WEBVTT - Does Your Computer Know Chinese? 0:00:00.160 --> 0:00:07.200 Brought to you by Toyota. Let's go places. Welcome to 0:00:07.400 --> 0:00:15.120 Forward Thinking. Hey there, and welcome to Forward Thinking. The 0:00:15.160 --> 0:00:17.520 podcast at looks at the future and says Confucius has 0:00:17.560 --> 0:00:21.239 a puzzling grace. I'm Jonathan Strickland, I'm La, and I'm 0:00:21.320 --> 0:00:25.040 Joe McCormick. So, um, we were talking like intelligent people 0:00:25.079 --> 0:00:31.080 just now, right, like intelligent people. Sure depends on how 0:00:31.120 --> 0:00:36.200 you define intelligence. Maybe people not. We're not all geniuses 0:00:36.240 --> 0:00:38.360 around here, but we have sort of what you might 0:00:38.440 --> 0:00:42.319 think of as a human intelligence. We're smarter than like 0:00:42.560 --> 0:00:45.839 pigs maybe, or like grasshoppers and stuff like that. I 0:00:45.880 --> 0:00:50.320 flatter myself to be considered smarter than my toaster most days. Yeah, 0:00:50.479 --> 0:00:53.000 so this actually got us into a discussion about being 0:00:53.040 --> 0:00:57.160 smarter than inanimate objects, and so far we're doing quite well. Yeah. 0:00:57.720 --> 0:01:02.640 I want y'all to imagine a scenario. Okay, imagine that 0:01:02.720 --> 0:01:09.000 you have a real good friend Ardia already stretching imagination already. Well, 0:01:09.120 --> 0:01:11.840 I know you don't have good friends, but we just 0:01:12.120 --> 0:01:15.560 pretend you have a really good relationship. But this guy 0:01:15.680 --> 0:01:20.560 named Artie, and he's funny, intelligent, compassionate. Um, not to 0:01:20.600 --> 0:01:23.800 mention the strongest man in the world. Uh. But you 0:01:23.920 --> 0:01:28.200 have been extremely good friends with Artie for many, many years. Um. 0:01:28.280 --> 0:01:30.800 And suddenly one day, while you and Artie are hanging out, 0:01:30.920 --> 0:01:34.080 maybe having some coffee in your living room, a bunch 0:01:34.080 --> 0:01:37.000 of people in lab coats bust into your living room 0:01:37.400 --> 0:01:41.680 and they deactivate Alreadie with a remote control. Uh. And 0:01:41.920 --> 0:01:45.600 the head scientist among this group, she flares her lab 0:01:45.680 --> 0:01:49.000 coat and informs you that Already is in fact and 0:01:49.200 --> 0:01:52.960 artificially intelligent robot. And to prove it to you, she 0:01:53.080 --> 0:01:55.520 opens up Already's head and shows you that it is 0:01:55.640 --> 0:01:59.600 full of dense circuitry where a normal person's brain would be. Pop. 0:01:59.720 --> 0:02:03.440 Quiz is what's your reaction in this scenario? Are you 0:02:03.720 --> 0:02:07.400 a horrified that you were tricked into a fake relationship 0:02:07.520 --> 0:02:11.679 with fake intelligence or are you be amazed that a 0:02:11.960 --> 0:02:16.480 synthetic intelligence is capable of being a real friend. I'm 0:02:16.480 --> 0:02:18.560 gonna go with Cee. I'm amazed they were able to 0:02:18.600 --> 0:02:21.440 burst into my living room because there's no doors, so 0:02:22.360 --> 0:02:26.959 live us through the walls. So I got like a 0:02:27.040 --> 0:02:29.079 wrecking ball comes in and it's got a bunch of 0:02:29.200 --> 0:02:32.920 lab coat people clinging to it, and so like scientist 0:02:33.040 --> 0:02:38.200 versions of Miley Cyrus. Maybe I think that I am 0:02:38.360 --> 0:02:41.639 d upset by this entire situation and that someone has 0:02:41.720 --> 0:02:44.760 just deactivated my friend. I think that I'm concerned about 0:02:44.840 --> 0:02:47.280 artist personally, which probably puts me into a B camp 0:02:47.400 --> 0:02:50.359 if you're if you're really being technical, Okay. But so yeah, 0:02:51.240 --> 0:02:53.919 I it's hard to know how I would feel, But 0:02:55.120 --> 0:02:56.720 I would like to say I think i'd go