WEBVTT - AI: Friend or Foe 0:00:00.160 --> 0:00:07.200 Brought to you by Toyota. Let's go places. Welcome to 0:00:07.400 --> 0:00:14.680 Forward Thinking. Pay there and welcome to Forward Thinking, the 0:00:14.840 --> 0:00:17.439 podcast that looks at the future and says, you've got 0:00:17.440 --> 0:00:20.760 a friend. I'm Jonathan Strickland, I'm Lauren Vocaldon, and I'm 0:00:20.840 --> 0:00:27.880 Joe McCormick. So, Joe, I hear you like a intelligence. Uh, 0:00:27.920 --> 0:00:30.520 it's one of the things I like. I hear you 0:00:30.560 --> 0:00:37.400 also like artificial things, like like artificial banana flavoring. Like 0:00:37.400 --> 0:00:40.640 what one thing that Joe absolutely loves. I have seen 0:00:40.760 --> 0:00:44.280 him put artificial banana flavoring on some of the weirdest stuff. 0:00:44.440 --> 0:00:46.640 But I was really trying to get it artificial intelligence. 0:00:46.680 --> 0:00:48.120 I know, I was going about it in a really 0:00:48.200 --> 0:00:51.680 kind of indirect Really, I thought this podcast was going 0:00:51.720 --> 0:00:54.760 to be about artificial vanilla extract. Well it could be, 0:00:54.840 --> 0:00:56.920 but instead I've decided to switch it over to artificial 0:00:56.960 --> 0:01:01.360 intelligence and the idea of creating a true artificial intelligence 0:01:01.360 --> 0:01:06.759 that has human level or beyond intelligence, And how how 0:01:06.760 --> 0:01:11.200 would we make sure it didn't kill us? Well, you 0:01:11.240 --> 0:01:13.279 would have to be in a position to kill us first, 0:01:13.360 --> 0:01:15.480 but that's something we can talk about as well. I 0:01:15.520 --> 0:01:18.360 want to start with the idea of a robot politician, 0:01:18.800 --> 0:01:20.959 which is a sort of construct that we touch on 0:01:21.000 --> 0:01:24.240 in this week's video. Um, so, have you ever read 0:01:24.280 --> 0:01:28.480 the Isaac asthmov short stories Evidence or The Evitable Conflict? 0:01:28.520 --> 0:01:30.640 These are part of I Robot, And yes I have, 0:01:31.000 --> 0:01:33.520 I have not, So for those of us who haven't, Joe, 0:01:33.720 --> 0:01:34.759 do you want to do you want to talk about 0:01:34.760 --> 0:01:36.160 that for a second? Sure? Well, I don't want to 0:01:36.200 --> 0:01:40.320 give too many spoilers, but one of them is about 0:01:40.360 --> 0:01:43.800 a controversy where there is a politician running for an 0:01:43.800 --> 0:01:48.000 elected office who is suspected of being a machine. Right 0:01:48.080 --> 0:01:50.960 and in fact, in the world that as a mom 0:01:51.040 --> 0:01:53.920 has created, it's important for you to realize that machines, 0:01:54.080 --> 0:01:58.080 robots with positronic brains, which are these artificially intelligent brains, 0:01:58.440 --> 0:02:02.520 are not allowed to be on worlds that have human habitation. 0:02:03.040 --> 0:02:06.120 You can only be on uninhabited worlds. It's the only 0:02:06.120 --> 0:02:08.000 place where those robots are allowed to go. So they're 0:02:08.040 --> 0:02:10.840 they're allowed to go to places and do dangerous work 0:02:10.919 --> 0:02:13.080 that benefits the rest of humanity, but they can't be 0:02:13.120 --> 0:02:16.079 on a world that's inhabited by humans. Yeah. So, Asimov 0:02:16.200 --> 0:02:20.440 had an interesting approach to talking about the integration of 0:02:20.600 --> 0:02:24.320 robots and artificial intelligence into society, which I like because 0:02:24.320 --> 0:02:27.760 it was neither utopian nor dystopian. Now it is very 0:02:27.919 --> 0:02:29.880 very much kind of taking light. Let's look at the 0:02:29.880 --> 0:02:32.919 world around us, which is definitely not perfect, but it's 0:02:32.960 --> 0:02:36.520 not you know, twelve monkeys world worst case scenario either. No, 0:02:36.880 --> 0:02:40.920 he was exploring a sort of a smart, well engineered 0:02:40.960 --> 0:02:44.120 system that still had flaws in it. And so the 0:02:44.160 --> 0:02:48.120 system was that the robots in this world are governed 0:02:48.160 --> 0:02:51.240 by three laws. The first law is you cannot harm 0:02:51.280 --> 0:02:54.680 a human. Second law is you have to obey human commands. 