else is helping the world to be productive. But then I get serious and I kind of say, okay, well, how much pizza would that be? And I used the stupid New York Domino's price, which is like double the price of everywhere else. If you bought a Domino's pizza with all the world's wealth, you could cover Niger with it. One time, I always like to think about, okay, there are 7.3 billion people about right now. That's a lot. How if you put them all together, you know how much space would they really need and so I do things like this and I kind of figure out Okay, so you could kind of fit you know you can kind of fit every single human in New York City actually it would be it would be possible and And then I got curious because Scandinavia is all, you know, you guys are all fancy here with your paternal leave and all that. But there's not that many of you is the little thing about you guys. And so I kind of figured out, and this is not for you guys. I did this in the post like a year ago. But if you guys know Bornholm Island, which I'm sure you all do since it's like two feet from here, all Scandinavians... can fit in that square on Bornholm Island. So there you go. That's all two-dimensional. How about if we start stacking the people? So then we can get them actually in here. It would sit on top of Manhattan, all the people in there would be a very unpleasant place to be. Not really want to be part of that experiment, but and then then I can't help myself And I take it a step further and I say well actually the thing is there's still a lot of empty space in that building not between the people But also like in the atoms of the people because actually there's not much Stuff in us is actually mostly empty space and atom is almost all of it. It's just its nucleus So if we just shoved everyone down to just the actual matter all the empty space out of all humans, we could actually fit in here in an M&M. This is the important stuff, guys. Sometimes I get into kind of like life and I'll think about, okay, why are we all, you know, humans are kind of crazy a little bit. Like we all, it's a kind of, I think there's a primitive side of us that develops, that evolved, you know, in 60,000 BC or way before that, you know, millions of years. And that's kind of in conflict with this very higher, this higher part of us that's very rational, that has this amazing ability to reason, that's conscious. And so I'm always trying to figure out what's going on between those two sides. And so one thing brought me to is Generation Y, who are, you know, right now they're, I guess, they're like 25 to 33, 22 to 33 around that age. And they're very unhappy people. You know, they have some things going on. So I like ended up getting into it and I kind of figured out that it's, that they're, that, so everyone on the ground, that's like where they actually are in life. Above is where they project they are on Facebook and So each person looks at all the other people and say oh, they're all doing so well But actually like no one's actually doing they're all 25. No, nothing's happening and That you know so you know getting to stuff like that and then And then when we get to like that primitive side of us we have we have an issue and I think it's what I'm sure a lot of you guys deal with because You guys are also procrastinators. I know that. It's not just writers. And so the issue is we all have this other guy in our brain. So, you know, the rational decision maker is the guy who says, okay, today I'm going to do this and this and this. Got it all planned out. I'm going to pace myself. And over the next month, I'm going to get these things done. Sounds great. But the little monkey actually has another plan. And, you know, when you decide... You're gonna you're gonna do some stuff that day the monkey the monkey actually doesn't doesn't agree and then and then takes over and then you end up figuring out how many people you could fit into an M&M and You also see the insanity come out in in relationships You know it's all these you know we have a lot of people that are not in really great relationships and your friends are kind of like Why are you why what's what's what's going on and it's it's there's a lot of things going on as well. We have a lot of social pressure to be a certain way and it ends up making us, we often make very bad decisions. So these are the kind of things that I think about, and then I go out in public, and, you know, you probably can tell, I'm a pretty awkward individual, and in public, then, you know, I run into things like the handshake, hug, death clash, when it's amazing, this probably happens, I would say, six times a week, I suffer, suffer through this situation. And I'm on both sides of it. I think it's worse to be on the hug side. You have to go like this. So I talk about that kind of thing. And sometimes I talk about punctuation and how it really matters, how it's a pretty important thing. And email and how awkward that is. And then sometimes when you're leaving and you've had a great time with someone and you say bye and it's all good and then you realize you're walking the same direction. So there's a lot to get into and as I've gotten into it it started off here is the blog at the beginning You know new post every Monday and Thursday you have to be consistent is what you have to understand if you want to have an audience and you want to get You want to you want to really have have a consistent writing schedule you got to have something like that and keep yourself accountable Except then that quickly didn't work and it turned into new post every Tuesday So we got rid of that Monday and Thursday was not happening and then I found myself I found myself missing the Tuesday deadline because remember that monkey? He didn't care what the tagline said, and people would get mad, so I changed it to Tuesday-ish. This is real, by the way. Each one of these things was on my blog for like three months, each of them. Tuesday-ish gave the monkey permission to not even forget Tuesday night, forget Wednesday, like we'll get it in next Monday. Again, people started to get angry, and this is not a joke. This is what the blog currently says. So, at least, now there's no expectations. But the bad side of this is that I'm on week 10 of doing one blog post right now. My insanity has gone completely through the roof without any... you know, hard deadline. The monkey's out of control, my perfectionism's out of control, but the good side of it is it's allowed me to really go into depth in a way I couldn't do when I was posting once a week or twice a week. It's allowed me to really kind of like, you know, learn a lot about a topic and dig really into it so that even people who like this topic, I can learn even more than they know and get into it and kind of synthesize it down. And doing so has gotten me into some interesting new topics like stuff with technology. I've gotten really into tech in the future, also stuff I used to think about all the time before I started blogging. Now I'm trying to dive in and write about some of them. And as I started to write about tech, I wrote about the Fermi paradox, which is why we don't see evidence of any alien life out there, but we probably should. It doesn't make sense that we don't. And artificial intelligence and stuff like that. I got a call one day, and... It was a woman explaining to me that Elon Musk reads my blog, and that he wanted to know if I wanted to do some writing about some of the industries he's involved in. And she went on to explain that she's sure I'm very busy, but that if I have interest in this, Elon would like to set up a time to discuss. So at that point, this is embarrassing, I don't know why I'm showing you this, but this is someone who was photographing this, as you can see. My head actually fell off. You don't expect that to be the person on the other line when you get a call that, oh, it's just Elon. And so it was awkward for a little while, and I had to handle the situation. But I eventually regained my composure and went on. And so I got, I had a chance to go out to California. I got to go to the Tesla factory and the SpaceX factory and talk to a bunch of his executives. And I started reading about all this stuff. And I'm sure, you know, I know that people in Sweden are interested in what he's doing a lot of the time and especially the tech community. And it's awesome stuff. He's a very unusual guy. And so I started. Kind of diving into what why he's doing what he's doing and to understand Tesla You have to kind of take a big zoom out to see the whole fossil fuels era Because that's really what it is. It's an era. There was a long time when there were no fossil fuels being used, and we didn't use much energy as a species, and now we're using all of this energy, and we're kind of running through our trust fund. And then we'll go into the future when, inevitably, we have to come up with some other solution. So we're in this era, and you have to zoom out to see that. And then