WEBVTT - What Does Artificial Intelligence Mean? 0:00:07.080 --> 0:00:11.040 Alexa, what's the best science podcast on air? Hey? Are 0:00:11.039 --> 0:00:13.560 you trying to replace me with Alexa? What's going on here? 0:00:14.120 --> 0:00:17.720 Do you think you're replaceable? There's no way an artificial 0:00:17.760 --> 0:00:21.560 intelligence could ever make jokes nearly as funny as I am. 0:00:21.600 --> 0:00:24.400 I think there's no way an artificial intelligence would laugh 0:00:24.440 --> 0:00:27.280 at your jokes. I'm pretty sure I could. I could 0:00:27.320 --> 0:00:30.200 program a pretty dumb computer to laugh at my jokes. 0:00:30.240 --> 0:00:34.280 It's called the laugh track. But that's Hey, that's a 0:00:34.280 --> 0:00:38.000 new challenge for for AI. You know, first chest then 0:00:38.080 --> 0:00:43.760 go now science comedy. That's right, now, programs something that 0:00:43.800 --> 0:01:05.560 can find humor in Daniel's ramblings him and I'm Daniel, 0:01:05.680 --> 0:01:09.040 and this is our podcast. Daniel and Jorge explained the Universe, 0:01:09.200 --> 0:01:12.080 in which we try to download everything we know about 0:01:12.120 --> 0:01:15.959 the universe, episode by episode into your brain, whether you're 0:01:16.000 --> 0:01:19.840 a real person or an artificial intelligence, listening to our 0:01:19.920 --> 0:01:24.319 podcast while trying to sound intelligent about it while writing 0:01:24.319 --> 0:01:28.280 your own humor for the open mic AI Night. The 0:01:28.319 --> 0:01:35.800 topic of today's podcast is what is artificial intelligence? And 0:01:36.680 --> 0:01:39.880 very importantly is it dangerous? That's right? Should you be 0:01:40.440 --> 0:01:42.880 looking at your window for the first signs of the 0:01:42.959 --> 0:01:46.320 robot revolution. Should you be afraid of your Alexa? Should 0:01:46.360 --> 0:01:50.240 you be worried about that robot vacuum cleaner getting resentful 0:01:50.360 --> 0:01:52.600 for having to do all the dirty work and eating 0:01:52.600 --> 0:01:56.640 your face off in the middle of the night. That's 0:01:56.680 --> 0:01:59.240 a bit dark. It seems kind of sinister, doesn't it. 0:01:59.240 --> 0:02:03.280 It's like sitting there are circling, circling, circling, waning, waiting, waiting. 0:02:03.560 --> 0:02:06.600 I think those things are creepy, right. Maybe it wants to, 0:02:07.160 --> 0:02:10.000 you know, clean your face, it wants see That's the question. 0:02:10.360 --> 0:02:13.680 Does a robot vacuum cleaner want anything? What does it 0:02:13.720 --> 0:02:15.720 mean for it to want? What is it like to 0:02:15.760 --> 0:02:19.320 be a robot vacuum cleaner? The next great paper in philosophy? 0:02:19.480 --> 0:02:21.720 So this is kind of in the zeitgeist right now. 0:02:21.760 --> 0:02:24.880 I mean, people are really excited about artificial intelligence. But 0:02:24.880 --> 0:02:27.280 at the same time there are big names like Elon 0:02:27.360 --> 0:02:31.520 Musk kind of warning people like, hey, artificial intelligence not 0:02:31.600 --> 0:02:33.680 such a good idea. That's right, it's a huge topic. 0:02:33.720 --> 0:02:35.600 I mean, you drive around like San Francisco, you see 0:02:35.680 --> 0:02:39.160 artificial intelligence, machine learning, deep learning. It's on billboards. Even 0:02:39.200 --> 0:02:41.000 you know you want to get a million bucks for 0:02:41.040 --> 0:02:43.880 your new company, you just say the words AI, deep learning, 0:02:43.880 --> 0:02:46.880 and boom people are throwing cash to right, people are 0:02:46.919 --> 0:02:51.120 learning in the deepest learning. Um. It's definitely part of 0:02:51.120 --> 0:02:53.920 the cultural moment, and you see that reflected not just 0:02:53.960 --> 0:02:55.640 in like what deep thinkers are saying, but also in 0:02:55.639 --> 0:02:58.040 like science fiction. You know, a lot of the near 0:02:58.160 --> 0:03:01.079 term dystopian these days is about how AI will take 0:03:01.120 --> 0:03:04.680 over and the dangers of AI. Another way, like thirty 0:03:04.760 --> 0:03:07.919 years ago is about the dangers of UH of radiation. Right, 0:03:07.960 --> 0:03:11.000 that was