with B. 0:02:57.600 --> 0:03:00.320 That I wouldn't just be horrified and say like all 0:03:00.400 --> 0:03:04.960 of those years were just a yeah, Now, I've had 0:03:05.040 --> 0:03:07.400 enough of those in in my life from from people who, 0:03:07.680 --> 0:03:10.120 as far as I know, are not robots. So I 0:03:10.200 --> 0:03:13.240 think I would mostly be amazed that someone had achieved 0:03:13.280 --> 0:03:17.760 the technical ability to simulate intelligence to the extent where 0:03:17.880 --> 0:03:21.560 I could not tell the difference between what is simulated 0:03:21.600 --> 0:03:24.639 intelligence and real intelligence. Okay. And so this ties into 0:03:24.680 --> 0:03:26.880 a concept we've talked about on the show before, which 0:03:27.000 --> 0:03:31.040 is the Turing test. What what is the Turing tests? Essentially, 0:03:31.240 --> 0:03:33.639 the basic version of the test is that you have 0:03:34.200 --> 0:03:38.160 a person positing questions over a terminal, so a computer terminal. 0:03:38.600 --> 0:03:43.080 They cannot see whomever or whatever is answering those questions. 0:03:43.600 --> 0:03:46.360 Those questions are being answered either by a person in 0:03:46.440 --> 0:03:49.360 another room at another terminal, who's typing those answers in 0:03:49.960 --> 0:03:53.760 or by a computer program that is responding automatically to 0:03:53.960 --> 0:03:57.680 those questions. If the person asking the questions is unable 0:03:57.800 --> 0:04:02.760 to determine with any real level of of of certain certainty, 0:04:02.840 --> 0:04:06.440 thank you words that don't come to mind certainty, then 0:04:07.320 --> 0:04:09.520 that machine is said to pass the Touring test if 0:04:09.560 --> 0:04:12.280 in fact it was the machine that was answering those questions. Okay, 0:04:12.320 --> 0:04:15.240 And so that's the basic Turing tests, say the back 0:04:15.280 --> 0:04:18.280 and forth text based chatbot. But you could sort of 0:04:18.360 --> 0:04:20.680 look at the Turing test in a larger view and say, 0:04:20.720 --> 0:04:24.680 it's the bigger question of can you simulate human intelligence 0:04:24.760 --> 0:04:26.800 in a way that people can't tell the difference? And 0:04:26.880 --> 0:04:29.080 in fact, Touring went so far as to say that 0:04:29.400 --> 0:04:33.440 if this machine was able to simulate human intelligence to 0:04:33.560 --> 0:04:37.320 that level, we would, in fact, we should, in fact 0:04:37.400 --> 0:04:39.960 say that it is intelligent in the sense that we 0:04:39.960 --> 0:04:44.760 should grant it the same kind of consideration that we do. Right, 0:04:44.800 --> 0:04:49.520 because because we cannot inhabit another person's being, we cannot 0:04:49.560 --> 0:04:54.640 be certain that that person actually possesses intelligence. We assume, yeah, 0:04:54.920 --> 0:04:56.600 we assume that they have the same things that we 0:04:56.839 --> 0:04:58.560 You know that they have the same things that we have, 0:04:59.279 --> 0:05:01.880 so we should extend the same courtesy to machines if 0:05:01.880 --> 0:05:03.680 they are capable of doing such a thing. Right, So 0:05:03.880 --> 0:05:07.320 this is what's known as sort of strong AI. Actually 0:05:07.520 --> 0:05:11.160 this is really known as weak AI because it's simulating intelligence, 0:05:11.320 --> 0:05:14.960 not necessarily possessing intelligence. Well, that what the question is 0:05:15.040 --> 0:05:18.920 about strong AI. The question is is um something that 0:05:19.200 --> 0:05:24.720 displays convincingly intelligent behavior actually intelligent. And here's where we 0:05:24.800 --> 0:05:29.240 come back to Artie. Okay, So Artie, your friend convinced 0:05:29.320 --> 0:05:32.200 you for years that he was a real intelligence. He 0:05:32.279 --> 0:05:35.320 seemed intelligent, he was, you know, funny, He