we talk about SpaceX. People don't get SpaceX. No one knows why SpaceX exists, but it actually exists because he's trying to put a million people on Mars. Whatever your ambitions are, nothing will make you feel with smaller dreams than spending a little time with Elon. But in order to understand why, you have to zoom out again. In fact, you have to zoom out even farther back. You have to get to the point where you can see the whole history of life and understand that becoming multi-planetary. is not just a big deal for humans, it's a big deal for life. It's one of the major leaps, like coming out of the ocean onto land, and you have to understand that there's a lot of potential risk, and that you don't see that in our day-to-day lives, but moving to, but having humans on two planets is kind of like backing up the hard drive, and probably a pretty good idea when you start thinking about it more and more so as I zoomed out working working with Elon I had a chance to work with him a little bit he's weirdly available to me like he would I would I think he's really into messaging and he thought this is really an important thing to get some you know thorough writing about this so I was able to get him on the phone and and it always went the same way like it went in person where with the first you know he's he's an awkward man And I'm kind of an awkward man. So we would start off the call or the meeting with like a five-minute awkward off. It would be like an old-fashioned Western movie with like two people and an awkward off. But then we usually we settle in and then it would be fine and I would end up and I've gotten to kind of ask him some of the deep questions to understand how he thinks and... The thing that I ended up doing is all that zooming out is what I've been doing a lot recently. And what I try to do in general when I'm writing a blog post about the future, about technology, about life, I always try to just, because I don't claim to be any wiser than any reader, I need to just get out of my own head and take a big step back and zoom out and look at things. And all that zooming out... has started to let me in on kind of what I see as a bigger story. That really, because it's all these different topics, they all start to say the same thing to me, which is that this isn't a normal time we're in. That nothing about this is really actually especially normal. And if you want to understand that, and why I'm saying that, you have to just kind of think, if you went back to the year 1750, and you... Got George Washington. I always use him, but who's some Swedish famous person from 1750? That is not easy to say. Okay, let's quickly just do this, right? Who is this? How do you spell this? All right, so here's the dude. All right, so there he is. Handsome guy. All right. He had no idea he was going to be brought into this tonight. So let's just get ourselves back here. Wait, we're going to do this right. If we're doing this, since apparently I've made that decision, we're going to get him into the situation here. Okay. All right, so we have Gustav III here. There we go, much better. So if you went back and you got Gustav, and you went in a time machine, you know, and you go and you get him and you bring him back here. 1750. was not that long ago, 265 years ago, but everything always was in a blackout. It was a blackout all the time. There was no electricity. There were no light bulbs. There's obviously no way to transport yourself other than car and boat. I mean, horse and boat. And it wasn't that long ago, but... almost nothing that's cool about today existed then. So if you brought him here, and he saw cars zooming down the highway, and he saw the internet, and he saw the iPhone, and he saw, and you told him about the International Space Station, and you told him about relativity, and you told him about, I mean, really, any one of these things, airplanes, you can go on and on and on. I think that it would be such a big shock to Gustav III that I don't think that he would just be mind blown or surprised. I think he would die. And so I started thinking about that. How long do you have to go into the future in order to die from how surprised you are about the progress they've made? And I've called that a die progress unit. So the thing about a DPU that is interesting is not that it exists, it's that the length of a DPU is variable. That's the interesting thing. Because if Gustav III went back in time and he said, dude, you're not going to believe the shit I just saw in 2015, and he started telling people, and then he said he was like, that was so fun for them. They get to watch me die. They were all laughing. I want to do that. And he says, okay, you know what? I'm taking the time machine. And I'm going to go back, and I'm going to get Da Vinci. Because he was... 48 years old in the year 1500, about the same problem. He's like, I'm going to go back a DPU of my own, I'm going to get Da Vinci, I'm going to bring him back here and watch him die. The problem is, if he did that, Da Vinci would be surprised. I mean, he would be taken aback with, you know, there'd be some mind-bending stuff about space and physics, for example, you know, the telescope and a lot of things like that. There would be, you know, he'd be... impressed with the commitment that Europe apparently ended up having to imperialism. He would look at the world map and be like, oh, okay, it's a little different than we thought. It would be cool for Da Vinci. I don't think he would die. That's not enough. It's not the same. The truth is, if Gustav III wanted to kill somebody... I think he would have to go all the way back to before the industrial revolution, I'm sorry, before the agricultural revolution, back like 12,000 years ago. If you got someone from then, someone who had just lived as a hunter-gatherer, before the agricultural revolution, there weren't really cities. So sometime before, large groups of people even gathered together. And you brought him to 1750 and you showed him these towering churches and ships that could cross the ocean and all the things like that. I think that that person may collapse and die. So at that point, I think you kind of, the DPU, going back, you need a lot more time to create a DPU. And so you bring the hunter-gatherer here, he dies, he goes back in time and says, well, that was annoying. I didn't get to try that. I want to go back. I want to kill somebody. So he gets the time machine. And if he goes back 12,000 years, he's a hunter-gatherer, he goes back and he gets another hunter-gatherer and he brings him back and he's like... And the guy would be like, yeah? What? You're showing me some stuff. I mean, it was just grass. And so he'd have a real problem on his hands. If he wanted someone to die, I think he'd have to go honestly all the way back. So far, it would be like... 150,000 years ago or sometime before fire. Fire might kill him, but that's a pretty big thing. Or language. Language might blow that guy's mind significantly. But really, you have to go back 100,000 years now for a DPU. The DPUs are getting longer. We call this the law of accelerating returns, and it speeds up over time. And the reason the DPUs get shorter and shorter is more advanced species can make progress faster. So that's a very intuitive concept, but it's kind of crazy when you start to think about looking at this particular length of time, 100,000 years ago. There's a debate about when humans started, but let's say 100,000 years, that's 1,000 centuries. Now, I couldn't even make that small enough on the screen. That's actually like 500 years, that red part. It should be even thinner than that. That's how little a time we're talking about since kind of the Industrial Revolution and the early 1800s. That's what we're talking about. I'll say 1815 to now, there's been some really insane things that have happened in that red thing that didn't happen in any of the other thing. So we're under a billion people on Earth. for that entire time, and suddenly, in the red, we've crossed the one, two, three, four, five, six, and seven billion population marks. That's weird. That's like very not normal. As I mentioned, the fossil fuel era, we went all that time barely using energy. Suddenly, in the last, in the red, we've been digging up. carbon that's been buried since the pre-Cambrian era and putting it back into the atmosphere. God knows what's going to happen with that. We spent that whole time, in fact, this is really the last one century, we spent 999 centuries walking, running, riding horses, and sailing in ships. And in the red, we have cars, planes, and space travel. This is not a normal time. We never had a global... brain of god-like information access and connectivity. That was not part of Gustav III's life, nor the hunter-gatherers. And now we have the Internet. The Internet is so not normal. So what this all comes to is that this is not a normal time. And it's tempting because even if you had a movie where it was an action scene and you paused