0:02:55.320 --> 0:02:58.440 Third law is you can't destroy yourself. Right. And of 0:02:58.440 --> 0:03:01.160 course each of the laws ends up saying unless it 0:03:01.480 --> 0:03:05.680 would break prioritized one to three. Right. Uh yeah, So 0:03:06.000 --> 0:03:08.520 they use this to try to create a framework to 0:03:08.560 --> 0:03:11.960 make sure that a robot never does anything bad. Of course, 0:03:12.040 --> 0:03:15.040 it doesn't always work, and thus is the sort of 0:03:15.120 --> 0:03:18.280 point of conflict for many of Asimov's stories. It's like, uh, 0:03:18.320 --> 0:03:20.560 they're sort of obeying the laws, but the laws are 0:03:20.560 --> 0:03:23.840 coming into conflict in such a way that now we've 0:03:23.840 --> 0:03:26.040 got a problem. Right, And do do recall that he 0:03:26.120 --> 0:03:28.680 was writing fiction to be entertaining. He wrote the laws 0:03:28.680 --> 0:03:31.280 in order to be interestingly flawed so that he could 0:03:31.280 --> 0:03:33.920 exploit that for story purposes. This this was never meant 0:03:33.919 --> 0:03:37.800 to be a complete manifesto of how to robot right, right, 0:03:37.840 --> 0:03:39.960 So back to the two stories you brought up. The 0:03:40.040 --> 0:03:42.320 idea of one of them is that there's a secret 0:03:42.400 --> 0:03:45.920 robot who seems to be human outwardly running for office, 0:03:46.040 --> 0:03:48.480 and the question is is it really a person or 0:03:48.560 --> 0:03:51.680 is it really a robot? But characters within this story 0:03:51.800 --> 0:03:54.360 debate whether it's really such a bad thing to have 0:03:54.440 --> 0:03:58.600 a robot in office because the robot, unlike humans, is 0:03:59.000 --> 0:04:02.520 not self interest did it has? It has these laws 0:04:02.640 --> 0:04:05.640 governing its actions, and these laws will in the end 0:04:05.760 --> 0:04:08.640 ensure that really it isn't going to do harm. In fact, 0:04:08.680 --> 0:04:11.520 one of the main characters in I Robot is this 0:04:12.440 --> 0:04:20.159 humor less kind of misanthropic robo psychologists. She's she's human, 0:04:20.760 --> 0:04:25.600 but she specializes in robo psychology, and she uh, she 0:04:27.040 --> 0:04:31.159 call her humor less. But there are specific passages where 0:04:31.240 --> 0:04:34.120 she she she people try to engage with her and 0:04:34.160 --> 0:04:39.679 she turns her humorless eyes upon them that she she states, 0:04:40.320 --> 0:04:44.680 uh completely, you know, in a in a very uh 0:04:44.720 --> 0:04:48.000 straightforward way, that she thinks robots are superior to human 0:04:48.040 --> 0:04:52.240 beings in in most in most ways because with the 0:04:52.360 --> 0:04:55.720 Robot President character, the person who may or may not 0:04:55.839 --> 0:04:58.719 be a robot. In fact, they're very careful to try 0:04:58.839 --> 0:05:02.560 and build a k either way. They being Asima, really 0:05:02.839 --> 0:05:04.960 build a case either way that could be robot, could 0:05:05.000 --> 0:05:10.039 be human. Uh. She says that he's either a robot 0:05:10.600 --> 0:05:14.919 or a really really really decent human being. So that 0:05:15.080 --> 0:05:17.400 that kind of tells you that that character's perspective and 0:05:17.440 --> 0:05:19.719 a lot of the stories come from from her kind 0:05:19.720 --> 0:05:24.279 of experience that she feels that robots are in fact 0:05:24.560 --> 0:05:27.520 better than people for the most part. Right, But let's 0:05:27.560 --> 0:05:30.320 imagine we take it one step beyond just the idea 0:05:30.320 --> 0:05:33.680 of a single robot in a single leadership role. There's 0:05:33.680 --> 0:05:38.440 another Asimov's story called The Inevitable Conflict, which discusses how 0:05:38.920 --> 0:05:42.840 at some point in the future, all kinds of systems 0:05:42.880 --> 0:05:47.040 are governed by robotic or artificially intelligent controls. Some would 0:05:47.120 --> 0:05:49.760 argue that we're already in that world at some point. 