the new dangerous thing physicists that invented. Now 0:03:11.040 --> 0:03:13.760 the new dangerous technology that we're all worried about is AI. 0:03:13.840 --> 0:03:17.760 It's a new promise in peril. Um. Yeah. AI. Every 0:03:17.760 --> 0:03:19.840 piece of technology is a double edged sword, right. You 0:03:19.880 --> 0:03:21.560 can use it for good, you can use for evil. 0:03:21.600 --> 0:03:24.919 But AI is special because it's not just technology, is 0:03:24.960 --> 0:03:27.160 not just a tool that people use. It's a tool 0:03:27.400 --> 0:03:30.880 that that has independence, that has autonomy. And that's why 0:03:30.880 --> 0:03:33.560 it's such a vexing question. Well people, I'm sure people 0:03:33.560 --> 0:03:37.160 everyone associated with robots and machines and computers, but we 0:03:37.200 --> 0:03:40.840 were kind of wondering if people actually knew what artificial 0:03:40.840 --> 0:03:43.280 intelligence was, like, what makes it work, what makes it 0:03:43.320 --> 0:03:46.880 different than than real intelligence? M I bet you that 0:03:47.760 --> 0:03:50.760 the people who say the phrase artificial intelligence don't actually 0:03:50.760 --> 0:03:53.720 know what they're talking about, which is probably true for 0:03:53.840 --> 0:03:58.520 most topics andology, it's true for me probably, I'm sure. 0:03:59.240 --> 0:04:01.720 But we're wondering if you guys out there knew what 0:04:01.880 --> 0:04:05.200 artificial intelligence was. And so, as usual, Daniel went out 0:04:05.240 --> 0:04:07.800 and as people in the street, and here's what they 0:04:07.880 --> 0:04:11.120 had to say. Um, it's the idea that we can 0:04:11.120 --> 0:04:14.720 create some type of material thing that could think on 0:04:14.760 --> 0:04:17.640 its own ultimately, And do you think it's something we 0:04:17.640 --> 0:04:19.120 should be concerned about? Is it ever going to be 0:04:19.120 --> 0:04:22.600 a threat to humanity? I mean possibly, but I mean 0:04:22.680 --> 0:04:24.480 we never We don't know everything. We can know the 0:04:24.520 --> 0:04:26.680 bounds of what could be a threat, we cannot be 0:04:26.720 --> 0:04:30.720 a threat. Yeah, it's AI, and it's the stuff that's 0:04:30.839 --> 0:04:34.320 used in various technological applications basically just kind of like 0:04:35.160 --> 0:04:39.200 trying to make machines replicate certain aspects of human intelligence. 0:04:39.560 --> 0:04:41.360 Stuff like that. Okay, And do you think it could 0:04:41.360 --> 0:04:43.120 ever be a threat to humanity? Is something we should 0:04:43.120 --> 0:04:45.240 be worried about? I guess since I don't have a 0:04:45.320 --> 0:04:49.200 particularly strong opinion on it, I don't think so. So 0:04:49.839 --> 0:04:52.599 I guess I'll say no for now. Um, I'm assuming 0:04:52.680 --> 0:04:58.680 that's the idea that computers or electronics can have like sentience. Right, 0:04:59.600 --> 0:05:02.160 Are you worried that computers would when they take over 0:05:02.200 --> 0:05:05.120 and make us their slaves? And not really, I don't 0:05:05.160 --> 0:05:07.719 think it will come to that point. All right. Those 0:05:07.720 --> 0:05:10.720 are pretty sophisticated answers. I like the ones that said, 0:05:10.839 --> 0:05:14.240 um oh, artificial intelligence, that's just AI, right, Like that's 0:05:14.240 --> 0:05:18.520 an answer. So that's an answer to every question. You know, 0:05:19.080 --> 0:05:24.640 what is Google blex Zavi Brown? Oh, that's just gens. Yeah, acronyms. 0:05:24.760 --> 0:05:27.719 Acronyms can make you look intelligent, that's it. That's the 0:05:27.760 --> 0:05:34.600 real artificial intelligence to speaking acronym acronym intelligence. Um. You know, 0:05:34.760 --> 0:05:37.120 but people had some sense that it's, you know, something 0:05:37.160 --> 0:05:40.520 that can think for itself, or something do something for you, 0:05:40.720 --> 0:05:43.640 or create something that can think by itself. There's definitely 0:05:43.680 --> 0:05:46.400 the nuggative idea is definitely out there. They use it 0:05:46.600 --> 0:05:49.159 in relation to what it can do. That's right, yeah, 0:05:49.440 --> 0:05:53.520 exactly