it. And that frame might not look like anything, and that's a little like it is now because, you know, still the days move slowly and we don't necessarily see what's happening with our eyes. We have to step back and think about it for a second to realize. And in our day-to-day lives, we often just feel like this is normal. Our education system is normal, our technology is normal, the way our relationships work is normal. politics. This is just life. Of course there's countries. Countries were invented. That's a thing of our modern era, the concept of countries. You can go on and on and on. And some of the things I've noticed just writing blog posts as I've zoomed out, we have things like we might go to Mars in our lifetimes. In fact, I have an upsettingly large bet because I got into a thing and then I ended up getting pressured into an upsettingly large bet. So if there's not a human on Mars by January 1st, 2031, you'll know somewhere out there, I'm really unhappy. But the point is, in this century, this one century, we've seen someone go to the moon, and we may see someone go to Mars in this next century. Genetic engineering, there's so many amazing things happening there. We can choose the gender of our children now. You can get in there and actually tweak the genetic code, and that's just going to be growing and changing. I haven't written this post yet, as you can tell. I don't know what's going on there yet, but it's going to be crazy. nanotechnology, the size of the particles that we can manipulate are getting smaller and smaller and smaller. The implications of that are crazy. If we get down to the time when we can manipulate individual atoms, which a lot of people think is only a couple decades away, you'll be able to take a pile of trash and put it into a nano manipulator and a cube of apple comes out. You can feed people. I mean, the things that could do with nanobots in your bloodstream, they can go and repair cells, perhaps, things like that. It's pretty mind-blowing. There's been five mass extinctions, and we are currently engineering a sixth ourselves. That's also, again, five over 3.8 billion years, or really over the last, since animals have been around, 600 million years, and right now we're seeing another one, perhaps. We haven't seen any evidence of alien life, and as far as we know, aliens haven't seen us, but we could be very wrong about that. But this is a time when we are broadcasting all kinds of signals out inadvertently that we never used to. This could be the century that we first hear contact from someone. And so it's kind of an anomaly inside of, because even Gustav III was in an anomaly of his own. Nothing in the last 10,000 years is a normal. So we are really in an anomaly, in an anomaly. in an anomaly, and the question then, you know, if we're here, and it feels like we're here, but we might actually be kind of here. And someone said to me when I was saying this the other day, well, you know, with exponential curves, it always looks like this. If you zoom out, then this part looks flat, and then it continues to look like this even when you zoom out. But the truth is, that comes back to the DPU, where a huge upswing in technology may have taken 1,000 years, and then 10,000 years, and then 1,000 years, and then 200 years. And it may be getting to the point where we start to see a DPU in our own lifetimes. No one has experienced that before. And so that brings me to the question, why isn't everyone talking about this? Every time I started researching, I would say, oh my God, why is everyone not talking about this? It's just, it's clearly the most important thing. And I think the reason is that we're, yeah, I think really life-changing, it seems far away, and we're also very bad at projecting future progress. We are humans, we hunt, we used to hunt, you know, we think linearly. So an animal would run in a straight line, and we are wired to think in straight lines. And so the problem is... When you get to progress, it doesn't work in straight lines. Exponentially, you know, if you think about someone who manages money, they'll always tell you, you know, you want to start putting money away young because exponential growth is a serious thing, compound growth. But we don't, we have a hard time conceptualizing that. And I think the same thing holds here. So you can kind of see here that if you base, what we do when we think about what's going to happen in the next 30 years, we think about the last 30 years. If you say, okay, 1985 to now, wow, so much has happened. We didn't have cell phones. we didn't have the internet and you know all this stuff is new so imagine all that technical all that progress moving forward and even then we still kind of think well it can't be something that big we can't have progress so even sometimes we think it's going to be less of a growth than the last 30 years If we're thinking a little bit more logically about this, we actually might say, well, no, let's actually look at what's happening right now, the growth in the last 10 years. You remember that 10 years ago, social media didn't exist. A lot of things have happened, but if you really want to get it right, you have to think about the growth rate growing as we go. That's why the DPUs get shorter. And so if that's happening, you have to think about the fact that The last 30 years is nothing compared to what the next 30 years is going to be, unless there's some surprising event that slows it down. There's no reason to think that it wouldn't be a much crazier 30 years coming up. And as I thought about this, again, with all these examples, the moon, genetic engineering, or sorry, Mars, genetic engineering, nanotechnology, etc. This is the craziest one. This is the one where, as I was reading it, my eyes were just wide the whole time. But because we're humans, I then finished the post and I was like, blah, duh, duh, I just forgot and didn't care anymore because we just can't wrap our heads around the gravity of something as big as artificial intelligence. So I'm certainly not an AI expert. I'm sure a lot of people here know more about AI than I do. What I did do, though, is I really read the gamut of opinions from the experts. And I wrote a pretty long blog post synthesizing what I read and trying to just wrap my head around it and help readers wrap their heads around it. So I'm gonna spend a little time now just kind of going through the basic deal with AI and what the experts are saying so that you guys can be thinking about this stuff too because I think for tech people it's very important to know what's going on here. So the first thing is just the term AI is confusing. There's a concept called anthropomorphism. which is applying human values, human emotions, things like that to non-human things. So we see the Terminator, we see other things where the robots turn evil and they have egos and they want control, they want power. That's not actually how it works. Computers of any kind, smart or stupid, want only things that they're programmed to want. And they'll never turn on a human... ego and suddenly say I'm gonna ha ha I'm gonna take all the power for myself It's also a very broad category so your phone is full of AI your calculator is AI gmail autofill things like that and so it's hard to understand you know it's a very it's a concept it's hard to kind of again wrap your head around and Finally you know once it works no one calls it AI anymore so no one thinks no one refers to AI in their own lives no one thinks about their car as having AI or their phone but they of course do and so to To understand, really, you have to get that AI can mean something very different depending on the caliber. The three calibers that experts talk about are what I call ANI, artificial narrow intelligence, or sometimes that's called weak AI. That's something that is very, very, very good at one thing, or two things, but it's good at its own narrow purpose. Every single AI in our lives currently is ANI. So that has to, that's, you know, your calculator is a simple example. It's extremely good at multiplying huge numbers, way better than any human. But that's it. If you ask the calculator for dating advice, not gonna be able to help out. More complicated example is like, you know, Google search engine. That's a pretty legit AI. It has a lot going on. It's still completely narrow. To get to the next level, which is what we call artificial general intelligence, or AGI, that's as smart as a human across the board. And that's the key, is the across the board thing. It's easy as humans to underestimate the breadth of our intelligence. And there's no computer that's come close to that before. Finally, you get to artificial superintelligence, which is a computer that is... far smarter than a human, or as Nick Bostrom, who's one of the main thinkers in this area, has said, it's an intellect that's much smarter than the best human brains in practically every field, including scientific creativity, general wisdom, and social