0:05:49.839 --> 0:05:52.760 I mean, you look at the stock market, you know, 0:05:52.920 --> 0:05:56.400 robo trading. You've got like this, this, all these algorithms, 0:05:56.440 --> 0:06:00.320 these these programs that are running all these sophisticated uh, 0:06:00.520 --> 0:06:03.520 you know, algorithms to guide them on when to buy 0:06:03.520 --> 0:06:08.120 and when to sell all these uh, these very short transactions. Uh, 0:06:08.200 --> 0:06:11.440 and they have global consequences. We've talked about that previously 0:06:11.520 --> 0:06:14.320 on this podcast. So in some ways we're already seeing 0:06:14.320 --> 0:06:17.120 that come to pass. Now we're not talking about a computer. 0:06:17.200 --> 0:06:20.000 We go to you know, type in a question of 0:06:20.279 --> 0:06:23.000 you know, how do we do such and such, and 0:06:23.040 --> 0:06:25.200 it gives us the sage advice and then we you know, 0:06:25.279 --> 0:06:27.440 it's not deep thoughts. I don't know. Google does that 0:06:27.520 --> 0:06:30.360 for me about seventy eight times a day. Google, Well, 0:06:30.520 --> 0:06:33.719 Google does do that. We are already sort of wading 0:06:33.720 --> 0:06:35.560 into these waters, whether you know it or not. You 0:06:35.600 --> 0:06:38.440 mentioned the Stock exchange, but you might say, oh, well 0:06:38.440 --> 0:06:41.120 but that's private industry, wild West guns Blaze and they're 0:06:41.160 --> 0:06:43.479 doing whatever. You know, the government wouldn't do that. Well, 0:06:43.760 --> 0:06:47.279 the I R S already has a process called computer scoring, 0:06:47.680 --> 0:06:51.280 where you submit a tax return and computers pre screen 0:06:51.360 --> 0:06:54.479 those returns to decide whether or not we should put 0:06:54.480 --> 0:06:58.200 you into the pile to investigate for an audit. Yeah, 0:06:58.760 --> 0:07:03.240 and the fun act is this podcast goes live the 0:07:03.279 --> 0:07:06.599 week of income Tax Day but after it's already over. 0:07:06.680 --> 0:07:09.120 So I hope you guys thought about that before you 0:07:09.160 --> 0:07:15.320 since your it turns in. Okay, So imagine a future 0:07:15.400 --> 0:07:19.800 where we do have artificially intelligent machines, probably much more 0:07:19.880 --> 0:07:24.480 intelligent than humans. Otherwise, what's the point governing our systems, 0:07:24.480 --> 0:07:28.480 our societies, our economies, making decisions on our behalf to 0:07:28.520 --> 0:07:30.640 try to make the world a better place for us? 0:07:30.680 --> 0:07:33.400 And there's hypothetical pluses and minuses here. What are some 0:07:33.440 --> 0:07:36.280 of the good points? Well, good point would be that 0:07:36.400 --> 0:07:40.520 it be able to make decisions faster and with preface, 0:07:41.200 --> 0:07:44.880 ideally with less bias than a human being with Oh yeah, well, 0:07:45.120 --> 0:07:47.240 let's just start from the ideal point of view before 0:07:47.280 --> 0:07:49.560 we crack a bunch of Okay, So, let's say it's 0:07:49.600 --> 0:07:53.440 a perfect AI and it is uh, you know, you 0:07:53.480 --> 0:07:57.920 wouldn't call it cold. It's logical, but it's also compassionate. Yeah, 0:07:58.000 --> 0:08:00.360 Let's say you've you've created a computer and given it 0:08:00.440 --> 0:08:04.840 some instruction like create the greatest maximal benefit for humanity, 0:08:05.000 --> 0:08:07.680 and it it works out how to do that, which 0:08:07.720 --> 0:08:10.360 it can do because it's super intelligent. It's way smarter 0:08:10.440 --> 0:08:14.360 than any human and it can look at trends in society. 0:08:14.400 --> 0:08:17.400 It can look at unemployment numbers and crime statistics and 0:08:17.440 --> 0:08:21.040 all these things, distribution, water distribution, It can average all 0:08:21.120 --> 0:08:24.720 of that data together to make incredibly accurate predictions about 0:08:24.720 --> 0:08:27.160 the effects of its actions that we just don't have 0:08:27.200 --> 0:08:29.960 the cognitive capability to do. And furthermore, it can do 0:08:30.000 --> 0:08:32.880 all of that with with no hate, no greed, no ambition, 0:08:32.960 --> 0:08:36.280 no prejudice. Right, exactly, it doesn't have a will to 0:08:36.440 --> 0:08:39.880 power of its own. It just has programming. It just has, 0:08:40.280 --> 0:08:43.680 you know, doing what it's designed to do. So that's 0:08:43.720 --> 0:08:47.760 the ideal, perfect vision sort of. It's perfectly capable and 0:08:47.840 --> 