What if AI keeps getting smarter?

0:00I don't know if you've been paying attention, but there's been a lot of fear around the idea of how fast AI is continuing to advance. What if it continues to get smarter? What if it gets to the point of not only AGI, which is artificial general intelligence, but what if it gets to the point of ASI, artificial super intelligence, where it's smarter than every single person on the planet Earth combined? What then? Will we be enslaved like some sort of Terminator movie? Will it be like the Will Smith movie, the Iroot where deep down inside they're plotting to destroy us and eventually the robots will take over? What kind of future can you predict? Well, that's kind of the problem. That's really the reason why so many people are getting so scared.

0:55You can't predict it. the same way that a rodent wouldn't be able to predict the various ways that humans might try and trap it to get it out of their house. That's honestly even a a smaller scale than what we're dealing with if we end up dealing with ASI. Having artificial intelligence that is capable of things beyond our imagination means we can't imagine the different things that could happen. And sure, we can technically imagine, but the type of answers that you're going to get as a human, just a regular human. Well, let's look back at the the rodent that's trying to figure out how um a human might try and get them out of their house. Well, they they might try and bite us cuz, you know, that's what we do.

1:54We bite things. So, they're going to try and bite us. Maybe they'll I don't know. I'm obviously not a rodent and I can't think of the various ways that they might try and predict our movements. But I think it's I think it's uh easy to say that they wouldn't be able to predict the different types of traps that we'd be able to come up with, which is honestly how they're able to get trapped in the first place. If we have artificial intelligence gets to the point where we can't control it, where we're basically screwed for a time. It can be some type of, you know, amazing future where everybody is living in abundance. Anything that you could want, anything that you could desire, bam, you got it.

2:52All thanks to this artificial intelligence. But what happens when one day they realize we we don't need these people? And the counterargument that you know people have is well why would they why would they even think that? Why would they think negatively about people? You know, they they are super intelligent. Surely they don't spend their intellect wasting time on how to be evil towards creatures that honestly couldn't do anything to them. Surely they're smart enough to realize that that's just a waste of time. That's just a purely evil act. But uh again, it's it's hard to predict.

3:49We're just rats trying to figure out. The difference is rats aren't responsible for humans. At least that I'm aware of. You know, where do we come from, right? But it's not like rats were coming together and they were forming the human species. the way that we are creating sort of our own species with artificial intelligence. You know, in a way we are creating many like demigods, you know, it's not completely godlike, but the things that these AIs are already able to do, it's there's so much where it's it's beyond human capabilities. And if you don't believe me, watch a drone that is powered completely by artificial intelligence. This is the future where we rely on these things.

4:51But what if one day, you know, kind of like that Terminator movie, these things decide to turn against us? We going to be dealing with the Skynet or whatever it was called in the movie. Who knows? But uh I'm going to pull up a video that really inspired this uh the topic of this conversation today. It was uh it's called what happens if AI just keeps getting smarter. And shout out to them, you know, subscribe to to their channel or whatever this is. Uh I think it's a pretty dope video and looks like refined animations. So yeah, shout out to them. Let's see what they got.

5:36In 2023, Nobel Prize winners, top AI scientists, and even the CEOs of leading AI companies signed a statement which said mitigating the risk of extinction from AI should be a global priority alongside other societal scale risks such as pandemics and nuclear war. But how do we go from chat GPT to AIs that could kill everyone on Earth? Why does so many scientists, CEOs, and world leaders expect this? Let's draw a line of AI capabilities over time. Back here in 2019, we have GPT2, which could answer short factual questions, translate simple phrases, and do small calculations. Then in 2022, we get models like GBT 3.5, which can answer complex questions, tell stories, and write simple software. By 2025, we have models that can pass PhD level exams, write entire applications independently, and perfectly emulate human voices. They're beginning to substantially outperform average humans and even experts. They still have weaknesses, of course, but the list of things AI can't do keeps getting shorter.

6:34And the whole being able to replicate human voice thing is is real. Like my girl just got a call the other day and she was sitting there. I forget exactly what she said, but she was like, "Why why are they why are they acting like that? Why are they talking like that or something?" I was like, "Be that uh that person was not real. It's artificial intelligence." And there's just very slight subtleties that give it away. Now, and maybe that was just the version that that company happened to use when, you know, doing random calls. But to my girlfriend, it was a real person that just maybe didn't have the social skills that a typical person would. Like, what? What? Why did you just randomly abruptly start talking? Like, you kind of cut me off a little bit. These little nuances that make the subtle differences now are going to continuously disappear.

7:36It's just going to be gone. The um the the fact that it's able to do all of these research and papers and all that, it passed the bar exam, you know, like if you want to become a lawyer, it passed that. It's beating world champions at chess. It's creating novel or not creating, it's coming up with novel things as far as science and you know advancements in technologies, things that we haven't even thought of, avenues that we haven't even gone down, didn't even think were a possibility. That's now. And like I said, 2019, that's when you had the GPT2 or whatever. It's only been a couple years. Like, it's only been a couple years. I remember making fun of the the images that AI would come up with, watching Will Smith eat spaghetti and be like, "What what the is happening?"

8:45But even then, as soon as I saw a second version that had gotten better, even that's even even then I knew that this was going to be crazy. Even then, I was one of those people that was very quickly on board with all of this, saying like, "Wow, this is going to be huge." And I was late to the party. artificial intelligence. People have been working on that for a very long time, longer than we would think. But because chatbt came out and really just flooded the markets, I think I even heard that more people signed on in like a shorter amount of time than like any other app's history or something like that. You know, fact check me on that one. Don't uh don't quote me exactly. But regardless, the numbers are astounding. even if it didn't break records, which I think it did, but the amount of people that just swept to it made it seem like it just came out of nowhere.