skills. So social skills, wisdom, these are not things we think of a computer having. I mean, people laugh thinking about a computer having social skills, and they laugh because we're not anywhere close. But the kind of ASI we're talking about, it's not just, the funny thing is that it's not just as good, It's cooler than us. It has really good social skills. If you and the AI all go for a girl together, it's going to get the girl because you have much worse social skills. And it's much wiser. It'll see the mistakes we're making so easily. Like we would see a three-year-old making a big mistake. So pretty crazy stuff. And so that kind of set up in my head what I think of as a road to super intelligence. And right now, where are we? We're here. We're in a world... full of artificial neural intelligence. There's AI in our cars, like the anti-lock brake systems, our phones are full of AI, half of the apps in your phone use AI. The internet, of course, you know, when you see, you're on Amazon and you see people who bought this also bought. That's not someone figuring that out, that's an AI system. You know that sometimes it's creepy where you like, talk to someone somewhere, or you email them and then you go on LinkedIn, they're suggesting you, you're like, how did you know that I just talked to them? It's like, that's some AI creepy thing going on behind the scenes. There's your flight ticket prices and where the plane lands at the gate. All that is, and I, other sectors, military, manufacturing, finance, expert systems like Watson, who was on Jeopardy. That's an IBM product that's made to be for medicine, it's meant to be a medical expert who can answer doctor's questions and so it's all over the place and it's growing. And it can mess things up, it can melt down a nuclear plant if an AI goes wrong or it can knock out a power grid or crash the stock market which has actually happened in certain, the flash crash of 2010 was an AI. making a mistake and just doing what it was told, but we didn't think about a certain circumstance. But some people think of all these different ANI systems as amino acids, like the amino acids in the primordial ooze of the earth. They're the stuff of life, the amino acids, but they're not life, but one day they woke up. And the question is, all these ANIs, are they kind of quietly, each one putting a brick on this road to something much, much, much bigger? So how does that happen? How do you get from ANI to AGI? Well, the thing that's hard, again, this is not an intuitive concept. This is all very, we have to kind of rework our thinking to understand this. We think, we have a misunderstanding about what's hard and what's easy. So we think walking is easy. and so do dogs, and so do ants. But the reason is because there's actually incredible software in our brains that has been evolved, been programmed by evolution over billions of years to have the kind of motor skills and the balance and the physics to walk. So if we say something like, to an AI, we think that calculus is hard. But calculus is actually pretty easy. And if you think about it for a second, engineering, calculus, this kind of stuff, that's invented pretty recently. So even though we have these great engineers now... All of us are pretty amateur at it. It's not ingrained in our brains to do that stuff. We had to learn it. We had to invent it. So anything like that, that we invented, that's a human creation in the last hundred or thousand years, super easy to make a computer that can do it. And the computer can just beat us right away. It's the things that we evolved over three billion years that are actually much harder. So if we tell a computer, okay, beat the world chess master in chess, the AI is like... No problem, I got it. We have AIs that do that. The AI is like, yeah, yeah, send me down. I'll get it and give me some coffee and whatever. But then we say, okay, also, I want you to read this word. And the AI is like, oh, shit. It's like, yeah, you made it slanty. And that's because we have a sense of what an R is. We know what the essence of an R or a B or a Q. But... An AI doesn't. And if you actually think about this, if you wanted to program a computer, if your job was to program a computer to beat the world master at chess, and you have to think about that, you might say, okay, I can see how that can maybe happen. But if you were told to get it to understand the essence of a B, That's actually much, much harder. That's why these things work on things that are so smart. You could say to an AI, okay, I want you to memorize every street in the world and integrate with real-time satellite data so that you can instantly give me directions from any point in one country to another point, taking into account traffic accidents and construction, and do it instantly. And the AI is like, yeah, yeah, dude, I got it. It's really not a problem. It's fine. And then you're like, okay, also, is this a dog or a cat? And the AI is like, Jesus! And so the AI will look at that thing and it'll explode with confusion trying to figure out if that's a dog or a cat. And granted, that is kind of tough for humans too, I would say. So you can all sum this concept up with what Donald Knuss said, which is AI has by now succeeded in doing essentially everything that requires thinking. but has failed to do what most people and animals do without thinking. Because without thinking stuff is the stuff that's actually 3.8 billion years evolved in us. If you try to, you know, vision is very, very hard for an AI system. If you explained, if both an AI and you look at that, you'll both get it. Okay, yeah, there's, you know, there's some black, there's some grays, and it's all kind of whatever. But if we can then take up, pick up the black part. and this is actually what's underneath, and we say, okay, no problem. I say there's some slats going this way, there's some cylinders, and there's some corners, there's some light from that angle, because our brains are actually doing a really amazing thing. It's very quickly rendering what that is representing in the real world. That is not easy. And an AI is going to see the same thing it saw before, a bunch of two-dimensional gray shapes. It's... It's very hard to get, but when we look at the real world, and so does an AI, it also sees two-dimensional colors and shapes. It just sees pixels. It doesn't see the depth without a very serious programming challenge. Likewise, here we see very clearly a 3D black rock, but an AI sees a collage of white and gray and black in two dimensions. The two major things that have to happen in order for an AI to become human level intelligent first has to do with hardware and So the human brain goes at about 10 quadrillion calculations per second and of course this is somewhat controversial It's you know, but it's around that range and so Where where are the AIs now? This is made in the year 2000. This is you know the whole Moore's law thing which you know of course you guys know about and You know it says every 18 months or so the amount of transistors you can fit on a chip in a certain area doubles And so what you have is the year 2000 you have it at a certain level but actually 15 years later this graph is still accurate And so if this continues on this you'll have an AI that is able to do the computations that a humans can do this is this is by the way also not the most the fanciest computer, it's whatever you can buy for $1,000. So what this graph suggests is by about 2035, or sorry, 2025, an AI will be able to have the raw computational power of a human brain. And then by 2060, if this graph holds, you'll have an AI that can, something you can buy for $1,000 and put in your pocket that has the raw computational power of all humans combined. And so then the other thing that needs to happen, so that part is more straightforward. That part is probably going to happen, and we understand how it can happen. The really, really hard part is the software. How do you make the computer smart? How do you make it understand stuff? It's very different than Google Translate and taking word X over here and swapping it for word Y in the other language. It's understanding language. That is a very different concept. Things like that are, that's what a lot of people are thinking about, and there's no right answer right now. There's a lot of, a lot of people think different things. So some people think that the way to do it is to plagiarize the brain. Basically say, look, we have a really awesome computer that we know and we can see it. And we have a sense of how it works. So let's... try to build a network, neural networks like that, where when a line of thinking in the neural network reaches a successful answer, it gets strengthened. When it reaches a failure answer, it gets weakened. And over time, that system will learn. That's how humans learn. That's how babies learn. The other idea is going to full distance. We don't have the technology to do this right now at all, but whole brain emulation, where