was a 0:05:35.440 --> 0:05:39.240 hundred indistinguishable from every other human you knew. And when 0:05:39.320 --> 0:05:42.040 he was activated, he would insist that he was in 0:05:42.160 --> 0:05:46.720 fact conscious, understanding language, would insist on all the other 0:05:46.839 --> 0:05:49.600 things that we would insist on as part of our 0:05:50.000 --> 0:05:56.480 sort of mental abilities. Was already actually intelligent and did 0:05:56.640 --> 0:06:01.760 already understand the relationship the language, all of these things 0:06:01.920 --> 0:06:05.200 that happened between humans. Or was he merely running programs 0:06:05.279 --> 0:06:07.520 that made it seem as though he was right? So, 0:06:07.640 --> 0:06:10.160 in other words, this is the difference between we weak 0:06:10.200 --> 0:06:12.880 AI and strong AI. Week A I would be running 0:06:12.920 --> 0:06:15.680 programs where you would be taking in various kinds of 0:06:15.720 --> 0:06:20.160 input and producing output that to an independent observer would 0:06:20.160 --> 0:06:22.880 seem to be this kind of natural processing of information 0:06:22.960 --> 0:06:25.160 and reaction in the way a human being would react 0:06:25.600 --> 0:06:27.880 week at week A. I would be the position that 0:06:28.040 --> 0:06:31.240 even though you could simulate it, it wasn't really intelligence, right, 0:06:31.480 --> 0:06:34.320 and strong AI would be that it actually has a 0:06:34.560 --> 0:06:37.520 level of intelligence that's comparable to human intelligence. It might 0:06:37.600 --> 0:06:41.200 not be identical, but it would be. It would truly understand, 0:06:41.360 --> 0:06:44.680 It would know what it knew, and not just be 0:06:44.880 --> 0:06:49.800 able to process information based on some complex series of rules. Right. 0:06:50.200 --> 0:06:53.800 So can a robot or a computer program, any kind 0:06:53.880 --> 0:06:59.200 of machine that's based on symbolic programming and computation actually 0:06:59.360 --> 0:07:02.720 no or understand anything. And this is a question that 0:07:02.960 --> 0:07:07.880 is debated quite heavily in artificial intelligence circles and cognitive science. Yeah, 0:07:07.960 --> 0:07:10.440 what we want to focus on today is one particular 0:07:10.560 --> 0:07:14.240 thought experiment that has tried to answer this question. And 0:07:14.800 --> 0:07:18.400 I would say that this is probably the most hotly 0:07:18.520 --> 0:07:23.680 debated question in AI and cognitive science in the past 0:07:23.960 --> 0:07:26.720 few decades, well at least the past decades certainly. Yeah, 0:07:26.920 --> 0:07:31.040 we're talking about John Searle's Chinese Room thought experiment, which 0:07:31.120 --> 0:07:35.080 is a really uh interesting approach. And there are some 0:07:35.720 --> 0:07:39.600 also interesting counter arguments to the Chinese room thought experiment 0:07:39.640 --> 0:07:41.920 that we'll have to kind of talk about. Yeah. As 0:07:42.000 --> 0:07:46.920 of about, computer scientist Pat Hayes had defined cognitive cognitive 0:07:46.960 --> 0:07:49.560 science as a field as the ongoing research project of 0:07:49.640 --> 0:07:54.120 refuting searles argument. Um. Yeah, So obviously this is a 0:07:54.280 --> 0:07:57.800 huge issue, hotly debated by people a lot smarter than 0:07:57.880 --> 0:07:59.480 we are and who have thought about it a lot more. 0:07:59.560 --> 0:08:01.720 So we're not going to settle it here today. We're 0:08:01.800 --> 0:08:04.040 just going to lay it out. We were just going 0:08:04.120 --> 0:08:06.280 to try to explain it and offer a few of 0:08:06.400 --> 0:08:08.960 the prominent thoughts in this field. Let me give a 0:08:09.160 --> 0:08:12.600 crack at explaining what this thought experiment is. Okay, okay, Well, 0:08:12.680 --> 0:08:14.560 first of all, just let's know