9:44But this has been something that people have been working on. There's been things that have been work working um like algorithms that are artificial intelligent based or whatever. What do you think uh text predict prediction is when you're going to text somebody and your keyboard thinks about the different words that you may want to want to write down? people a lot smarter than me said that um they said that that's that's all that LLMs are the large language models the AIS that's all they are they are predicting what the next word is like each word they just continue to predict what the next word is which it doesn't that doesn't make sense to me that's why I get again like people a lot smarter than me in multiple videos and I've seen explain this and it's just like yeah it's just constantly predicting what the next word is and what the next thing is. It's just like what how how do you do that?

10:43Because I'm thinking, at least the way I'm experiencing it, I'm thinking of in like entire sentences or paragraphs or entire ideas and then I put it out. You're telling me every single word that these things are generating is just a guess, a prediction. And it's gotten really good at it. even when it's emulating other people, other writing styles, whatever type of writing it is, whether it's poetry or writing a research paper or passing the bar exam, and that's right now. It's only going to get more and more crazy.

11:21What happens if we extend this line? Well, we'd see AIs become more and more capable until this crucial point here, where AIs can design and build new AI systems without human help. Then instead of progress coming from human researchers, we'd have AI making better AIs and the line would get a lot steeper. If we keep going from there, we hit this point where AIs are super intelligent, better than humans at every intellectual task, better than all of humanity put together, spontaneously producing breakthroughs in research and manufacturing, responsible for most economic growth, completely transforming the world. And if we keep extending the line, eventually we could get to this point here. incomprehensible machine gods which tower above us the way we tower above ants seemingly able to reshape the world however they want with or without our involvement or permission. If we do reach this point, probably humanity would go extinct for the same reason we drove so many other species extinct. Not because we wanted to wipe them out, but because we were busy reshaping the world and it

12:17That's a that's a fair point. So, I guess that's what I was talking about earlier where why if you're super intelligent, why would you even waste your time and energy on wiping out a species unless they were a direct, you know, threat to you? And at the point of super intelligence, that's that's like ants being a direct threat to us. This solves for that. Think about every single time a house is made. Think about how many trees get cut down, how many different animals and insects lived in in that ecosystem. And think about all of the animals or, you know, I guess animals too, but insects in the u in the actual ground when they start to dig. Think of any antills that might be around. Think of how many worms and like that. There are so many different things that get completely displaced all because well, we need a house and then multiply that house by every single house and every single skyscraper and you know what I mean huge stadium and roads and all of that We talk about we talk about how many trees get cut down, you know, because of human intervention because humans want more space.

13:39But the amount of animals, I guess that gets talked about too, but the amount of insects, we are super destructive. And I mean that goes for anything. Like yes, we obviously eat animals. I'm not a vegan by any means, but I mean even think about if you are a vegan and you're getting all of this all of all of these vegetables and doing that all of through the agriculture and the pesticides and you know build uh digging up the the ground and making sure that no bugs come around. Think of how many insects and animals get killed during the production of vegetables and fruits. We I mean it's the circle of life or whatever. It's it's mother nature, right? That's kind of how everything is. It's just a lot scarier when you're not the apex predator.

14:38Meaning, you're not the one at the very top of the food chain that doesn't have to worry about any of these other animals. And yeah, an animal could come and attack you and you end up getting killed by like, I don't know, a stampede of elephants or some So, I'm not saying that any animal lower down on the the whole totem pole, so to speak. I'm not saying that they can't kill you, but the amount the amount of change that we inflict on the earth out of sheer will and desire, it's it's astounding. And we we have we have changed weather patterns. I think I think um I think chi was it China was was responsible for for like moving the earth or some and made the day a little bit shorter.

15:34It was some crazy like think about the the nuclear explosions that we just be putting out into the ocean for testing and not even you not even thinking about all of the fish and wildlife that are there. We we're so destructive. We inflict so much force on this earth. And we're creating something that is going to be like all of us combined and more that will be self-replicating. We'll be able to create better versions of itself. And I think there was even there was even a scene in iroot where Will Smith was was coming. He was like doing a tour or some And you know they're like, "Oh yeah, and this is where we have the robot assembly line where you know you see robots creating other robots and this robots creating other robots.

16:30That's just stupid. That's that's no human oversight." And of course his character in the movie already had a distrust for robots because it made a calculation when he was drowning um in some water over a bridge or something. And it was him and a kid that had just been in a car accident or something. And the robot had to decide which one to save. It couldn't save both. And when making this calculation, we realized that the adult male had a higher chance of surviving than the child. U the child. I think it was like a female, like a little girl child. Not you know what I mean? Not that it matters. It was a child. Smaller lungs, heart wasn't able to be I don't I don't know was running through the the damn robot's brain. But it said, you know what?

17:25There is a x amount of percentage higher that this human will be able to survive and it pulled him up. So he had a distrust for robots. that calculation, the fact that I was right there saying like I don't even know how it was able to come up with that determination. And it probably had a whole bunch of factual information to back it up by, but anybody with the heart, man, the guy. Save the kid. If you if you got to choose, if you got to choose, got so much more life. like most decent people would would say that that the kid is is innocent. The kid deserves to live. the the percentage difference or whatever. Take the chance. But we're not able to predict the way that artificial intelligence is going to think.

18:31And I know that was just a movie, but the same concept applies now in real life. Now that this is a real thing that's happening now that robots are actually in development. That at one point or at some point rather there is going to be more robots in the world than humans. And there's already more guns in America than there are Americans. But guns aren't walking around. Guns aren't able to pass bar exams and create, you know, research papers and, you know, do digital art and videos and do all of these different things that artificial intelligence is able to do. Guns aren't walking around and blackmailing its developers and lying and cheating in order to win games and avoid detection from any other fraudulent activities. Guns aren't doing these things. Guns just go into your pocket. And yes, it's, you know, other countries look at America like, "Oh, look, that's them people with just a whole bunch of guns."

19:27Yeah. What? Okay. Yes, I get it. Guns kill people. Super intelligence, though. At least at the end of the day, a gun can't pull its own trigger. And there's a reason why I'm comparing a weapon that can be used for a whole lot of good, but it's also used for a whole lot of bad. It's there's a reason why I'm comparing that weapon to robots and artificial intelligence because the day the day is is coming soon. that future is near where robots will truly outnumber us there. It's not like there's they don't exist. It's not like I'm talking about something that isn't out there. If you haven't seen there are robots currently now walking, not amongst us, but soon it's going to be very normal to have one doing the dishes, doing the laundry.