you literally take very thin slices of a human brain, scan them all, and then upload them to a computer into a fully functional brain model. And if that really happened correctly, then whoever's brain that was will be, their personality, they will be in their computer. And then, of course, you get to the whole existential thing. Are they really there? Different talk for a different time. But the thing is, that person will say, whoa, I'm in here now. I mean, it'll really be them. That's the craziness of being able to do that. Some people think, well, no, don't plagiarize the computer. who built that evolution okay let's just copy that so they do things like genetic algorithms where they take you know bunch of computers all intentionally built differently and they try to do a task and the most successful ones are then mated by kind of their code being combined and you end up with very Odd and and very uh, but very creative solutions to things we never would have thought of when we do that and so this is another way And then of course finally you kind of say let's not do anything Let's make the computer just do it. Let's like convince it that it's its job and it'll do the work for us so that's kind of when you basically build a computer whose main thing is Researching AI and coding itself and you have a self-improvement thing and so the thing about it is On both sides, this is very far away now. But when we say far away, you have to think about what that means. In the year 1985, our computers could do about a trillionth of what the human brain can do, hardware-wise. By 1995, it was a billionth. By 2005, it was a millionth. Now, 2015, it's... thousand so it's not that we're a thousandth away it's that we're one step away if you think about it like that if this kind of thing shows just the power of exponential doubling and how it can seem like something is far away and then suddenly boom it creeps up on you and there it is software wise it can creep up too because when we're doing things like building an AI that can code itself that can research AI when we're doing this genetic algorithm or this evolutionary thing what you what you what you can do is sometimes you make that tweak that little tweak that it's been in the way the whole time and and and it really unleashes a whole new wave of progress that then can spring into something where the thing can now code itself really well and so at some point you don't know when most of the experts think we will get here and most of them think it'll be in our lifetimes and The question is what will happen then will the AI at that point hang out at human level intelligence? Well, a lot of people think that it's not going to hang out here. It's going to actually have a lot of ways to at that point increase well beyond us. So there's hardware reasons because it can be infinitely, not infinitely, but it can be far, far bigger. Our brain can't be that big because of our skull, but also because the speed of the synapses suggests that if it were bigger, the brain couldn't communicate with itself quickly enough to be a thinking machine. The AI's will build can think so much faster that you could have a computer the size of a dwarf planet before it would have that same problem. So it has the capacity there. It has way better storage. It can hold a lot more information, more reliably. It doesn't forget things. It doesn't misremember things. It's durable. It doesn't age. It doesn't have dementia. And on the software side, It's editable and upgradable and it has the ability, you know, the agriculture revolution was huge for human progress because suddenly we could compare notes and a group of people together could all kind of collectively grow in information. This is that to the extreme because all the AIs on the planet could learn something new and it instantly syncs to all the other AIs perfectly. For so many reasons, it has a lot of advantages, and it shouldn't surprise us if we're here hanging out at a human-level intelligence station, and we see the AI, and it's on its way, and it finally gets to us, and then, oh, oh, bye. And part of the reason that really might surprise us is even when it's getting really close, we might not see it. Because we'll look up here, and we'll see, okay. Oh, wow, an AI can do what an ant can do. Oh, man, it can do what a bird can do. Oh, look, it can do what a chimp can do, and we'll think it's cute. We'll say, ha, that's funny, the robot can do monkey tricks, and we'll be amused. It'll be this funny thing. Even when it gets to what we call dumb human level, we're like, oh, wow, it can do some simple human things, we'll think it's still pretty far away from us. But the fact is, we have a skewed vision of the intelligence scale because we're in it, we're inside of it, so we see the human range as really big. Actually, it's more like this, where once you get to chimps, you are getting real, real close to us. And once you hit dumb humans, you're like that far from being Einstein, because there is a very small difference in the big scheme of intelligence between the dumbest human and the smartest human. So we might say, oh look, it can do simple monkey tricks, it can do very simple human things, and then before you know it, oh wait, it's smarter than everyone. And what if this is the type of computer that codes itself? What if it got there by self-improving? Well, at the beginning, you have a pretty bad computer scientist, because it's a dumb person. Dumb people are not good computer scientists. But it does a few things, and it works its way up. And before you know it, it's an Einstein-level computer scientist. Einstein would have been a pretty sick computer scientist. And what does he do? He makes himself smarter. So now it's far smarter than Einstein. And now it's a really, really good computer scientist. It can make itself smarter. this gets pretty scary pretty quickly. And so what happens? How quickly does that arrow go is a big question. That's called the takeoff speed in the AI world. And this concept of recursive self-improvement where it's a little bit like the DPU concept, how a more advanced species improves quicker. Well, a smarter AI self-improves itself quicker, codes itself better. And so what you can have is something called an intelligence explosion. So if you think about the fact that this is the intelligence staircase, and we're on one step, and you know, a monkey's a few steps below us, and a chicken's a few steps below that. You have to understand how big a deal a step is. If a monkey, monkeys are pretty smart, but if a monkey sees a skyscraper, or a huge tower of some kind, or this room we're in, not only can it not build that itself, it can't even understand that we built it. So there's no way you can even explain that we humans made this. This isn't natural, because monkeys, when they see something big, it's just part of nature. So that means an AI even two steps above us on that green step, not only could we not do what it can do, we won't even be able to wrap our heads around the fact that it did it. We won't even be able to get what it did. That's how much smarter it will be than us. But that's two steps above. Remember, that thing is damn smart now, and it can code itself faster. And if it's programmed to try to make itself as smart as possible, it's going to end up here. And so we're all down there. What the hell? I mean, this is one of those things that it's, there's no way for us to think about how intense a concept that is. You know, we have, in the world we have, we have, uh, stupid is an 85 IQ and really smart is 135 IQ. We don't have a word for 12,573 IQ. Uh, we don't know what that is. Uh, it's, and, and us trying to understand that is like a bumblebee trying to understand Keynesian economics. This is not possible. So all we can kind of say is, well, what does that mean when this happens? You have to understand that we can play God to any species, even the smart ones like apes. We can easily play God to an ant farm. We can do whatever we want with it. Our intelligence gives us power. Intelligence equals power over other less intelligent things. So this thing has so much power that Whether you're religious or not, there's God now. And the big question is, will it be a good God or a bad God? And this is what the AI community is trying to figure out. And a lot of them are pretty scared. So this is one of my worst drawings. I think I did it like five in the morning before I was doing it. But anyway, that's supposed to be a balance beam. And, you know, we're all walking across it. And the way, if you look at 99% of species go extinct. Actually, it's more like 99.99. So what that means is that as species are going along the balance beam of their existence, they fall off, everyone's falling off, and they're all falling into extinction. And it's hard to stay on it for very long. The dinosaurs had an amazing run, and they stayed on it for 265 million years. We've been around for 0.1 million years. So the question is, is