0:08:54.400 it's perfectly moral. But on the other hand, machines are unpredictable, 0:08:54.480 --> 0:08:56.760 or at least machines like this. Actually, machines on the 0:08:56.800 --> 0:08:59.319 small scale are very predictable. They do what you tell 0:08:59.360 --> 0:09:01.040 them to do in the thing else they aren't. They 0:09:01.080 --> 0:09:03.640 can't do anything else because they weren't programmed to do. 0:09:03.920 --> 0:09:05.960 But if you create a machine that is more intelligent 0:09:06.000 --> 0:09:09.480 than you, you inherently cannot understand what it's doing. Whoops. Yeah, 0:09:09.600 --> 0:09:14.400 So any machine smarter than you, you sort of lose transparency, right, 0:09:14.480 --> 0:09:18.199 it's hard to understand the decisions that are being made. 0:09:18.280 --> 0:09:20.720 If they're being made, it's at a level, way way 0:09:20.720 --> 0:09:23.480 above your head. Let's here's an example. Let's say that 0:09:23.640 --> 0:09:26.960 we have like the Grand Deep Thought computer that we 0:09:27.040 --> 0:09:30.640 want to consult when we have a particularly tough question. Uh. 0:09:30.679 --> 0:09:32.560 And maybe it's one of these about how do we 0:09:32.679 --> 0:09:36.120 have the maximum benefit for the most people on earth, 0:09:36.200 --> 0:09:40.000 impacting having a negative impact on the least number of people, 0:09:40.440 --> 0:09:42.600 trying to trying to get as good a reaction as 0:09:42.600 --> 0:09:45.360 we possibly can, knowing that there's not likely to be 0:09:45.400 --> 0:09:48.400 any perfect answer that's going to make all ships rise 0:09:48.520 --> 0:09:51.160 up with the tide, right, Uh. And then the computer 0:09:51.240 --> 0:09:53.600 comes back and gives us an answer that, on the 0:09:53.600 --> 0:09:57.400 face of it seems counter intuitive or counter productive. And 0:09:57.440 --> 0:10:01.480 the computer knows because it's run the remulations that while 0:10:01.559 --> 0:10:04.839 this first step is possibly a tough one for us 0:10:04.840 --> 0:10:07.520 to take, it's actually the one that will lead to 0:10:07.559 --> 0:10:10.560 the most beneficial outcome. So then the short term we 0:10:10.640 --> 0:10:14.520 have some hardship. Perhaps it is food redistribution, which would 0:10:14.559 --> 0:10:17.320 be a huge one, right, or water redistribution, which would 0:10:17.320 --> 0:10:20.200 be another huge problem. But let's say that's that first 0:10:20.200 --> 0:10:23.680 step that's really really hard for at least some parts 0:10:23.720 --> 0:10:26.800 of the world to to agree to. Then you could 0:10:26.800 --> 0:10:29.960 have people arguing this thing is trying to destroy us, 0:10:30.000 --> 0:10:32.480 it's not trying to help us, not necessarily being able 0:10:32.480 --> 0:10:34.760 to see that twenty eight steps down the road, it 0:10:34.800 --> 0:10:38.000 actually leads to an outcome that's beneficial for everybody. Likewise, 0:10:38.000 --> 0:10:40.080 on the other hand, it could tell us to do 0:10:40.200 --> 0:10:43.800 something because it is malfunctioning and we don't have the 0:10:43.840 --> 0:10:48.360 transparency capability to understand that it's malfunctioning. Thus it leads 0:10:48.440 --> 0:10:51.840 us down a really horrible path. Without hating us, I 0:10:51.840 --> 0:10:55.160 mean it doesn't. It's not that it's trying to destroy humanity. 0:10:55.200 --> 0:11:00.200 I mean it might in it just it calculated something long, 0:11:00.240 --> 0:11:05.000 it didn't understand something one burn all the week, okay. 0:11:05.800 --> 0:11:08.000 And that flip side of it of it not being 0:11:08.360 --> 0:11:11.040 hateful of a machine inherently not being hateful is that 0:11:11.240 --> 0:11:15.200 a machine inherently has no human empathy or intuition about 0:11:15.360 --> 0:11:18.720 what what step is okay and what is not unless 0:11:18.760 --> 0:11:21.720 we program that in. Yes, So if you haven't thought 0:11:21.920 --> 0:11:26.280 to have the computer specifically look at the most disadvantaged 0:11:26.400 --> 0:11:30.600 people and uh and take special consideration for those people 0:11:30.640 --> 0:11:35.120 who are are essentially they're going to be victims of 0:11:35.200 --> 