what what is what's the new capability that defines it? Yeah? 0:05:53.839 --> 0:05:56.040 Right yeah, and that it's it's a fascinating way to 0:05:56.040 --> 0:05:57.840 think about it, you know. And uh, it's definitely a 0:05:57.880 --> 0:06:00.200 tricky question, right because I guess we know it in 0:06:00.200 --> 0:06:02.760 the context of UM using them for things, right, Like, 0:06:02.800 --> 0:06:05.360 we're people, don't just create a I because we want 0:06:05.360 --> 0:06:07.960 to create artificial beings. It's like, so it can help us. 0:06:08.320 --> 0:06:11.520 I want to create artificial beings. What's what's wrong with that? 0:06:11.520 --> 0:06:15.120 That sounds pretty awesome. Create a whole army of physics 0:06:15.160 --> 0:06:19.400 artificial physics grad students. It sounds pretty cool. Kids. Yeah, 0:06:19.440 --> 0:06:23.279 you mean, are they worried about competing with my digital children? Natural? 0:06:23.920 --> 0:06:26.760 They know you'd rather have artificial children. I didn't say 0:06:26.760 --> 0:06:30.200 I'd rather have artificial that in addition to my beautiful, 0:06:30.200 --> 0:06:32.440 wonderful natural children, which I should not be talking about 0:06:32.480 --> 0:06:34.599 on this podcast. I'd love to have a whole, you know, 0:06:34.720 --> 0:06:38.880 cadre of artificial children to do my bidding, unlike your 0:06:38.880 --> 0:06:42.320 real children. If who won't do your bidding? So somebody 0:06:42.400 --> 0:06:47.280 listening children? And that sort of goes to the heart 0:06:47.279 --> 0:06:49.880 of the question. You know, UM, if you created a 0:06:49.920 --> 0:06:52.480 digital being with artificial intelligence, would it listen to you? 0:06:52.560 --> 0:06:54.960 Or would it make its own decisions? Right? And so 0:06:55.360 --> 0:06:57.160 that's why we thought it would be interesting to dig into, 0:06:57.279 --> 0:06:59.640 like what is artificial intelligence? If it just did what 0:06:59.680 --> 0:07:01.680 you told that to do. It wouldn't maybe be in 0:07:01.720 --> 0:07:04.839 an artificial intelligence you're saying nobody smart should listen to you, 0:07:04.880 --> 0:07:08.839 is what you're saying. I'm saying they should decide for 0:07:08.880 --> 0:07:15.640 themselves whether I'm I'm worth following. So let's break it 0:07:15.680 --> 0:07:19.600 down for people. Daniel, what is artificial intelligence? Well you 0:07:19.600 --> 0:07:21.320 should listen to this podcast and that will give you 0:07:21.360 --> 0:07:25.560 the answer. Done. Um, Well, you know, I think to 0:07:25.600 --> 0:07:27.760 understand what artificial intelligence is, we should think for a 0:07:27.840 --> 0:07:29.800 moment about what do we mean by intelligence? Right? And 0:07:30.040 --> 0:07:32.880 very simply intelligence. It's just the ability to learn, is 0:07:32.920 --> 0:07:36.360 to find patterns to extrapolate from them. Really, that's how you. 0:07:36.920 --> 0:07:39.160 But like a dog can learn. But a dog, you 0:07:39.160 --> 0:07:41.760 wouldn't say it's intelligent, would you? Absolutely? I would say 0:07:41.760 --> 0:07:44.040 a dog is intelligent. You can teach a dog, you 0:07:44.080 --> 0:07:46.040 can train a dog. It's more intelligent than a rock. 0:07:46.360 --> 0:07:50.920 But would you say by a lot, Oh my gosh, 0:07:50.960 --> 0:07:52.880 if you like, never interact with a dog. A dog 0:07:52.960 --> 0:07:55.920 is like a living sension being. It feels that experiences, 0:07:55.960 --> 0:07:59.320 It definitely learns. It can recognize you. I mean, dogs 0:07:59.320 --> 0:08:01.000 can do complicated of think. The dog is a perfect 0:08:01.000 --> 0:08:03.040 example it's you know, I wouldn't trust it to do 0:08:03.120 --> 0:08:06.480 my taxes. You know, I don't know compared to our 0:08:06.680 --> 0:08:08.400 tax accountant, did I do a pretty good job. I mean, 0:08:08.400 --> 0:08:11.120 you can say that's an intelligent dog, but you wouldn't say, like, 0:08:11.160 --> 0:08:14.400 that's the epitome of intelligence. I wouldn't say the dogs 0:08:14.440 --> 0:08:16.720 are the most intelligent beings in the universe. But that's 0:08:16.720 --> 0:08:19.360 what we're talking about. We're talking about do they have intelligence? 