where it came from. 0:08:14.640 --> 0:08:17.800 It was from John Searle in his nineteen eighty paper Minds, 0:08:17.960 --> 0:08:21.840 Brains and Programs, and that was published in the Behavioral 0:08:21.920 --> 0:08:24.320 in Brain Science. He has he has updated it a 0:08:24.360 --> 0:08:27.800 couple of times since then, uh, and also responded to 0:08:27.960 --> 0:08:31.080 various criticisms leveled against his argument. But here is the 0:08:31.200 --> 0:08:36.040 basic thought experiment now, imagine that you, Joe. I'm going 0:08:36.080 --> 0:08:37.640 to use you as an example. The Joe, you don't 0:08:37.720 --> 0:08:40.679 happen to know Chinese, do you know? Okay, well, then 0:08:40.679 --> 0:08:42.760 you're a perfect I wanted to use you as an example. 0:08:42.840 --> 0:08:44.559 It's too late. I'm the one giving the example, so 0:08:44.840 --> 0:08:47.559 you are the example. I also don't know Chinese, so 0:08:47.679 --> 0:08:49.440 either one of us would work in Okay, you can 0:08:49.559 --> 0:08:52.959 use me. Okay, So Joe does not know Chinese. Uh, 0:08:53.120 --> 0:08:56.360 he is put into a room. That room has a desk, 0:08:56.520 --> 0:08:59.520 it's got some paper, some pencils or racers, it's got 0:08:59.559 --> 0:09:01.679 some file in cabinets and as a door with a 0:09:01.760 --> 0:09:04.920 slot in it. And the way that Joe interacts with 0:09:04.960 --> 0:09:08.680 the outside world is that occasionally someone puts a piece 0:09:08.720 --> 0:09:11.600 of paper through that slot that has a Chinese character 0:09:11.800 --> 0:09:14.800 drawn on it. Now, Joe also has an enormous book 0:09:15.000 --> 0:09:18.120 full of instructions on what to do every time a 0:09:18.240 --> 0:09:20.480 piece of paper with a character comes in and he 0:09:20.559 --> 0:09:23.280 looks at the character, goes through the book, finds the 0:09:23.360 --> 0:09:25.920 character in the book, follows the instructions on what he's 0:09:25.920 --> 0:09:28.840 supposed to do when that character is inserted, which usually 0:09:28.880 --> 0:09:31.679 involves him drawing a different Chinese character on a on 0:09:31.720 --> 0:09:34.000 a sheet of paper and then slaying that back through 0:09:34.080 --> 0:09:38.079 the slot. Now from anyone who is outside of that rooms, 0:09:38.200 --> 0:09:42.000 a Chinese speaking person who is writing down a specific 0:09:42.080 --> 0:09:44.640 instruction on a sheet of paper and slides it through 0:09:44.720 --> 0:09:47.080 that slot and then gets the result that Joe has 0:09:47.160 --> 0:09:50.400 driven written on the other sheet of paper. To them, 0:09:50.480 --> 0:09:53.920 it appears that whatever is inside that room understands Chinese 0:09:54.480 --> 0:09:58.560 right from their perspective, if if the instructions are good enough, yes, right. 0:09:58.640 --> 0:10:00.679 So what we're sort of a mad jenning is this 0:10:00.840 --> 0:10:03.000 is a room with a person in it that passes 0:10:03.080 --> 0:10:06.319 the Turing test. Yes, yeah, because you the person on 0:10:06.400 --> 0:10:08.800 the outside has no idea what's on the inside of 0:10:08.800 --> 0:10:11.280 the room. All they know is that they have passed 0:10:11.320 --> 0:10:14.480 in instructions that were in Chinese. They've received output that's 0:10:14.520 --> 0:10:18.880 also in Chinese, and everything follows according to a sets 0:10:18.960 --> 0:10:21.880 set of rules, predetermined rules, and and the output is 0:10:22.000 --> 0:10:25.319 precise and convincing enough that the person outside thinks that 0:10:25.520 --> 0:10:29.559 whatever is inside is a true native Chinese speaker. So 0:10:29.880 --> 0:10:33.599 so the question then in this experiment is does this 0:10:33.760 --> 0:10:39.280 mean that Joe actually knows Chinese? And Searles, uh, Well, 0:10:39.320 --> 0:10:42.120 Searles says