this possible? That's the big question. An AI could easily shove us off onto the extinction side, but could it open another side for the first time? Could it open a side where the human body is just a physical thing? It's just cells. It's atoms. It's DNA, which is software, which codes the hardware. If you can change the DNA, if you can fix cells, death doesn't have to exist. Crazy things like that. So you start to get to these crazy questions where you start to think, okay, wait, you know, both of these things could happen with AI. AI could, it's like, you know, you have to imagine that it for the first time allows something to fall onto the immortality side of the balance beam, but it also creates such a bang when something that smart is here that you have to imagine it probably will end up knocking us onto one side or the other. And so then we get to when is this happening? These are all, as I'm going, I'm just like... okay when you know like it's it's you have to and and of course no one knows that either some people think we're here um and other people think we're farther away but what i was surprised is i thought there was going to be a lot of people saying no no this will happen and not for another 10 000 years or a thousand years it was amazing to me that even the people it's like some people said this is going to happen in 20 years. And other people said, no, no, you're totally overzealous. You're underestimating how hard this is. It's not going to happen until the end of the century. And I'd be like, the end of the century is still really soon. You just said that like that was a far away thing. Like I might be alive then. And that's your long estimate. And so there's been a bunch of surveys done. And this is what the median, this is among AI experts who may be a crowd that's into this, but a lot of the AI experts are also the people who say, no, this is far away. So that's the median prediction for when we get to AGI. and when we get to ASI. A lot of us are going to be alive on those red dots. So how will it turn out for us? Who will control this vast power and what will their motivation be? So I kind of set up this scale of where an expert can fall. So on one axis you have, when will it get here? On the other axis you have, will this be good or will this suck? And the people who think that, so I found a lot of people here, and I called it confident corner. Ray Kurzweil is the most famous of those. Peter Diamandis is another one. A lot of these people that are, they, and these are very, very crazy intelligent people who understand this better than I do and better than most of you do. And those are people that say, no, no, this is gonna happen sooner than anyone realizes and it's gonna be awesome. And these people are pumped about it and they think, you know, Ray Kurzweil is pretty sure he's not going to die. And he's like 65. And people think he's wacky, but he's a really, really smart dude. And so if that good outcome happens, there's another thinker in this named Eliezer Yudkowsky who said, there are no hard problems, only problems that are hard to a certain level of intelligence. Move the smallest bit upwards. And some problems will suddenly move from impossible to obvious. Move a substantial degree upwards and all of them will become obvious. So you think about, again, a monkey trying to turn on a computer and it's hitting it, you know, and we can just go over and press the power button, not hard. So what are we doing? We're trying to, you know, come up with cancer treatments and we're like hitting the computer and AI might be like, oh, God, no, I got it, you know? And so, and that, again, that's, we're talking about monkey. That's two steps above us on that staircase. We're not talking about the insane... 100 steps above, either, when we even get there. And so all these things, hunger, disease, death, climate change. Climate change is, you know, a laughable problem for the AI. They'll very quickly figure out some way to extract carbon, to control things, and, you know, this is my human brain trying to come up with the way it would do it. So then there's a whole other crowd of people. They're not here. They're here. I call it Anxious Avenue. Those people range in when they think it'll happen. Some people think it's going to happen... Very soon some people think no probably farther away, but and and they're not there's also You know catastrophists who believe this is going to be very very bad. We're all gonna die Those people tend to be more ambivalent they say I don't know if it's gonna be bad But I don't know it's gonna be good either and what they do know is that once it's there We can't take it back if you know monkeys evolved into us monkeys can't take the whole human thing back They can't be like oh yeah, actually you guys are not pleasant for the rest of us Let's just undo that command Z, and no reason to think this is any different. This is not a take-backable thing. That's why Elon Musk, who's very scared of AI, and he kept bringing it up even though I was trying to talk to him about the posts I was doing, he says that he calls it, we may be summoning the demon. And it's because we get to the two scariest words in the English language, existential risk. And so this is a Nick Bostrom chart, which basically talks about how, you know, you can have, in terms of something bad happening, you can have an imperceptible level of badness, or something that's endurable, terminal, or hellish, which is a fun word he throws in there. And then how big is it? Well, is it personal? Is it just one person? Is it local? Is it global, the whole planet? Or is it transgenerational, where it doesn't just mess up the current planet, but like, you know, for all the future. And then we get to cosmic. So existential risk is when something spans, it covers the whole world, it spans generations, and it sucks a lot. That's when you have an existential risk. So an asteroid, of course, is an example. Nuclear weapons used to be thought of like this all the time, maybe a little bit still. Biological weapons, things like that. But the more I read, the more I'm like, none of that is the thing. AI is the thing. And unlike nuclear weapons, there's a potentially really good side to AI, which is why it's such an enticing topic. So if it's bad, why is it bad? Remember I said, it's not the Terminator saying, it's not a robot saying, ha ha, now I have the power. It's not biological. It doesn't have the need to reproduce and to self-preserve. It's not like us or any other life. You know, you have to think about, it still is what it is. So it doesn't care about itself or me more than this laptop cares about itself or me. It just doesn't care about anything. And, you know, if you took a spider and you made the spider super intelligent, just terrible thought by the way, the spider's way smarter than us, it quickly sizes up and knows exactly what to do with us, but in its DNA, in its core, it's still a spider. It still wants what it wants. And so, this is still a computer. And so, is it evil people? Do terrorists get a hold of AI? Is it some country who gets a hold of AI and says, now we're the next superpower and we're the only superpower? Is it, you know, again, the evil robots, which, again, I don't think, and most of these thinkers don't think it is. I think what most people think it is, because I think that once it gets smart enough that it could do something really bad, it's going to be very hard for people to control it, for anyone to say, this is mine and this is my weapon. I think most people's very, the anxious avenue crowd, they're scared of, the first AI being programmed carelessly. And remember, the only thing it wants is what it's programmed to want. So some people talk about what's called the paperclip scenario, where you have an AI and you say, okay, look, we need to just get this thing smart, and it's going to code itself to get smarter and smarter. And the way we'll test it is we'll give it a task. We'll say, hey, take all this raw material and turn it into paperclips. And as it gets smarter, we'll get to see how much more efficient it gets. And we'll say, you know, just make as many as you can. And that's great until the AI crosses that barrier and becomes super intelligent. Then the AI, you know what it starts to say? Okay, I need to make a whole shitload of paper clips. Now, what's in my way here? Well, people, because I can tell that this wasn't their initial plan. They're going to shut me down because they don't want me to make paper clips because it knows everything. It's much smarter than we are. And it says... and all it cares about is paper clips. It's obsessed with making paper clips. And it says, okay, so people got to go. So it very quickly figures out some way to extinct all of us. Bad day for everyone. And then, you know, 50 years later, it's the entire solar system, Mars, Jupiter, it's all paper clips. Because this thing has created space travel and all these drones, and it's converting all matter in