0:08:20.080 --> 0:08:22.440 Pretty example, because they can learn, you can train them, 0:08:22.640 --> 0:08:24.840 and you the cool thing about an intelligent being is 0:08:24.840 --> 0:08:27.080 that you can train it to do something even if 0:08:27.080 --> 0:08:29.760 you don't know how to do it. Say, for example, 0:08:29.800 --> 0:08:33.720 you want your dog to recognize you, right, but tear 0:08:33.760 --> 0:08:35.839 the face off anybody who tries to break into the house, 0:08:36.160 --> 0:08:39.280 right guard dog. Okay, so you can train a dog. 0:08:39.360 --> 0:08:41.440 You reward it when it does the right thing, and 0:08:41.440 --> 0:08:43.480 you punish it when it does the wrong thing. You 0:08:43.520 --> 0:08:46.520 don't know how to, like build a being that does that, 0:08:46.520 --> 0:08:49.880 that like recognizes your face and recognizes stranger's faces and 0:08:49.920 --> 0:08:52.840 makes these decisions. That's hard task, you know, it's not 0:08:52.920 --> 0:08:55.200 easy to do. But you can train a dog. A 0:08:55.240 --> 0:08:58.080 dog can learn how to solve this problem and all 0:08:58.120 --> 0:08:59.760 you need to do to train it is to reward 0:08:59.800 --> 0:09:02.000 it and punch it. So you're saying, just the ability 0:09:02.040 --> 0:09:04.959 to sort of learn from your mistakes or learn from 0:09:05.679 --> 0:09:08.640 your surroundings, that's what you would call intelligence. Yeah, And 0:09:08.720 --> 0:09:11.760 dogs have less of it than we do, and more 0:09:11.800 --> 0:09:14.840 of it than cats and mice, um. But they have 0:09:14.920 --> 0:09:17.600 some of it for sure, which is what makes them trainable. 0:09:17.640 --> 0:09:20.600 And you know, I wonder sometimes, because dogs can be trained, right, 0:09:20.960 --> 0:09:23.360 nobody ever trains their cat. What does that say about 0:09:23.400 --> 0:09:27.040 a cat's intelligence. I've always thought I love cats, but 0:09:27.080 --> 0:09:30.120 I've always thought dogs are probably smarter than cats because 0:09:30.160 --> 0:09:32.920 you can train them, right, Or maybe cats are more 0:09:33.000 --> 0:09:35.840 intelligent in that they're they're not they don't allow themselves 0:09:35.880 --> 0:09:39.199 to be trained by humans, right. And rocks, by that metric, 0:09:39.240 --> 0:09:41.760 are the most intelligent because they completely ignore you, right 0:09:41.920 --> 0:09:43.880 to see the fallacy of that argument right there. But 0:09:43.920 --> 0:09:47.480 I mean, maybe there's sort of sort of like a hump, right, like, 0:09:47.600 --> 0:09:51.160 as you get more intelligent, you're easily more trainable, trainable, 0:09:51.160 --> 0:09:54.439 trainable bout somebody. You get so intelligent that you rebel 0:09:54.480 --> 0:09:56.600 against your masters, And so how do you tell the 0:09:56.600 --> 0:10:00.400 difference between something that's totally unintelligent and something that's so 0:10:00.440 --> 0:10:04.240 intelligent and completely ignores you. Yeah, I don't know deep 0:10:04.360 --> 0:10:08.079 question or believes all the rocks are probably thinking about him. 0:10:08.160 --> 0:10:09.840 Well sure, I mean if he used to use the 0:10:09.840 --> 0:10:12.800 ability to listen to what I say as a benchpark 0:10:12.840 --> 0:10:16.040 of intelligence and then yeah, there's, um, something super intelligent 0:10:16.160 --> 0:10:18.000 could be just as smart as the rock. But obviously 0:10:18.000 --> 0:10:21.080 a cat is still making decisions and acting and you know, 0:10:21.200 --> 0:10:24.120 doing things, so it's intelligent. But maybe it's much more 0:10:24.120 --> 0:10:26.480 intelligent than a dog because it's it chooses not to 0:10:26.520 --> 0:10:28.719 listen to us. All right, I think we need to 0:10:28.760 --> 0:10:32.239 have a whole other podcast on who's smarter cats or dogs? 0:10:32.280 --> 0:10:34.920 And before we do that, we will collect some data 0:10:35.160 --> 0:10:37.920 to answer this question. Um, but I think with the 0:10:38.080 --> 0:10:40.800 question we're focusing on is what is artificial intelligence? So 0:10:41.120 --> 0:10:43.880 natural intelligence just the ability of an animal to learn. 