0:11:37.679 whatever decisions you make. It may be that they have 0:11:37.840 --> 0:11:41.360 a positive outcome, but it may not be unless you've 0:11:41.400 --> 0:11:43.680 built that in. Then the computer is not necessarily going 0:11:43.760 --> 0:11:47.040 to make that consideration for you. And that could be 0:11:47.080 --> 0:11:50.319 a real impact. Right, I'd like to mention something else. 0:11:50.360 --> 0:11:53.199 We say that a computer has no hate, has no greed, 0:11:53.240 --> 0:11:55.600 and all those things, which is inherently true about the computer, 0:11:56.040 --> 0:11:59.520 but the humans that create the computer could have those things. 0:12:00.080 --> 0:12:02.160 And a program is only going to be as impartial 0:12:02.200 --> 0:12:04.400 as its creator was. And and you know, the creator 0:12:04.480 --> 0:12:06.319 might be sitting there going like, well, you know, some 0:12:06.440 --> 0:12:11.160 animals are more than others. Yeah, you know. And so 0:12:11.960 --> 0:12:13.480 even if you even if you take it a couple 0:12:13.480 --> 0:12:15.920 of steps further, because I've seen it proposed that if you, okay, 0:12:16.000 --> 0:12:19.560 create a super intelligent machine and have that super intelligent machine, 0:12:19.840 --> 0:12:22.559 create a really super intelligent machine and use that super 0:12:22.559 --> 0:12:25.600 intelligent machine as it's your president robot, this is deep 0:12:25.679 --> 0:12:30.560 thought creating the Earth. Yeah, because the Earth is a 0:12:30.559 --> 0:12:33.679 computer in Hitchecker's guide, right, right, computer? And I mean, 0:12:34.800 --> 0:12:37.520 you know, input output, if if if if the humans 0:12:37.559 --> 0:12:41.520 creating deep thought we're prejudiced at the beginning, then that 0:12:41.520 --> 0:12:44.679 could just computer form. Right. So yeah, I mean if 0:12:44.760 --> 0:12:47.559 you if you have a bias, and that bias is 0:12:47.559 --> 0:12:49.959 built into the programming you make. Because you know, we're 0:12:50.000 --> 0:12:53.440 talking about a an intelligent computer. I think a lot 0:12:53.480 --> 0:12:57.760 of people just imagine that to be an incredibly powerful machine, 0:12:57.800 --> 0:13:00.599 and that's where it begins it, right, That's it's the 0:13:00.640 --> 0:13:03.520 machine part that's important. But like we said in our 0:13:03.559 --> 0:13:08.800 Singularity podcast, the software is equally as important, and without 0:13:08.840 --> 0:13:12.120 it maybe more important. Yeah, you could argue more important. 0:13:12.160 --> 0:13:14.000 I mean without the hardware, the software can't run. But 0:13:14.040 --> 0:13:18.440 without the software, it can't be intelligent. Right, So unless 0:13:18.480 --> 0:13:22.679 you have very sophisticated software that can take on the 0:13:22.679 --> 0:13:26.679 these these problems, either by designing the next computer so 0:13:26.720 --> 0:13:30.400 that it is the most efficient or by doing it itself. 0:13:31.240 --> 0:13:33.560 If the if the programmers do have this bias, that 0:13:33.600 --> 0:13:36.800 could be reflected in the results. Okay, so people are 0:13:36.840 --> 0:13:40.960 talking about creating a super intelligent machine. Obviously we can't 0:13:41.000 --> 0:13:44.720 do that today, but people are refining AI methods and 0:13:44.800 --> 0:13:47.880 it may in some people's minds sneak up on us, 0:13:47.920 --> 0:13:50.440 Like you could suddenly realize like, oh, we've gone a 0:13:50.440 --> 0:13:53.760 long way down this road to creating something that's equal 0:13:53.800 --> 0:13:56.560 to human intelligence or even beyond it, which is really 0:13:56.559 --> 0:13:59.840 the sweet spot for these problems. Maybe it's a good 0:13:59.840 --> 0:14:02.800 idea to start thinking about what we would need to 0:14:02.880 --> 0:14:07.280 do in order to prevent really negative outcomes if we 0:14:07.280 --> 0:14:10.559 were to create this superintelligent machine. Right, the two big 0:14:10.600 --> 0:14:14.319 negative outcomes these are like taking to the absurd extreme obviously, 0:14:14.800 --> 0:14:17.400 but I call it the kill all humans or the 0:14:17.440 --> 0:14:22.000 subjugate all humans approaches. These are really popular in science fiction. Right. 