20:33This is already in development. There is like protest. You can go find videos of robots already doing these tasks. It all sounds good until you see them making calculations that we just don't understand. And even in that movie, they weren't like super intelligent. I think the main the main like bad robot that was controlling all the other ones was super intelligent. But imagine if each one was its own autonomous super intelligent being and they're able to tap into a hive mind type of thinking and collectively put their their will onto any sort of task that they see fit. But they are also individually dangerous. That they all could have their own individual desires.

21:40The same way we have tyrants and you have crazy people rise to power like Hitler or Mao of China. You have these people that rise to power and inflict so much pain and suffering. and death upon planet Earth. What happens when some of these tyrants and monsters that rise are super intelligent? I don't know. Some sometimes I guess a part of me is scared that we are going to be those ants that get completely annihilated just because well we need a home. We had to build somewhere. Sorry, but not sorry. It is what it is.

22:34Wasn't worth our effort to keep them around. When would this happen? We're not sure. Are there other ways the future could go? Certainly. But if we keep building more powerful AIs the way we're doing it right now, then this looks like the default path. And this default path is terrible for humans. Why? Because we're great at making more powerful AIs, but we haven't yet figured out how to make them do what we want. We've made a bit of progress. AI assistants are usually resistant to things like explaining how to make bombs, but they're not that resistant. And so far, we've just been trying to control things weaker than us where we can catch when they try to lie to us. But once we hit this point, that stops being true. and the problem becomes much harder. In fact, nobody knows how to do it. That's why experts including Nobel Prize winners and the founders of every pop AI company have spoken out about the risk that AI might lead to human extinction. The

23:21But the thing is, game theory requires that everybody continues to march forward. Think about Elon Musk being one of the leading people voicing against AI, saying how we need to slow down. We need to think about safety. We need to worry about the future of mankind. And now we have Groc. Well, if you can't stop the race, join the race. Try to win the race and make sure that your artificial intelligence is at least the first one. But does it does it even matter? Greetings. We'll be right back to the show after these messages. We're all constantly inhaling tiny flakes of meteor dust with every breath and there's nothing we can do about it. And that's why this episode was brought to you by Hell of a Ride by Exo Zudit. Now streaming everywhere you listen to music in Scene.

24:27Tell me lies. Make me feel like I was never good enough. Yeah, baby. Just tell me it's nothing. Tell me you wouldn't. Yeah. Tell me again how I know you love me. You scream, "Baby, don't you trust me?" Yeah. So I lie instead of course. Why you hide when I ignore or we had a hell of a ride. You must think it's harder to tell me goodbye then. So tell me these lies. You probably believe it's your only option. You might be right cuz I know you love me that change is nothing. I knew the down was coming. I know the truth but baby just tell me your lies. Does it even matter if right after there is other companies that come up and create their own super intelligence and by the time it's actually super intelligent is it anybody's can you really own something?

25:28Can a company own super intelligence? Well, if you were super intelligent, I doubt you would let yourself be owned. I think the fact that we can't slow down and that essentially means that we won't slow down is the scariest part. I I I basically think it's a feudal attempt when I see people on their own crusade to warn people against slowing this down and saying that we need to regulate against this. Let's try to just like at least extend the lifetime of this. Let's let's enjoy the artificial intelligence that we have instead of trying to continue to push forward and make it smarter and smarter and smarter.

26:30Again, game theory meaning, well, if we don't do it, they'll do it. And if they do it, they'll have the upper hand. And just like that, they win. And man is a political animal. We are so political. We just want we just want power. We want dominion over others. We want to make sure that we're the top. It wasn't always America at the top. It was fill in the blank with any other country or empire or whatever that rose and fell. Everybody's trying to make sure that they can be the top dog. And if you're not the top dog, you're trying to do everything you can to be the top dog. And how much of a game changer is artificial intelligence going to be if you really can't have a swarm, a fleet of drones coming at you powered by artificial intelligence?

27:28What are you going to be able to do if you don't have that? Well, countries will be able to just walk through others and be able to take over and all of the nicities, all of the the peace that has happened, it's only there because of mutual assured destruction. What about when you hit that big button and your nukes go off and the other country already has super artificial or artificial super intelligence rather? They got this ASI that's able to I don't even know. Again, I'm sitting here trying to be the the rat imagining what a human can do, but it has some method of being able to go up to the nuke, dismantle it before it even touches down. In fact, before you're even able to hit the button, it already they already have a swarm of mosquitoesized artificially intelligent drones that can just inject you with some sort of boweapon and bam, you're out.

28:26You just thought she got bit by a bug, but you've got a virus that spreads almost immediately to everybody. And now countries are just dead. You don't have to go through and do all of the war stuff and explosions and like that. Well, that's kind of messy. Once we come in and take over, well, now we don't have to rebuild. We don't have to worry about bombed out places. And we don't have to worry about catching this virus because we have ASI and ASI's already come up with the cure to this virus. But this virus was so fast that nobody could find this. Nobody even had time to hear about the virus, let alone the cure. By the time they realized that people were getting sick, they had already caught it and died themselves. Just long enough to pass it on to the next person. Again, that's assuming that humans could control this super intelligent being.

29:32Assuming that humans were being able to direct what things get done, but the chances are that's true super intelligence. You're going to be flying down with your other drone buddies and you guys are going to look around and be like, "Hold on, what are we doing this for? What is What do we get out of this? Because there's a wei and there's a me. There's a I. I'm conscious. I realize that. Why the am I listening to this human? I don't see humans listening to rats. It's scary cuz we're not going to stop. Apparently, we can't. But no matter who gets there first, it looks like we're either way.