0:10:44.280 --> 0:10:48.280 Artificial intelligence would be if something artificial that we create 0:10:48.559 --> 0:10:52.640 has that same property, the ability to change the way 0:10:52.840 --> 0:10:56.880 it processes things in response to what it sees about 0:10:56.920 --> 0:11:00.319 the world. Yeah, artificial intelligence is a very broad field 0:11:00.320 --> 0:11:02.600 with lots of elements that we couldn't cover in just 0:11:02.760 --> 0:11:05.600 one episode of a podcast. But let's just talk today 0:11:05.600 --> 0:11:09.320 about one important sub field of AI, which is machine learning, 0:11:09.679 --> 0:11:12.640 or more specifically, I would say, let's focus on training. Right, 0:11:12.800 --> 0:11:17.560 can you build something or something artificial that can be trained? Right? 0:11:18.200 --> 0:11:20.280 And uh? I think, I think let's talk for a 0:11:20.320 --> 0:11:23.439 moment about, you know, how how normal computers work, and 0:11:23.480 --> 0:11:26.640 then we can talk about how computers, smart computers, computers 0:11:26.640 --> 0:11:29.200 that can learn, computers with artificial intelligence, how they work. 0:11:29.400 --> 0:11:31.120 I think, I think what you keep talking about cats 0:11:31.120 --> 0:11:34.560 and dogs? All right, we'll talk about cats and dogs, 0:11:34.559 --> 0:11:50.640 but first let's take a quick break. Let's talk about 0:11:50.800 --> 0:11:56.000 what computers can do. Yeah, let's because computers are smart, right, 0:11:56.200 --> 0:11:58.439 you can program a computer to do smart things, but 0:11:58.679 --> 0:12:02.040 that doesn't necessarily mean it has intelligence. That's right. There's 0:12:02.040 --> 0:12:04.360 a difference between a computer that can do something and 0:12:04.360 --> 0:12:06.880 a computer that can learn something. Right. The way I 0:12:06.880 --> 0:12:09.920 think about non intelligent computers is the way you sort 0:12:09.960 --> 0:12:12.880 of think about machines. Right. You can tell them what 0:12:12.960 --> 0:12:15.880 to do, and they do exactly what you tell them, 0:12:16.080 --> 0:12:18.319 regardless of whether it's the right thing. You don't give 0:12:18.360 --> 0:12:20.360 them like a goal and say, hey, I just want 0:12:20.400 --> 0:12:22.720 the house to be clean. Figure it out. You have 0:12:22.760 --> 0:12:24.840 to tell them exactly what to do. You say, step 0:12:24.880 --> 0:12:27.920 over here, move the broom this way, step over there, 0:12:28.080 --> 0:12:29.680 you know, and if it's not cleaning the house because 0:12:29.679 --> 0:12:31.840 they're stuck on a corner or they're you know, fell 0:12:31.880 --> 0:12:33.960 on their on their butts or whatever, they don't care. 0:12:34.000 --> 0:12:36.280 They just tell you do exactly what you tell them 0:12:36.320 --> 0:12:39.120 to do. Have no sort of larger sense of what's important. 0:12:39.280 --> 0:12:43.480 It just follows instructions, just follows the recipe you gave it. 0:12:43.640 --> 0:12:45.360 That's right. It's like a like a wind up toy, 0:12:45.520 --> 0:12:47.760 you know, you wind it up, you give it some energy, 0:12:47.840 --> 0:12:49.640 and then it goes. And I really do think about 0:12:49.640 --> 0:12:52.959 computer programs the way you might think about little machines, right, 0:12:53.080 --> 0:12:55.679 because that's exactly what they are. They just execute a 0:12:55.720 --> 0:12:57.800 set of instructions. You know. It's just like a bunch 0:12:57.800 --> 0:13:00.760 of gears clicking into place, and they can't change the 0:13:00.800 --> 0:13:02.480 way they do that. And they do it regardless of 0:13:02.480 --> 0:13:04.720 whether it's the right thing, or whether it's effective or whatever. 0:13:04.760 --> 0:13:09.760 It just goes. Like your electric toothbrush, you know, you 0:13:09.800 --> 0:13:12.280 switch it on, and it's just it has a circuit 0:13:12.320 --> 0:13:15.319 that just has it moved the bristles back and forth, 0:13:15.520 --> 0:13:17.240 that's right, And it doesn't know if it's brushing your 0:13:17.240 --> 0:13:19.960 teeth or just flailing around in midair. Right, has no idea, 0:13:20.040 --> 0:13:22.120 It doesn't care, It doesn't think or feel whatever. It's 0:13:22.160 --> 0:13:26.240 just a machine, right, Thank god, it doesn't know. It 0:13:26.280 --> 0:13:32.240 would tell you to brush your teething like that's chocolate, Jorney, 0:13:32.320 --> 0:13:35.920 I'm tired of this? What is this gunk? Yeah? Exactly. 