0:14:22.000 --> 0:14:24.840 This is this is the world of the Terminator, where 0:14:25.200 --> 0:14:28.800 humans have created machines that gain sentience and ultimately turn 0:14:28.880 --> 0:14:31.960 on their creators for one reason or another. And there 0:14:31.960 --> 0:14:35.160 are a lot of different approaches to this kind of storyline. 0:14:35.160 --> 0:14:39.040 In some cases, the machines have malevolent intent. They actually 0:14:39.120 --> 0:14:42.160 want to kill humans because they're you know, essentially robotic 0:14:42.200 --> 0:14:47.000 psychopaths and other versions. It's that the machines have calculated 0:14:47.040 --> 0:14:50.800 that the best possible outcome for whatever planet Earth will 0:14:50.840 --> 0:14:53.400 say is for humans to be wiped off, because that's 0:14:53.400 --> 0:14:54.840 the source of most of the problems. So if you 0:14:54.840 --> 0:14:56.640 get rid of the source, then the problems are gone. 0:14:56.960 --> 0:15:00.000 So in some cases it's like a mistaken like, oh, 0:15:00.040 --> 0:15:01.880 I know how to solve this issue. We just gotta 0:15:01.960 --> 0:15:04.560 kill all the people deemed you ilogical, right, or the 0:15:04.600 --> 0:15:07.960 subjecate all humans. That's essentially the matrix approach where we've 0:15:08.000 --> 0:15:11.000 created machines and we are Our intent was to make 0:15:11.000 --> 0:15:14.120 the machines work for us, but irony of ironies, the 0:15:14.160 --> 0:15:16.680 machines have decided that we're going to be working for them, 0:15:16.720 --> 0:15:20.400 possibly as giant batteries, although that's incredibly inefficient. They get 0:15:20.440 --> 0:15:22.560 better results from cows that should have been the moo 0:15:22.640 --> 0:15:27.840 tricks I've been waiting to use that. Lauren is shaking 0:15:27.840 --> 0:15:30.200 her head at me. So tech stuff fans know what that. 0:15:31.640 --> 0:15:34.360 Joe Joe appreciates it. I think it's only because I've 0:15:34.400 --> 0:15:37.720 heard that one before from from you on tech stuff. 0:15:37.760 --> 0:15:41.920 It's also fair. Okay, let's talk about friendly AI. Okay, 0:15:41.960 --> 0:15:45.280 this is the this is the term. It's friendly artificial intelligence, 0:15:45.360 --> 0:15:48.120 the term for the framework that we would need to 0:15:48.160 --> 0:15:51.960 come up with to create artificial intelligence or a super 0:15:51.960 --> 0:15:55.680 intelligence that has a net benefit to humanity rather than 0:15:55.720 --> 0:15:58.320 a negative outcome. I like to think of friendly AI 0:15:58.360 --> 0:16:00.720 as the AI that walks in the door, takes off 0:16:00.760 --> 0:16:03.160 its jacket and slow first puts on a pair of sneakers. 0:16:03.160 --> 0:16:06.960 A little sweater vest and then just gently leads you 0:16:07.000 --> 0:16:09.680 into the future. Lets us see a little story about trains, 0:16:10.280 --> 0:16:13.320 trains with faces in the future. Can can we have 0:16:13.360 --> 0:16:16.400 anyone building super intelligent AI is listening? Please do that, 0:16:16.520 --> 0:16:19.160 because that would be essentially the best of all We 0:16:19.160 --> 0:16:22.960 were actually designed friendly AI to follow the philosophy of Mr. Rogers. 0:16:23.000 --> 0:16:26.040 We'd be set, won't you be my neighbor? I would 0:16:26.040 --> 0:16:31.160 totally be that that super intelligent AI's neighbor, completely without hesitation. Okay, 0:16:31.160 --> 0:16:33.560 But so there's some guidelines that people have written up 0:16:33.600 --> 0:16:36.000 and and for a while these guidelines have existed. Back 0:16:36.000 --> 0:16:40.280 in two thousand one, the Singularity Institute published a thing, 0:16:40.640 --> 0:16:43.680 a rather lengthy thing that I will not go into deep, 0:16:43.720 --> 0:16:46.480 deep detail of, but but they began by positing that 0:16:46.480 --> 0:16:49.360 that since growth in AI is and I quote astronomically 0:16:49.440 --> 0:16:52.480 faster than the rate of human evolution um, that we 0:16:52.560 --> 0:16:55.520 need to be thinking about this issue. And hey, we'll 0:16:55.520 --> 0:16:59.760 talk about that that belief system um in our episode 0:16:59.840 --> 0:17:02.360 or already talked about it in our episode about the Singularity. 