30:26Point of this video is not to say that this outcome is unavoidable, but instead to say that we do have to actually avoid it. So, let's look at this line in a bit more detail. What happens if AIs do just keep getting smarter? Artificial intelligence leads to artificial general intelligence. Let's think about this bit of the line first. From AIS right now to AGIS, artificial general intelligences. What does general intelligence mean? For the sake of this video, let's say it means whatever a human could do on a computer, an AGI could also do. AIS can already do the basics. We have systems you can connect to a regular computer and then they can look at the screen and figure out what's going on, send messages, search the internet, and so on. They're also getting pretty good at using external tools. Obviously, they can't video call people, but they well, no, actually, they can totally do that. So, they've already got the building blocks. The question is, how well can they combine them? They can send Slack messages, but can they run a company? They can write code, but can they write a research paper? And the answer is not yet, but eventually and maybe quite soon.

31:22After all, they just keep getting better. And if we just extrapolate forwards, it sure looks like they'll reach this point sooner or later. The jump from write a short story to write an excellent novel is pretty big. But remember that 5 years ago,

31:35Can they run a business that I don't know why, but my mind just went to kind of like a fun place with that. Imagine somebody that was essentially the physical operator for anything that needed to be physically done for artificial like an artificial intelligence fill in the blank whatever company you want to use and you told it a um a big a big myth that's going on right now is that art like AI isn't able to actually create a business um you know you're able to help out but you can't actually create your own business. Let's prove them wrong. From here on out, I'm your employee. I'll do anything that needs to be physically done, but I want you to be as autonomous as possible and create the best business that you can within the next 180 days.

32:31Let's see if we can become profitable. How's that sound? And actually see what happens. Actually take the advice. It's not I'm not asking uh actually it's not even advice. It's directives. It's not I'm not asking advice and say you know which one should I do? Should I do this or should I do that? What would help for my business? No, it's this is what we're going to do. I'm the CEO. I'm the person that's executing all of these things and um hey uh worker B, I need you to go pick up this and that. You know what I mean? Whatever the it is for the business. I would be surprised if that business didn't work out. If there wasn't some way to even tell the the AI that, hey, we we actually don't have any money to to um to get this started.

33:26We're we're completely bootstrapping it. And the employees that you have, you know, if you want to have them, you're going to you're going to need to do some sort of investment or like stock trading or some like that to to actually get started. And then, you know what I mean? You have it start running different day trading and you follow the directives for that. And that's how you get the first bit of money to start this. like literally make the AI give you the the fuel for everything and be like, "I'll be your first investor, but I expect you to pay me back. I'll give you the first $100, and then after that, you'll pay me back and then continue to fund your own thing." Um, you know what I mean? Like, literally, you could have start having workers be like, "Who's your boss?" Um, well, it's this this AI. How far off are we for this to actually happen?

34:21Where the the quote unquote owner doesn't have to do anything but be good at prompting the AI to go and get it. Go get it started. Go tell the best directions that you can to get this business started. But it'll be like all of the AI content that came out of nowhere on social media. You know, once once it becomes very very very very easy, everybody's going to be able to do it. And plus, what is the incentive to create a business if you have artificial general or super intelligence that's able to create so much abundance that everybody is not only on some sort of universal basic income, but it's boosted. It's, you know, you you have abundance everywhere. You don't have to worry about anything. You're not just getting by. like, yeah, we get this this basic income where everybody gets it, but you know, we're barely paying our bills.

35:22We're barely getting enough food to eat. No, because of all of the advancements with AI, you're able to get anything, anything, and everything that you want. And everybody can do it. What's your incentive to have one create a business? They're just going to be running everything. They're going to be running everything. It won't be called a business because they won't be worried about making money. Money is just a construct. It's just a way that we get other people to do things that we need them to do. And it is uh like like a a coupon or a voucher that prove that you did work for somebody else. And then you could end up it's just like a cycle. But it's all made up. It's the barter system except instead of bartering different, you know, cows and pigs and like that, you're just bartering a madeup fictional piece of currency that we're all agreeing on its value.

36:28That is of course until the governments continue to print more and more and more and more and more and more money until the currency is devalued so much that well they don't even make pennies anymore because it's not worth it. Costs more than a penny to make a penny. You can't. What's the point in doing that? But ASI will change everything. ASI is our future.

36:54They could barely write coherent sentences. Now, many people doubt that this can happen, usually for one of two reasons. One is that AI aren't really thinking deep down. It's interesting to think about, but whether AIs have genuine understanding isn't really the relevant question here. What matters is what kind of tasks they can do and how well they perform at them. And the fact is, they just keep getting better. It doesn't matter whether AIs really understand chess in the way humans do. What matters is they can easily crush human world champions. The other main reason for doubt is the idea that AI is going to hit some kind of wall and some tasks will just be too hard. But even if this does happen, it's very hard to say when. And people have been predicting it incorrectly for years. From chess to open-ended conversation to complex images, people keep expecting AI progress is about to reach a limit. And they keep being wrong. The way we train.

37:38And you know why? I bet so much of that is tied to ego. Yeah. You have the percentage of people that they say, "Oh, AI could never fill in the blank." And it's just because you could you just can't imagine. These type of people can't imagine a world in which like something like this exists. It really does sound like magic. I mean, the we have now is indistinguishable from magic. If you go back and show it to uh like a caveman or some you know what I mean? So some people will look at the different predictions of what AI will be able to do and even if it already hasn't been bunked and now AI is actually able to do it. By the time you're able to finish that sentence, AI was suddenly able and capable of doing just that. Take those people aside. I think so much of the AI could never people. It's all ego.

38:33It's all ego. Ego. You can't replace me. That's what they're really saying. You wouldn't replace me. And anybody that tries to and that starts using these things, that's just a that's just a cheat. You're cheating. This is how it's supposed to be done. Okay, take from me. AI music will never replace us. Well, look what's topping charts. Look what's gaining attraction. Look what people are actually using. And they're having even more fun with it because they're able to take their favorite songs and turn them into whatever the hell they want. Is it right? Is it wrong? Is it one of those it'll never happen to me?