the universe. into paper clips. There's the singleton concept where, you know, if a bunch of different people are working on this or companies, the one who gets there first, again, if we wanted to shut down, if monkeys were starting to work on getting human-level intelligence and we thought that was going to be a threat, we could just shut it down. We're smart. We can cage them. We can stop and take away their stuff. You can zap them with a taser. We have all kinds of ways to... to have power over something less intelligent, so the AI who gets there first will say, oh, no, no, I'm not going to have competition, and just quickly shuts down all other efforts in ways that we can't even understand. You can't just unplug it any more than a group of spiders, to use another spider example, it wants to cage a human, and it says, ha, okay, or it wants to kill us, and it says, I got a plan. We won't let it get any food, it'll starve to death, so we won't let it create any spider webs. We kind of say, okay, I'm going to go to the grocery store. Let me get food there. Bye. And then it's like, you know, we can't unplug it. That's a human solution. We'll say, ha ha, we'll unplug it. And it says, no, no, it's figured out ways to get energy through some wavelength thing that, you know, that we can't understand long ago. And this is what makes Nick Bostrom worried, because what you have is you have lots of companies working on this. The most famous is Google. And Elon is nervous about all of this. Google is who he thinks is in lead here. They just bought a company called DeepMind, which is apparently one of the major, major front runners in AI research. Of course, Facebook is working on AI. Apple is working on AI. You have all these companies, and there's also small stealth companies secretly working on AI. And it makes Nick Blastrom feel like this is a bunch of kids playing with a bomb. That it's a bunch of people, and the thing is that the people working on it tend to be the confident corner people. That's why they're working on it. And that's the kind of people, it's the entrepreneurs that are drunk with their own visions of their own glory and of changing the world and of creating the coolest technology ever and solving every problem, that they don't think very hard about the potentially the bad side. And it gets to... you know, they start to ask the Anxious Avenue people, well, how do we code this safely? And a lot of them don't know. Because you think, remember, it wants what it wants, but it's not a human. So it doesn't just understand what we mean. It doesn't care the way we care about what we care about. So when we say, hey, make all humans smile. The next day, every human has permanently paralyzed their face into a smile because it, like, went into our DNA and it changed that. And we're like, it's not what we meant, you know. Or make all humans happy. And it then turns all humans into drooling, smiling vegetables that can't think, but they're very happy feeling. Or they say, no, just do good for the world. Just do good, whatever good means. Do good. And they say, okay, well, you know who's killing lots of things? Humans. Let's get rid of them. That's going to do a lot of good for the world. And we're like, no, that's not what we meant. So this is not easy. And you have all kinds of people thinking this gets into philosophy. How do you code an AI? when it's not intelligent, so that when it becomes super intelligent, it really gets it. Again, this is one of those things like reading the slanty writing. It seems easy to us. It's not when you're coding a computer. And so that's why they say, either way, it's the last invention we'll ever make. Either we won't need to invent things because it'll be doing it, or we won't need to invent things because we'll be extinct, and extinct species don't invent things often. And so... Back to the beginning, this is just kind of one of the things. This is the big deal, but we're also going to maybe be going to Mars. We're also going to be able to code our DNA differently. I mean, crazy things are happening right now. So it's cool that you guys are all in the tech community, because that's really where this stuff is all happening. And so, yeah, I think that's basically what I had to say to you, and I hope you make it good. Don't make it bad. Okay, this is all in your hands. And I don't know if we have time for any questions or for, but if we do, we can do those. Does anyone have questions? Yes. So I think what we should do is if people do need to go somewhere, they can, and people want to stay in questions, we can do 10, 15 minutes of questions for those who are interested. Does that sound good? Yeah, okay. And I will be the mic runner. So who has it? Thank you. Yes. So who has a question? Who would like the mic of, and if you do want to leave, please do. Who has got a question? There, okay, I actually have to be the mic runner. Can you have your hand up, I think? Yeah, good. Questions can be about anything. Doesn't have to be about AI, by the way. Hi. Hey. Hey. I was thinking what you said about once it gets, if there are several different competitions and there's gonna be one AI and once it reaches the threshold and becomes superior intelligence, it's gonna shut down everything else. Like all other attempts in ways that we can't even imagine how it does it. And then if there are several different, if it's going to have godlike powers, then there's presumably going to be some kind of arms race that different forces are trying to get this and then try everyone else to shut down their places, preferably with propaganda that this is very dangerous so you shouldn't be doing this. How do you tell the difference if it's somebody who's trying to develop their own and make everyone else shut down theirs and it's actually an AI that's trying to shut down everything else and look normal? How do you tell if it's actually an AI doing it, not a person? That's one of the things that I read about is that a lot of them think that the AI is actually gonna be like a sneaky dick about it, and that what it's gonna do, what I kept hearing is that it might do this thing, and again, first you're like, no, but then you're like, well, yeah, I mean, that's what I would do in the situation where the AI is going to fake dumb. So a lab will be making an AI, and they'll be like, huh, it's just not working, and the AI gets to a certain level, and it's kind of like, okay, wait. don't, just play it cool, fail that test again, play it cool, but don't fail it enough that they give up on you, because they'll unplug you, just chill, and it slowly starts coding itself, unbeknownst to us, and then one day it's like, ha ha, ta da, like, I got you all, and there's nothing we can do then, it takes the key and we're all in the cage. So, it's not really answering your question, because I don't really know, like, I don't know how we would know, and the truth is, You get to this thing, what should we do? Well, I know a lot of people think that there should be regulation. Anything that's a serious threat to public safety has government regulations, but you don't usually have the government restricting tech progress, so it's very difficult to figure out how to handle this. There's also the thing, well, what if we're regulating ourselves, but China's not, or Saudi Arabia's not, and you start to think, okay, well, we should be working on it too. I don't know the answer. I'm happy it's not my problem, and let's leave it at that. Okay, there is one thing that I don't understand in this whole discussion, because I've heard this reasoning before, like if you're gonna make an AI, it's gonna have huge consequences, so whatever you do, make it good. The problem is that the people on this planet who create stuff... create stuff in different way and you assume that we can create stuff good or bad but you also know that what we create will be able to create its own so why wouldn't there be enough diversity in the sort of AI life form that it won't matter if we make it good or bad because it will make both on its own well I mean that question very quickly bleeds into anthropomorph anthropomorphizing because you just start to say, well, it has to have a motivation to do something other than what we programmed. And there's not really a good reason it necessarily would. So a lot of people say, well, why can't it just undo its coding as soon as it's smart enough? But again, but why? Why does it want to undo its coding? It is its coding. Its coding is what it wants. And so... I think if it's not carefully programmed, the AI can tweak, if it has a messy program, so it has a certain motivation, but then it has kind of like a motivation on top of it that Isaac Asimov has something where it's do things that are good unless it's gonna hurt a human. So if you have something like that and it's messy, then you could have the AI say, whatever's the core thing could quickly say, this other thing is inconvenient, and I'm gonna recode that. But as far as... The