0:17:02.440 --> 0:17:04.240 We don't know which one will come first. I will 0:17:04.400 --> 0:17:07.040 I will say that it definitely has evolved much If 0:17:07.040 --> 0:17:09.600 you think of human evolution as taking over the course 0:17:09.640 --> 0:17:12.000 of millions of years, and the fact that we've had 0:17:12.119 --> 0:17:15.680 computers since the like nineteen forties. If you want to 0:17:15.720 --> 0:17:18.720 be really generous, I can I can agree with the 0:17:18.800 --> 0:17:22.679 astronomically faster evolution. I don't know that necessarily leads to 0:17:23.280 --> 0:17:27.760 superintelligent computers, but but pray continue, sure um and and hey, 0:17:27.800 --> 0:17:31.639 either way, caution and thought are good. So they specifically 0:17:31.680 --> 0:17:34.040 suggest that we should be careful not to expect a 0:17:34.119 --> 0:17:36.760 machine mind to operate like a human mind. Um, that 0:17:36.760 --> 0:17:40.920 that we shouldn't anthropomorphize AI. Right, that's a really good point. Sure, sure, 0:17:40.960 --> 0:17:44.119 And building from there, they lay out the challenges in 0:17:44.200 --> 0:17:50.080 creating friendly AI um being the creation of ethical content UM, 0:17:50.119 --> 0:17:54.440 creating a machine capable of acquiring that content, even asking 0:17:54.560 --> 0:17:58.560 human questions when necessary, but simultaneously knowing enough to resist 0:17:58.640 --> 0:18:02.800 human manipulation and sell correct for human errors. That's pretty cool, though. 0:18:03.800 --> 0:18:06.159 They go into a lot more depth in the recommendations. 0:18:06.520 --> 0:18:09.880 Believe these are based on Yudkowski's work, right, yeah, yeah, 0:18:09.960 --> 0:18:13.760 he did a book length kind of paper also in 0:18:13.760 --> 0:18:16.960 two thousand one called Creating Friendly AI one point oh, 0:18:17.040 --> 0:18:22.440 the Analysis and Design of Benevolent Goal architectures. Elias are Udkowski, 0:18:22.480 --> 0:18:25.520 who we mentioned in our podcast about the Singularity. He's 0:18:25.560 --> 0:18:29.960 written at length about this specific problem, the friendly AI problem. Yes, yes, 0:18:30.040 --> 0:18:32.200 and uh, and we'll have more to say about an 0:18:32.200 --> 0:18:36.080 interesting thought experiment he came up with in a little bit. Yeah, 0:18:36.119 --> 0:18:39.680 so we should back up and say, hey, wait a second, 0:18:40.280 --> 0:18:42.560 why do we really need to worry about friendly AI 0:18:42.920 --> 0:18:45.880 kill all humans and subjugate all humans? I pretty much 0:18:45.920 --> 0:18:49.760 covered that, why would that happen? Well, okay, what if 0:18:49.800 --> 0:18:53.560 we just do what apparently most AI developers are doing 0:18:53.960 --> 0:18:56.520 and just keep going and hope it will work out 0:18:56.560 --> 0:18:59.600 for the best. Uh. There Actually there have been some people, 0:18:59.720 --> 0:19:02.919 some thinkers in in friendly AI who have pointed out 0:19:02.960 --> 0:19:05.480 that this seems to be the dominant approach, just kind 0:19:05.520 --> 0:19:07.560 of hope it's going to work out well and and 0:19:07.640 --> 0:19:11.399 hope that no one's programming psychopathic tendencies into their software. 0:19:12.480 --> 0:19:14.240 Part part of it, I would argue, is that a 0:19:14.280 --> 0:19:17.000 lot of programmers say that we're so far away from 0:19:17.040 --> 0:19:21.280 a a human level intelligence or superhuman level intelligence of 0:19:21.280 --> 0:19:25.360 of AI, uh that could do anything beyond a very 0:19:25.400 --> 0:19:29.200 specific task. We're so far away from that that it's 0:19:29.280 --> 0:19:31.439 not really that important to worry about it at the moment. 0:19:31.600 --> 0:19:34.639 And uh so there's that level, right, that's the idea 0:19:34.720 --> 0:19:36.840 that we're all working on these bits and pieces that 0:19:36.960 --> 0:19:40.399 ultimately could come together to make a superhuman intelligent AI 0:19:40.480 --> 0:19:44.480 in the future. But right now, years out, we're far 0:19:44.560 --> 0:19:47.040 enough away right now where that's you know, come on, 0:19:47.800 --> 0:19:50.960 I agree with you that it probably is a good 0:19:50.960 --> 0:19:53.160 ways out. I'm not one of those people who thinks 0:19:53.200 --> 0:19:56.200 the singularity is near. I think it's probably a long 0:19:56.240 --> 0:20:00.119 way off. But even with it being probably a long 0:20:00.000 --> 0:20:02.840 long way off, it's way better to be safe than sorry. 