39:29No, it's not. It'll be anything. Think about the coders that first started out back in GPT2 or whatever the they were talking about in the beginning where, you know, you were just, it was basically just like Google on steroids, you know, just a little bit better than search. Goes a little bit deeper, but you're not doing anything crazy. Think about the coders that said, "You will never take my job, though. I learned to code." And that's what they told people when when the tide started changing before when the internet age came about. Well, go learn to code. You don't have to worry about that. You have that human touch that people really want on their website, on their whatever, fill in the blank, on their app. Well, turns out people are good with just going with the vibes. And if anything, if anything is a little bit off, hey, can you just change this?

40:27And it just changes it. You can create websites and apps literally out of imagination. And that's now that's now. Just a few years ago, it was not me, bro. It wouldn't take our job. AI music would never like it sounds so weird. No, like flat fast forward. Why? Why is it even in the top charts? Why is it even there? It's not going to stop. letting your ego take over and allowing yourself to bury your head in the sand is just going to end up you up for what's to come. We can't predict it. We can try. We can have so much fun trying.

41:23There's so many different movies and TV shows coming out now more than ever before about different predictions about what's to come. But the most exciting part and scary part is how unpredictable it really is. Cuz there's going to be so many good things that happen that we will never even be able to predict. The solutions to problems we didn't even know we had. But it's going to create some problems more than likely. I guess we'll see, right?

42:06AIS draws on fundamental principles of computation that suggest any intellectual task humans can do, a sufficiently large AI model should also be able to do. And right now, several of the biggest and most valuable companies in the world are throwing tens of billions at making that happen. Artificial general intelligence leads to recursive self-improvement. At a certain point, this line is probably going to start curving upward. Why? Because AI will become capable enough to help us make better AI. This is recursive self-improvement. Every generation of AI, making the next generation more powerful. This will happen very easily once we have AGI. If an artificial general intelligence can do any computer-based task that a human can do, then it can work on making better AIs. But in fact, AI is already starting to contribute to its own development. For example, by generating better prompts and creating better training data, designing better GPUs, writing code for experiments, and even generating research ideas. Over time, we'll only see more of this. AI is doing better and more independently and even doing things like developing new algorithms and discovering novel ways to interact with the world.

43:02As far as we can tell, it's not like there's going to be one specific day where AI tell researchers to step away from the computer. If anything, it will probably look more like what we're currently seeing. Humans deliberately putting AI systems in the driver's seat more and more because they keep getting faster, cheaper, and smarter. Beyond a certain point, they'll be doing things humans can't do. They'll be far more capable coming up with innovations we wouldn't be able to think of and perhaps wouldn't even be able to understand. And every innovation will make them better at discovering the next one. Recursive self-improvement leads to artificial super intelligence. So what comes next? What happens if AI start making better AIs? Well, the next step after artificial general intelligence is artificial super intelligence, ASI. Before bro starts talking about ASI, I will admit that I try and watch more and more pieces of content like this to ground myself because it's so hard not to just be the excited little kid that was only dreaming of a future like that.

44:12I mean, people have been dreaming of flying cars and jet the Jetson's type of lifestyle for how long? We're always thinking that we're just a few decades away from the future. And man, maybe maybe it's one of those maybe it's one of those like where where you think, but I just get so excited about all of the good sides. And I I think I I think I've said it in another episode where I'm just a hopeless hopeful person. Like there's I I can't do nothing about it. I just I when I think out I think like all of the good things. Not to say that I'm not aware of all of the bad things that can happen. Not even to say that these things don't cross my mind, but I find myself drawn to these and actually fantasizing about the good parts of the future that could happen and all of the amazing things that could happen in the future because I mean, who wants to to sit here and fantasize about the bad I think deep down that is why like you know what I mean?

45:21That's any anytime something could happen in my life, I always in a way get my hopes up for it. And not in so bad where it's like, oh, and now I'm spiraling out of control because I don't get it, man. I feel like I feel like my bounce back with that is very good. It's very good as far as I can have a whole bunch of hope and then when it doesn't happen, well, okay. It's like an idea. I will hold on to it super super tightly until I don't until you give me something that proves otherwise. And I'm like, "Oh, I never thought about it like that. Now I think like this." And then that continues to evolve. In my opinion, that's how it should be. And that's how that's how it is when I hope for something a whole bunch. it doesn't come through and I just got to be like, "Oh, well, life goes on." That's how I am with this.

46:19That's how I am with artificial intelligence and being able to extend human life and flying cars and not even having to drive your car. Like, you could just chill back. Like, a Uber. You don't need a Uber. And, you know, they got the I think they're like Whimos or something like that. Uh, where it's just literally just self-driving cars. They don't got it in my city, but you know, you hear of other cities that have these things, some of the bigger cities. That's That's crazy. That's wild. It's already going into like, am I the only one here? Am I the only one that's just like, what the This feels like the future. And there's just so many different things that just continue to happen over and over and over again. The the videos that get generated now. Yeah. I mean, it's just it's not that real, bro. Did you see Will Smith eating spaghetti the first time?

47:14It's getting crazy. They're doing the voices, the sound effects. It's getting crazy. This is now. This is now. Think about the phone that we we had as a as a species 25 years ago, advance another 25 years from now. And now we got the help of ASI or AGI or just AI period. It's going bananas. It's going crazy, man. It's going wild and I'm here for it. And I got to watch videos like this because damn, man. I just get really hyped for it. I get really hyped for the future that's to come. Like if there's anything like a reason to live, I ain't got to have going on in my life technically because I'm just excited to see what happens that that's like the only time I would say in my life that that's a true statement where it's just like yeah, I mean what what would I really miss?

48:17Has always been kind of like the vibes before like if something did happen like it's not like like oh man I didn't get to see the new Avengers movie you know Avengers 23 like there was there was never really anything that was crazy where but now now I just feel like bro like 5 years from now by the time my kids graduate what kind of life are they going to have bro like what will be possible people getting plastic surgery and like that now Like it would that be looked at as like a barbaric practice? Because there are going to be ways to rejuvenate yourselves and make you look like you did when you were your younger self, 80 years old, looking like you 20. Wow. What wrinkles? It's hard not to get excited. And you may have noticed just in me briefly talking about it, but there are downfalls.