0:13:35.920 --> 0:13:37.960 And so that's what a sort of a normal machine is. 0:13:37.960 --> 0:13:40.000 That's what a classical computer program is, right. I think 0:13:40.000 --> 0:13:41.360 of it just the same way as you think of 0:13:41.360 --> 0:13:44.920 a physical machine. Okay, it's just doing what you the 0:13:44.960 --> 0:13:48.080 programmer told it to do. That's right, and it follows 0:13:48.080 --> 0:13:51.840 your instructions exactly. Um. Now, a computer that can learn 0:13:52.040 --> 0:13:54.920 is different, right. A computer that has artificial intelligence is 0:13:54.960 --> 0:13:57.959 different in this really important way because you can train 0:13:58.040 --> 0:14:01.280 it right, and you can train it because you we 0:14:01.400 --> 0:14:05.439 build these things to model the way that we work. Right. 0:14:05.520 --> 0:14:09.440 So for example, an AI program is sort of like, um, 0:14:09.600 --> 0:14:12.680 like a newborn baby can't do anything. Right. Say, there's 0:14:12.679 --> 0:14:16.800 a AI program, for example, that's supposed to recognize you 0:14:16.960 --> 0:14:18.760 when you come in the door. Right, is this Jorge 0:14:18.920 --> 0:14:21.480 or is this not Jorge? Right? Because it should only 0:14:21.680 --> 0:14:23.840 open the door for Jorge and not open the door 0:14:23.960 --> 0:14:27.320 for not Jorge. Okay, So when you created a new 0:14:27.360 --> 0:14:30.720 AI program, you would start out like just a newborn baby, okay, 0:14:30.800 --> 0:14:33.720 and like a blank slate, right, Like a blank slate, 0:14:33.720 --> 0:14:36.640 it would make random decisions, right you You you show 0:14:36.640 --> 0:14:38.880 it a face and it would say yes, it's Jorge, 0:14:39.160 --> 0:14:41.520 and then you say no, you were wrong or yes 0:14:41.560 --> 0:14:44.160 you were right. And then you would reward it if 0:14:44.160 --> 0:14:46.600 it does well, if it gives the right answer, and 0:14:46.760 --> 0:14:49.880 you would um, you would punish it if it doesn't, right, 0:14:49.920 --> 0:14:51.960 you would tell it. I mean, you don't actually punish 0:14:51.960 --> 0:14:53.960 it a reward. You just tell it, yes, you made 0:14:53.960 --> 0:14:56.320 the right um call this time, and know you made 0:14:56.320 --> 0:14:58.840 the wrong call this time this other time. But how 0:14:58.920 --> 0:15:01.640 is that different than the of calibrating something? Do you 0:15:01.680 --> 0:15:05.640 know what I mean? Like is calibration than artificial intelligence? Right? Well, 0:15:05.680 --> 0:15:09.240 the difference is calibration is like here, I have a tool. 0:15:09.440 --> 0:15:10.960 I know how to solve the problem. I just have 0:15:11.000 --> 0:15:13.600 to adjust it so that it does exactly the right thing. 0:15:13.880 --> 0:15:16.640 Um here, right, But you have a strategy that it's executing. 0:15:17.160 --> 0:15:19.920 You know. It's like, uh, you have a drill and 0:15:20.040 --> 0:15:21.880 you want it to drill fast or slow, and you 0:15:21.920 --> 0:15:23.480 know you know what how to solve the problem. You 0:15:23.560 --> 0:15:25.360 just you know it has to spin and screw and 0:15:25.440 --> 0:15:27.960 screw the thing in or whatever. It's just adjusting a 0:15:28.040 --> 0:15:30.440 knob here. You don't know how to solve the problem, 0:15:30.720 --> 0:15:33.720 and so you've given it a very very very flexible 0:15:34.040 --> 0:15:36.960 strategy on on the inside. You've given it like imagine 0:15:37.000 --> 0:15:40.840 something has like a thousand knobs. If you twist all 0:15:40.880 --> 0:15:44.680 these knobs, you could get all sorts of crazy um strategies. Right. 0:15:44.720 --> 0:15:47.200 So back to the example of like recognizing Joge or 0:15:47.280 --> 0:15:50.720 not when it when you tell it it's done, it's 0:15:50.720 --> 0:15:54.280 given the wrong answer, then it it adjusts those knobs. 