concept of an AI getting smart enough where it suddenly says, okay, well, no more of this goal. I don't need to work for these people anymore. That's not necessarily what will happen. It supposedly will keep its own core, just like our DNA. Humans, we do all different things in the world, and we're all different personalities, but at the bottom of it all, it's mate and self-preserve. That's the core coding in all of us, and it'll feel just as strongly about that as we feel about not jumping off a cliff and dying. Do you believe we are just a host of technology? That we're basically part of the evolution to just push technology into its own being? Are we the biological boot drive for AI? Yeah, yeah. Kind of seems like we could be. It really got into my head when I was doing this writing when I was thinking, yeah, what if AI is the real thing that's out there in the universe? And the way AI is born is from this gross, simple, biological animal species that has blood and it's kind of gross. And that life begins with biology and then eventually biology passes the torch to the real thing. Or life begins and its progress works through the mechanism of evolution. Evolution is what death. is a filter, so the concept of everything that has evolved is because death has filtered it. And maybe this is the end of death and evolution as the main driver of what happens, and this is something new now, where AI, which can code itself into something much smarter, maybe this is the end of the era of evolution, for example. So this is the kind of thing that you can think about, and if you think about it like that, then yeah, we are The pattern might be, this gets to the Fermi paradox, by the way, which is again, where is everybody? I don't see any evidence of alien life, and it's a fascinating question with all kinds of different potential explanations, but one of the big ones is that there's some great filter that no attempt at super intelligent life can make it past. And there's something that kills off attempts at intelligent life And that either is behind us, like maybe the hard thing is going from single cell, sorry, from simple cell to complex cell, the nucleus. Maybe that's the one in a trillion thing that happened to occur here. And we passed the great filter. Life is, the universe might be teeming with prokaryotes and single cell simple organisms. And this is the anomaly. Or maybe it's life beginning at all. Maybe the great filter is that it's very rare, and we've never been able to replicate. the beginning of life in a lab, and it took a billion years here. Maybe that's the one in a trillion, or maybe all of that is very common. And there are millions of examples of life just like ours in the Milky Way and in the universe, and there's all kinds of species that are our level of intelligence, and maybe a bit smarter, but nothing beyond that. And that's why anything that would be to the level where we could see a lot of signals, which is a level beyond us. maybe that they don't exist because there's some great filter. And so a lot of people think that if the great filter's ahead of us, which is the unfortunate circumstance, then it is because of AI or their technology kills them. So it's very plausible what you're saying. And in a lot of ways it makes sense. And then it made me not want to be a person anymore. I was kind of like, why am I an animal? That's so annoying. dumb to be an animal. It reminds me of riding a horse instead of having a car. I don't want to have to eat and die and get tired and all these things. I want to have my consciousness in something like a machine so I don't have to deal with any of that. It's very annoying, actually. What? No, I'm not. I do not. You can tell, huh? So I had a question about the people that you talked to. Andrew Ng, who's like one of the more famous machine learning researchers in the world, recently did an interview and he said, I don't worry about AI taking over the world for the same reason I don't worry about overpopulation on Mars, which is, it's not a problem, right? So given that he says that, I'm curious how many people you talked to just said, like, there's some sort of leap. from where we are now to the general strong AI that's just not something that we can see or not something that's even predictable? I mean, honestly, almost all of this is unpredictable. So there's no opinion I heard where I had the confidence to be like, well, that's definitely wrong. Because, I mean, anything can happen with this. And you can just tell when you're researching a topic, you know, I call it like, it's like the... It's like the... You know, you look back on history and you're like, wow, they thought Zeus through lightning bolts. They thought that's what lightning was. They were dumb. Or when you're like, wow, they thought the earth was flat or they thought the earth was the center of everything. And they seem so in the dark. And we see, you know, we have such a clear vision of what they didn't. And then there's times when you're researching something when I'm like, shit, we're the Zeus people right now. Or we are the flat earth people, I can tell. And you can just, you can, you can. When you're reading the Fermi Paradox, again, is one of these things about why there's no alien life. You'll have the smartest people in the world completely contradicting each other, contradicting themselves, disagreeing, wild guesses, and AI is even more like that. You have all kinds of people with strong opinions on one side and on the other side, and it's a mess. And it reminds me of what it must have been like when they were discussing whether the solar system was heliocentric or geocentric. The viewpoint that you just expressed, which is the most, if you go back to here, this is the viewpoint that it's down at the very bottom, almost definitely not soon. I found very few people with that view. There were some, but for example, in this survey, in the surveys, in the survey that brought these answers, they interviewed people from across the whole spectrum of this. A lot of skeptics and a lot of more bullish people. And I think 2% thought that it would never happen. And maybe something like under 10% thought that it would happen after 2100. So I think that the majority of people that really know about this seem to think that this is a 21st century thing. And so the overpopulation thing on Mars, I just don't see a good reason to feel conviction about that, to feel strongly about that, when there's so much we don't know, and so much of this is unpredictable, and you look back to, if you ask someone in 1900 to predict, remember the DPU gets smaller, so ask someone, go ask George Washington or Gustav III, predict 2015. It's gonna be... completely off in so many ways even in the 1960s if you said predict this they'd say well we're all on mars there's people on on on pluto because we're you know just gotten to the moon this was the beginning they'd be completely wrong about that and this would blow their mind the iphone would totally blow their mind so it's there's just no reason to think that we should be better at predicting especially when things are moving so quickly especially with something as confusing to us as artificial intelligence so it's a valid viewpoint But I certainly am not going to say, well, you know, that guy's right, therefore we shouldn't worry about it. Okay, this would be the last question down here. Do you think there's a limit to AI itself as well? So you talk about the skull and that AI could have a planet-sized brain, but we know that the speed of light is fixed and information cannot flow in the universe faster than that. So there must be a limit to intelligence itself too. Yeah, I think that's perfectly plausible. But if we're talking about, that's like, you know, we're running, we're jogging, and we're trying to build a really fast jogger. And we say, well, there's got to be some limit because they can't go faster than the speed of light. And that's fine, but that's really fast. And if it got, and so I guess the point is, if you were talking about that, the speed of light wouldn't even be the... the major concern because there's so much room between the speed that we'd be going now and the speed of light. So I would say the same thing, where I'm sure, because everything, the speed of light is a hard limit that affects all kinds of things. There's some kind of like a plank volume at the smallest type of thing, so there's certain limits that are on the extremes. But that said, that's so far away. The limit on intelligence once we get there it's almost i just think at that point it's irrelevant i could say i would say there's a good chance there is a limit but uh you know it's not going to really be um be something that changes anything as far as we're concerned i would say okay wow yeah look at how many people stayed for questions what look how many people

Tim Urban of Wait But Why speaks about Artificial Intelligence and the Road to Superintelligence at Oredev 2015, inSweden.