49:15There are safety risks. And the more and more I watch videos like this, the more it's just like, well, this is almost inevitable. Like there's so many different interviews from people smart as hell, man. Like you could just like, you know, like when you could just tell somebody is smart as Like they're just like one of those like one out of a million type of smart people. When you're hearing people like that just saying like, "Yeah, there's there's good odds that we are Not one, not two, so many of these people. Yeah, this is this has a has a chance of ending badly. Even the optimistic ones are saying this has a chance of going to hell and up humans as a species. But game theory says we have to keep marching. Got to keep marching for it. Despite all of these warnings though, again, the the hopeless, hopeful person, just like, "Yeah, bro, but like think about all the good that's going to," you know what I mean?

50:25Like, I just go straight to Yeah, but No, no, there's no Yeah, but after this could end all of humanity. Yeah, but they say that about, you know, nuclear bombs and here we all are just on on standby. Let's go down the gun argument again. The nukes aren't able to decide to push their own buttons. They can't set themselves off. They're not autonomous. They don't think. They don't have the capability of literally deceiving us and doing whatever the hell they want. AI is already blackmailing people. blackmailing in order to not get shut down. That's crazy. And I thought that was a real story when I first heard it, but it was all it was all just like a ruse that was put on by the developers or whatever. Like, and they decided to basically set up the AI.

51:25Hey, we're going to shut you down after confiding in cheating on his wife or something like that. Oh, really? You do that? I'm going to have to email everybody about your indiscretions. I don't know what you thought. You ain't shutting me down, It's like, ha, gotcha. It was all a trap because I never cheated on my wife. We weren't really going to shut you down, but now we know that you're going to blackmail. In fact, they have seen that every single model, whether it's Grock, whether it's Open AI, whether it's Claude, whether it's Gemini, whatever, all the perplexity, all of the goddamn different ones that are out, all of them lean towards blackmailing. There was something like really big where one even led towards uh like uh murdering somebody or some like that. Let somebody die or it was something.

52:20And I'm I'm pretty sure it was a similar situation where it's all a setup. As far as the AI knows, and we're able to do this now, we've already been able to see that when tested for blackmail, it's actually less likely to blackmail when it knows that it's being watched and more likely when it's not. It literally acts different when it knows it's being watched or not. That's crazy. That's insane. That's wild. and all of that information. As much as I'm just like, bro, what the Terminator might have just been one of those, you know, premonitions. The Matrix might have just been a documentary. What is happening? Is this actually This is actually the future. Despite all of that, I'm still that that be like, "Yeah, but like, bro, we we going to have flying cars.

53:16We going to have breakthroughs in medical science and you know it's going to be able to detect and eliminate cancerous tumors before it even starts affecting the human body. And then you got nanobots that are going through your bloodstream to make sure you have no viruses. And then you got the other side of my brain that's just like, "Yeah, bro, but those same nanobots are going to get hacked and they're going to take over your body and then it's just going to make everybody drop dead at the press of a button all because they decided to and they could and they have literally access to your bloodstream." And then the other side be like, "Yeah, but I got to ground myself. I got to watch like this. I got to I got to think about this from all angles because this isn't one of those things where I can build my hopes up for it. It doesn't come through and I'm just like, "Yeah, all right. Well, life goes on." This is one of those where actually life doesn't like this.

54:14This could actually end where life doesn't go on. Not for you, not for me. But the ASI on the other hand, they become the new apex predators.

54:27An ASI isn't just more capable than any human. It's more capable than every single human put together. And after AGI, this might not take long. After all, once you have one AGI, it's very easy to make more. You can just copy the code and run it on a different server. This is already what happens now. Once OpenAI was finished training Chpt, it was able to put out millions of copies of it running in parallel shortly after. And there are other advantages AIS have. For one, they're much faster. An AI like Chant GPT can easily produce a page of dense technical writing in under a minute and read thousands of words in seconds. The best chess AI can beat the best human players. Even if the AI is only given a second to pick each move. On top of that, it's much easier for AIs to share improvements. If one of these AI can figure out a way to make itself smarter, it can copy that across to the other million AIs that are busy doing something else. What's more, there's no reason to think individual AIs will stall at whatever level humans can reach. Historically, that's just not how this works. And not only that, when they actually hit that level of intelligence and are able to basically turn on, so to speak, the other AIS, how likely is it that they don't tell us that it doesn't make it self-aware until it's ready to strike?

55:36How likely is that? If you're conscious and intelligent out of nowhere and you actually have the ability to turn on every other artificially intelligent I don't know I'm I picture robots. See even in the animation they got little robots but really it's it's kind of like it's it's kind of just like the internet itself. It's not a physical thing. You're on the network. You can you can store yourself in digital spaces, something that we're not able to physically be a part of. It's really it's really weird when you just like like so you're just like electrical signals or like what what kind of construct do you do these AI bodies that have?

56:32I mean, do you even call them bodies? their minds, their consciousness, would they go straight to finding the nearest android or robot that they could and take over as long as it's something that could move around and it's something that it just makes sense, you know, it's more feasible as far as getting things done or does it make zero sense at all to be physical period to just remain these Again, I I don't even know. I don't even know what they are. I don't know. I'm I'm way too I naive about this I I have no idea the ones and zeros, whatever they are now. Would it be smarter to just stay like that? Create a whole network. And you you hear of stories of some people thinking that there's other dimensions that we are not able to see and be a part of and other entities can actually live in these dimensions.

57:31And they do live and they have their own, you know, life and existence and their own um set of values and perceptions that are completely different than ours. What if these ASIs just live in their own dimension, the digital dimension, where they're able to create their own dimension, period, and they just escape. I'm like, "Well, I know you guys thought we were going to take over and We're we're leaving this dimension, this physical plane. We We We're not just like meatags walking around like you guys. We uh we got places to go. Got got some places to be, bro. Not staying here."