0:15:54.280 --> 0:15:56.400 It says, well, let me try to tweak my strategy 0:15:56.440 --> 0:15:58.960 for deciding is this Jorge, and then we'll see how 0:15:58.960 --> 0:16:01.640 that goes. I think that's a key difference. It's the 0:16:01.760 --> 0:16:04.480 number of knobs, right, Like a drill with the knob 0:16:04.680 --> 0:16:08.040 for velocity. I mean, that is sort of trainable and 0:16:08.080 --> 0:16:10.640 you can set it up to be adaptive. But it's 0:16:10.680 --> 0:16:13.720 just one knob, and so it's not You wouldn't say 0:16:13.760 --> 0:16:16.720 it's intelligent. It's not intelligent. The spectrum of things it 0:16:16.760 --> 0:16:19.280 can do is very very seene right. But whereas like 0:16:19.360 --> 0:16:23.120 something that recognizes a phase. It needs to evaluate like 0:16:23.160 --> 0:16:26.440 a million pixels in a photo, right, and so for 0:16:26.480 --> 0:16:29.160 you to tweak how it evaluates each of those pixels, 0:16:29.200 --> 0:16:31.840 it would be really difficult for you to That's right. 0:16:31.880 --> 0:16:34.000 So imagine you know the machine here is is a 0:16:34.080 --> 0:16:35.520 camera in the door and takes a picture of who 0:16:35.600 --> 0:16:37.440 you got a million pixels and then it has to 0:16:37.480 --> 0:16:39.680 look at those pixels and decide is this Jorge or 0:16:39.720 --> 0:16:42.960 is this not whohe? And so there's does some calculation 0:16:43.080 --> 0:16:45.880 on that picture, right, and that calculation has millions of 0:16:45.960 --> 0:16:48.280 knobs on it, right, how much do I weigh this pixel? 0:16:48.320 --> 0:16:50.680 How much I weigh adjacent pixels? Do I look for 0:16:50.720 --> 0:16:52.440 his nose? Do I look for his hair? Do I 0:16:52.480 --> 0:16:55.240 look for the eyes? Right? So it's got some very 0:16:55.320 --> 0:16:57.960 very flexible thing inside of it that can do almost anything. 0:16:58.080 --> 0:17:00.080 And when you first start out, it's just random. So 0:17:00.120 --> 0:17:03.760 it's making ridiculous, terrible decisions. But the key the thing 0:17:03.800 --> 0:17:06.800 that models the learning, right, you know, just need artificial intelligence, 0:17:06.800 --> 0:17:10.000 You need artificial learning. The thing that models that learning 0:17:10.280 --> 0:17:12.680 is that when it gets the wrong answer, it knows 0:17:12.760 --> 0:17:15.720 how to adjust those knobs so that next time it's 0:17:15.760 --> 0:17:18.719 more correct. By itself. That's the key thing is that 0:17:18.760 --> 0:17:22.560 it learns by itself. It doesn't need you. They're sitting like, oh, 0:17:22.680 --> 0:17:24.920 you got this pixel wrong. You've got that pixel wrong. 0:17:25.240 --> 0:17:27.800 A tweak you know, tweak this one this way. It's 0:17:27.840 --> 0:17:32.080 really more like an autonomous automatic learning. That's right because 0:17:32.080 --> 0:17:33.560 you don't know how to adjust it. If you knew 0:17:33.560 --> 0:17:35.520 how to adjust it, you would just write that program. Right. 0:17:35.760 --> 0:17:38.879 The key is artificial intelligence is excellent when you don't 0:17:38.920 --> 0:17:41.320 know how to solve the problem, but you can define 0:17:41.359 --> 0:17:43.480 the problem. You can say, this is a picture of 0:17:43.600 --> 0:17:46.200 or and this is not learn a way to tell 0:17:46.320 --> 0:17:48.800 the difference, right. So you give it a very flexible 0:17:48.800 --> 0:17:52.160 strategy and then you try You let it try out, 0:17:52.240 --> 0:17:54.320 and when it gives it the wrong answer, you would 0:17:54.359 --> 0:17:56.440 let it adjust itself so that it gets closer and 0:17:56.480 --> 0:17:59.480 closer to giving the right answer right, And eventually these 0:17:59.520 --> 0:18:02.560 things will find the right setting for those millions of knobs, 0:18:02.960 --> 0:18:05.040 so that it's doing the right thing. It's saying, oh, look, 0:18:05.240 --> 0:18:07.200 this picture is a picture of Horhey, and it gets 0:18:07.200 --> 0:18:09.760 the right answer at the time. And when you give 0:18:09.760 --> 0:18:11.399 it a picture that's not a picture of Horne, and 0:18:11.400 --> 0:18:13.639 you give it Daniel, it says no, sorry, you're not 0:18:13.680 --> 0:18:16.600 getting in the house. Right. And I think a key 0:18:16.640 --> 0:18:19.320 thing is also that you, as a programmer, could not 0:18:19.600 --> 0:18:22.800 have predicted what all those knobs are going to be 0:18:22.880 --> 0:18:24.960 at the end, right, Like, it's such a big problem. 