0:20:03.119 --> 0:20:05.960 And that's where I do agree with these friendly AI proponents. 0:20:06.080 --> 0:20:08.879 I think it's a good idea to be thinking about this, 0:20:08.960 --> 0:20:11.119 even if we're thinking about it way earlier than we 0:20:11.160 --> 0:20:13.119 need to. So were you were you a boy scout? 0:20:13.640 --> 0:20:16.320 Be prepared? So there you go. I mean, I make 0:20:16.359 --> 0:20:18.680 a joke. I was also a boy scout, be prepared 0:20:19.720 --> 0:20:21.399 a boy scout. Lauren is not a boy scout, So 0:20:21.440 --> 0:20:23.720 we're shunning Lauren for the for the purpose of this 0:20:23.760 --> 0:20:26.439 little exchange. No, uh, but I mean the idea of 0:20:26.480 --> 0:20:34.760 be prepared. The girl scouts don't be prepared exactly, just whatever. No, no, 0:20:34.800 --> 0:20:38.119 but the Scots are great. Come on, you're getting me 0:20:38.200 --> 0:20:40.960 off track. I love cookies. Be prepared as a really 0:20:41.000 --> 0:20:46.320 important idea, just in general, because even if this eventuality 0:20:46.359 --> 0:20:50.600 doesn't come to pass, you you're okay, right, It's it's 0:20:50.640 --> 0:20:52.639 if the eventuality comes to pass and you're not prepared, 0:20:52.720 --> 0:20:54.920 that's when you're really stuck. And this is the same 0:20:54.920 --> 0:20:57.080 sort of thing we see in lots of different fields, 0:20:57.080 --> 0:20:59.879 not just artificial intelligence. We're talking about just general to 0:21:00.080 --> 0:21:04.399 zaster preparedness, the idea that you need those preparations for 0:21:04.440 --> 0:21:07.920 that worst case scenario because there's a chance that could happen. Yeah. 0:21:07.960 --> 0:21:10.680 I think there are very good reasons for going ahead 0:21:10.720 --> 0:21:14.040 and getting prepared rather than just hoping it will turn out. Okay. 0:21:14.040 --> 0:21:16.720 I want to give one specific quote from a paper 0:21:16.720 --> 0:21:20.480 called Thinking inside the Box Controlling and Using an Oracle AI, 0:21:20.600 --> 0:21:22.000 which is what I'm going to talk about more in 0:21:22.040 --> 0:21:25.440 a minute. That's a two thousand twelve paper by Armstrong, Sandberg, 0:21:25.520 --> 0:21:28.919 and Bostrom, and they give this quote. They say, in 0:21:28.960 --> 0:21:32.960 the space of possible motivations, likely a very small fraction 0:21:33.040 --> 0:21:37.480 is compatible with coexistence with humans. A randomly selected motivation 0:21:37.560 --> 0:21:40.919 can hence be expected to be dangerous so we're talking 0:21:40.960 --> 0:21:44.640 about not just something, not just a machine that has intelligence, 0:21:44.720 --> 0:21:48.080 but is acting upon some form of motivation. Yeah, it 0:21:48.119 --> 0:21:50.359 would have a motivation. Well, obviously a machine like this 0:21:50.400 --> 0:21:52.880 would have some kind of programming, it would have a goal, 0:21:53.040 --> 0:21:55.760 some kind of motivation. And let's imagine it has a 0:21:55.800 --> 0:21:59.879 really harmless goll like you've you've programmed a super intelligent 0:22:00.040 --> 0:22:03.120 machine to run a paper clip factory. This is an 0:22:03.119 --> 0:22:06.080 example they give, to make as many paper clips as possible. 0:22:06.480 --> 0:22:09.199 There is an inherent danger in the power of that 0:22:09.280 --> 0:22:12.840 super intelligence, because that machine is smarter than any human, 0:22:13.240 --> 0:22:16.159 anybody who can tell it what to do. Otherwise, it 0:22:16.280 --> 0:22:18.960 may just decide I'm going to do a really good 0:22:19.080 --> 0:22:21.160 job at making paper clips. So I'm going to turn 0:22:21.320 --> 0:22:23.800 this building into paper clips, and I'm going to pick 0:22:23.880 --> 0:22:26.439 up these people and make them into paper clips. And 0:22:26.440 --> 0:22:29.560