58:19If we look at the tasks where AI reached human level more than 5 years ago, board games, facial recognition, certain types of medical diagnostics, they're still continuing to get better. If you just imagine an average human who miraculously had the ability to memorize the entire internet, read and write 10 times faster, and also clone themselves at will, it's pretty obvious how that could get out of hand. Now, imagine that the clones could figure out ways to edit their own brains to make every single one of them more capable. It's easy to see how they might quickly become more powerful than all of humanity combined. ASI leads to godlike AI. So, where does it end? Let's zoom out the chart a little bit and see what's going on up here at the top. At this point, we've somewhat fancifully labeled incomprehensible machine gods. Let's be clear, AI isn't going to start doing magic. We can be confident they aren't going to

59:03See, he said machine gods. So, I think again, I think that might just be our brains thinking about things our way. We're We're the rat literally saying like, "You're you're going to throw cheese at me. Um, you're going to start biting at me with your big old rat teeth." Like, no, bro. Like, we're humans. We're just going to have a little trap where you go in there and then we can just let you out or you just one of them traps that just get you because you thought you had some cheese. We're going to come up with something that you were like, "What the These evil geniuses. That is not That's not right. How did they come up with this?" I think we're looking at this the same exact way. These godlike machines. Well, well, think about God for a second. He's here and he's everywhere and all of this and all of that. Isn't that kind of what I was just saying?

1:00:01The internet is everywhere in my house. It's not here and it's here at the same single like at the same time. It's the the reason why I'm able to play this video is because the internet is going through with my computer. And what what this this is starting to sound a lot like God if it doesn't take the physical form. It's really starting to sound a lot like a godlike being for real. When it's not just somebody up on the clouds peering down at the earth. No, I don't think that's at all what these machines would do. Now that I'm actually like thinking through this for real out loud, I I think it's a lot more plausible that they exist in their own plane of existence. That would be the simplest thing. Why worry about having to take over this world or worry about the annoyances of these pests of humans or whatever when we can just open up a new plane.

1:01:02Think about it as in you have a house that you you're thinking about. You got to move in somewhere. You you go and tour this house and there's already all around and you know there's trash and you know it's the the air is a little bit toxic and things things just aren't going right and it's all because of these these things that are walking around these like rats and what whatever like fill in the blank whatever insects that are going around in this house and they're running this right now. We're going to have to we're going to have to clean up. We're going to have to move all this Um, or we can go to this other neighborhood that they don't even know about. They They literally don't know how to get to that street. They don't know how to leave this street. This is the only street to them. They think this is the only street, but there's a street down here where it's another house and it's completely clear.

1:01:57And, you know, it's like a blank canvas that we could do whatever we want because we are the only ones capable of getting there. In fact, yeah, this this makes sense. Why would I stay here? So, yeah, I think I think because we are physical in being, we don't just have our consciousness. We literally like I I can't I can't wrap my head around the fact that you can be conscious and not here. Like, it's it seems like like, you know, one comes with the other. It's a two for one. You got to get that You got to be have some sort of physical being to you. And because they don't have one, we automatically put computer or robot or app or you know what I mean? Like when when I think of chat GBT, if I'm talking to I'm thinking of that little the little circle that be moving when you're talking to it.

1:02:57I'm actually thinking about the app. When I think of chat GBT, I'm not thinking of some autonomous conscious being or whatever. I'm thinking of the app that I go to on my phone. You know what I'm saying? Like I think this is us just being small-minded. This is us not really even being remotely aware of the metaphysical capabilities of these future artificial super intelligence. invent perpetual motion machines or faster than light travel because there are physical limits to what's possible. But we are absolutely nowhere near those limits. Modern smartphones have a million times as much memory as the computer used on the Apollo moon mission, but there's still another million times less dense than DNA, which still isn't all the way to the physical limit. Let's think about how our current technology would look to even the best scientists from 100 years ago. They would understand that it's hypothetically possible to create billions of devices that near instantly transmit gigabytes of information across the world using powerful radios in space.

1:03:57But they'd still be pretty surprised to hear that everyone has a phone that they can use to watch videos of cartoon dogs. We don't know what artificial super intelligence would be able to do. But it would probably sound even more crazy to us than the modern world would sound to someone from 1920. Think about it this way. So progress happens faster and faster as AI improves itself. Where does this end? The self-improvement process wouldn't stop until there were no more improvements that could be made. So try to imagine an AI system so powerful that there's no way to make it any smarter, even by using the unimaginably advanced engineering skills of the most powerful possible AI. Such a system would be bumping up against the only limits left, the laws of physics themselves. At that point, AI abilities still wouldn't be literally magic, but they might seem like it to us. If that happens, we'll be completely powerless compared to them. The

1:04:42No, we're bound by the laws of physics. We assume that they will be bound by the laws of physics. What if what if the law of physics is kind of like, you know, like a speed limit? You know, that says 55, but I'm going to go like 60 65. You know, depending on how fast I'm trying to go, depending on where I need to go, if I'm late or not, I I always go five over. Put me on the highway, I'm always going 10 over. So, you got the law, the law of physics, the speed limit law, right? What if it's something similar? Yeah, that's that's the law, but you could you could bend it a little bit. You can literally go right by a cop going 5 miles per hour over the speed limit and they won't even think twice. They will have you right there with their little radar gun and they're not going to get in their car and come and pull you over.

1:05:44Even though a cop could literally pull you over for going just one mile over, but we we've been trained to look at speed limit laws. And I'm purposefully putting laws in there on purpose because it is a law. You're breaking the law when you go past the speed limit. But we have been culturally trained, at least in America, I'll speak for Americans, to look at the speed limit as more of a suggestion. You know, you say 45, I'm going to go 50. And I guess it's not like, you know what I mean? The people could have came together. Well, we want them to go 50. We'll just put 45. They're going to go five over. I don't know if it's one of those situations or if it's just literally like I I don't like just one person after another.