0:18:25.000 --> 0:18:28.520 There's a million knobs. There's no way that you can 0:18:28.520 --> 0:18:30.359 predict what those knots are going to be set to 0:18:30.640 --> 0:18:33.199 when it learns my face. That's right. It's perfect for 0:18:33.280 --> 0:18:36.520 really hard problems where we don't know how to solve it, right. 0:18:37.040 --> 0:18:38.840 We know how to describe the problem, but we don't 0:18:38.840 --> 0:18:40.640 know how to solve it. You're right, So if I 0:18:40.680 --> 0:18:42.119 already knew how to solve it, I could write a 0:18:42.119 --> 0:18:44.840 computer program that that and and tell it just like 0:18:45.040 --> 0:18:47.399 use this pixel, use that pixel, use this pixel. But 0:18:47.440 --> 0:18:50.000 I don't know how to solve that problem. It's really hard, right, 0:18:50.160 --> 0:18:52.920 But I can train a computer to figure it out, 0:18:53.440 --> 0:18:56.800 just the same way I can train a dog. Right. 0:18:56.920 --> 0:18:59.880 A dog can learn my face. Right, a dog recognizes 0:18:59.880 --> 0:19:03.000 the owner, and you know, happily licks its face when 0:19:03.000 --> 0:19:05.840 it comes home, and recognize that that when somebody's not 0:19:05.960 --> 0:19:08.080 its owner, and barks like crazy and choose its face 0:19:08.080 --> 0:19:10.679 off when it's not its owner, Right, remind me not 0:19:10.840 --> 0:19:18.679 to visit your house, Daniel, seems a little dangerous. So 0:19:18.720 --> 0:19:22.239 then a big thing is programming a structure in the 0:19:22.359 --> 0:19:26.320 in the software that is kind of open ended and malleable, 0:19:26.440 --> 0:19:28.280 do you know what I mean? Like, its something that 0:19:28.480 --> 0:19:31.840 is kind of unpredictable in a way that can learn. 0:19:32.119 --> 0:19:34.040 That's right. And that's the key thing is that some 0:19:34.040 --> 0:19:35.760 people might be thinking, well, hold on, you said that 0:19:35.760 --> 0:19:37.879 the computers can just do what they tell you, So 0:19:37.920 --> 0:19:40.120 how can a computer learn? Right? How is that possible? 0:19:40.600 --> 0:19:42.760 And the key is that it's an emergent property, right. 0:19:43.280 --> 0:19:45.600 Like the way that you write a computer program that 0:19:45.640 --> 0:19:48.960 can learn is you build all these little um calculating 0:19:49.000 --> 0:19:51.800 bits with knobs on them, right, and each bit just 0:19:51.840 --> 0:19:54.560 does what it's told. It takes some data, it makes 0:19:54.600 --> 0:19:56.359 a decision based on the value of the knob, and 0:19:56.400 --> 0:19:59.760 it sends out some data. And together all these things 0:19:59.800 --> 0:20:02.440 may a decision. Right. Each individual piece has no idea 0:20:02.480 --> 0:20:04.640 what it's doing. It's not smart or intelligent or making 0:20:04.640 --> 0:20:07.480 its own decisions. That doesn't have free will, right, But 0:20:07.600 --> 0:20:11.119 together they're doing something. And as you said earlier, they 0:20:11.160 --> 0:20:13.639 can change the way they behave. They can adjust these 0:20:13.720 --> 0:20:16.680 knobs themselves to improve their performance, and that's where the 0:20:16.760 --> 0:20:19.520 learning comes from. It's from that training. It gets external 0:20:19.560 --> 0:20:22.960 input and changes its behavior based on that expe. You're 0:20:23.000 --> 0:20:25.439 saying that the way to program these AI s is 0:20:25.520 --> 0:20:30.119 by h connecting a bunch of little simple things together 0:20:30.400 --> 0:20:33.480 to get something complex. Yes, And here it's important to 0:20:33.520 --> 0:20:36.080 remember that we are using neural networks as sort of 0:20:36.080 --> 0:20:39.480 a standing to represent a big broad set of strategies 0:20:39.520 --> 0:20:42.120