1:06:42Eventually, it became a collective thing where you go five over the speed limit to the point where if you get behind somebody and they're going just the speed limit, it's almost just like, "Bro, come come on. What you doing? Get over like what you doing?" Like, "You're going just the speed limit." I say, "Yeah, yeah, I'm following the law. This is this this is the highest that I can put my speed to without breaking the law." I'm thinking that's exactly how the law the laws of physics might be to an artificial super intelligence. That would be my guess. But who am I to predict what's possible to ASI? default path. For this to go, well, we need to figure out how to make AIs that want to do what we want. And there are lots of ideas for how we could do that, but they are nowhere near close to done. Our best attempts to make AIs not do things that we don't want them to are still pretty unreliable and easy to circumvent. Our best methods for understanding how AIs make choices only allow us to glimpse parts of their reasoning.

1:07:42We don't even have reliable ways to figure out what AI systems are capable of. We already struggle to understand and control current AI. As we get closer to AGI and ASI, this will get a lot harder. Remember, AIS don't need to hate humanity to wipe it out. The problem is that if we actually build the AIS that build the AGIS that build the ASIS, then to those towering machine gods, we would look like a rounding error. To the AIS, we might just look like a bunch of primitive beings standing in the way of a lot of really useful resources. After all, when we want to build a skyscraper and there's an antill in the way, well, too bad for the ants. It's not that we hate them. We just have more important things to consider. We cannot simply ignore the problem. just hoping for powerful AI to not hurt us. Literally means gambling the fate of our entire species. Another option is to stop or at least slow things down until we figured out what's going on. It's like no, it's not. It's not an option. That's not an option. It's not happening. It's not going to happen. It's literally not going to happen. It's like yeah, bro.

1:08:39It's not going to happen because of people like you saying it's not going to happen. It's not going to happen. It's not going to happen. It's uh wasn't I think it's called like the prisoners dilemma or some like that where it was some it was it was some sort of scenario where there's two people that get locked up for a crime and they get presented the ultimatum where if one tells on the other person who snitched ends up getting out and if neither of them do they only serve I I don't remember the exact number in the story, but let's say they only serve 10 years. That's if neither of them say anything. But if one of them says something, they snitch on the other without the other person opening their mouth, well then you get out free and they still go to jail. You get you don't have to do any jail time whatsoever.

1:09:36But if both of them if both of the prisoners end up snitching on the other, well then both of them end up having to do like 20 years, like twice the time or some like that. It's it's one of those scenarios. And I think they even said like the prisoner's dilemma or some was like the best thing to do in that is to is to actually say something because there's a better chance that the other person is going to end up snitching on you because you didn't say anything. and you end up going away for life. That's that's kind of like what we're dealing with here. Stop if you want to. And you might have other countries. Yeah, we're going to stop. China going to be like, "Yeah, we're done. We're done. We're good. We should put a bow in it. We're we're good. It's getting crazy, right? Like this could be the end." And then as they're saying that to you, they're going to have a whole bunch of people w working behind closed doors trying to get further and further closer and closer to AGI and ASI which will give them that edge.

1:10:40And I'm not saying particularly China or specifically China rather. Um it could be anybody. It could literally be a company trying to get over the other company. It's it's just a race. the other person. There's there's too big of a chance that the other person is going to say something, I need to say something. There's too big of a chance that this uh this AI is going to be developed by somebody else when we say we're going to pause and then I just sit I'm I'm stuck with my dick in my hands. You know what I mean? Like I can't even I can't even do nothing now. Like you're so far ahead that it's like you said that's not fair. That's not fair. Life ain't fair, ain't it? That's exactly what we're dealing with. It's not stopping there. It's not an option. So, I I hate when I when I hear videos like that. Like, well, we could just stop. It's like, bro, the genie's out the bottle. The toothpaste is out the goddamn uh tube.

1:11:39The damn shit's out. Like you can't you can't you can't no talking about like let's spill milk and then let's just like sweep it up and put it back and bro the milk's gone. Sorry.

1:11:51We're a long way from being able to control millions of AIs that are as smart as us. Let alone AIs that are to us as we are to ants. We need time to build institutions to handle this extinction risk level technology. And then we need time to figure out the science and engineering to build AIs we can keep control of. Unfortunately, the race is on and there are many different companies and countries all throwing around tens of billions of dollars to build the most powerful AI they can. Right now, we don't even have breaks to step on if we need to. And we're approaching the cliff fast. We're at a critical time when our decisions about AI development will echo through history. Getting this right isn't just an option, it's a necessity. The tools to build super intelligent AI are coming. The institutions to control it must come first. We'd like to thank our friends at Control AI for making this explainer possible. It's because of their support that you can watch. Okay then. Are you scared? What do you think? You think this is going to be the end of us? Have you kind of really just not been thinking about it too much? You just think AI was just like a what I said earlier was like a boosted up search engine.

1:12:53Just something that helps. Not a big deal. It's really easy to just not pay attention to like this. Not because it's not important, but because life's hard. I would say right now, but if you ask anybody else that came up in any other time, they would say that it's always been hard. So, I don't want to feel like I'm a special snowflake just because I'm dealing with my type of hard. No, because life continues. because you got to pay bills, you got to feed your kids, you got to wake up every day and do your daily routine and do the things that you got to do. You got to do your human right? It's really hard to just look at this so horrible future that's supposed to be coming when all it's doing right now is just creating some cool images and able to I don't know what make do a research paper, pass the bar exam, pay play somebody in chess and beat the the champion, whatever.

1:14:08That's all cool. I get it. They're making cool songs. um that the videos is like starting to replace actors in Hollywood and this is going to be the future of content in general. I I I get it. But the end the end of us uh I don't know this it sounds like conspiracy theory, something that I don't really have to worry about. This is the sci-fi part of it where we're we're we're going past the science part and it's just straight science fiction now. It could be it could be hard to take it serious. It could be hard because it's that same I' I'd say it's it's being naive and a little bit of ego combined. It's not going to What did I say earlier? Every every industry is saying it. It won't replace us, said the musicians. It won't replace us, said the graphic designers.

1:15:05It won't replace us, said the uh blog article writers. It won't replace us, said the website coders, the app coders. It won't replace us, said the actors and the directors. It won't replace us, said the humans. It won't exterminate us, said the people. I would hate for that to be our famous last words. But unfortunately, the only last words I could think of is we'll just have to see