no, you don't. You don't know Chinese. You 0:10:42.160 --> 0:10:45.400 don't understand Chinese. You're following instructions, but you don't understand 0:10:45.440 --> 0:10:48.160 the actual language. Actually, I might make one changes. I 0:10:48.200 --> 0:10:50.800 think it's the obvious point is that I and that 0:10:50.960 --> 0:10:54.520 thought experiment don't actually know Chinese. The question is does 0:10:54.640 --> 0:10:58.679 this translate to a computer in the same situation, because 0:10:59.240 --> 0:11:01.520 this with this supposed to be an analogy for the 0:11:01.600 --> 0:11:04.680 situation that a computer or turing machine is in sure, 0:11:04.920 --> 0:11:08.880 it receives input, um it has a set of instructions 0:11:08.960 --> 0:11:13.760 that it follows in in computer readable language. It processes 0:11:14.000 --> 0:11:17.480 that input through those instructions, and then it gives output. 0:11:18.040 --> 0:11:21.599 But it doesn't have any actual understanding of what is 0:11:21.679 --> 0:11:24.040 going on at any given time. It it knows to 0:11:24.200 --> 0:11:27.520 follow specific instructions because of a program, but it doesn't 0:11:27.679 --> 0:11:30.280 understand what's happened. It doesn't understand the question more or 0:11:30.360 --> 0:11:32.920 less much less the answer, right. And so that was 0:11:33.040 --> 0:11:36.440 Searle's point, and it wasn't He wasn't against week Ai. 0:11:36.600 --> 0:11:40.520 I mean, he obviously believed that maybe a computer could 0:11:40.640 --> 0:11:43.760 simulate human intelligence so well that we couldn't tell the difference. 0:11:44.240 --> 0:11:47.160 He was just pointing out that UM it seems that 0:11:47.440 --> 0:11:54.199 based on simple programmable instructions and symbolic computation, that a 0:11:54.280 --> 0:11:59.000 computer could never really understand but it was dealing with. Yeah, 0:11:59.120 --> 0:12:04.160 so understanding is the crux of this argument. And also 0:12:04.880 --> 0:12:09.719 it's problematic because often people who level criticisms against it 0:12:10.120 --> 0:12:13.079 say that we need to define what understand means, and 0:12:13.160 --> 0:12:14.960 then it gets a little metaphysical. But we're not going 0:12:15.000 --> 0:12:17.199 to go too far into that because obviously that that 0:12:17.240 --> 0:12:19.719 can take hours of discussion. Well, but I mean, part 0:12:19.800 --> 0:12:21.920 of this problem is the is the semantics of the 0:12:22.040 --> 0:12:24.160 terminology that we use for this kind of thing, And 0:12:24.400 --> 0:12:27.120 and which is a little bit funny because because syntax 0:12:27.280 --> 0:12:30.959 versus semantics is considered um kind of the key point 0:12:31.120 --> 0:12:34.120 of this argument, right, Since what the computer clearly does 0:12:34.280 --> 0:12:38.439 understand is syntax, it's got instructions and data to deal with. 0:12:38.960 --> 0:12:41.840 What's up for debate is whether it gets semantics, whether 0:12:41.920 --> 0:12:45.040 it understands the meaning of the symbol. So, for example, 0:12:45.120 --> 0:12:49.160 we've talked about the semantic web, this idea of a web, 0:12:49.440 --> 0:12:51.840 some way of interacting with the Worldwide Web. I love 0:12:51.920 --> 0:12:54.640 to use that old term because I was around when 0:12:54.640 --> 0:12:57.720 it started, but a way to interact with the web 0:12:57.880 --> 0:13:00.959 where it quote unquote understand ends what you want and 0:13:01.000 --> 0:13:04.320 it's pulling exactly the stuff that you want. But when 0:13:04.360 --> 0:13:07.040 we talk about actually making the semantic web come to pass, 0:13:07.360 --> 0:13:10.640 it's all about things like meta data and including all 0:13:10.720 --> 0:13:13.200 of this information about information so that a computer can 0:13:13.240 --> 0:13:16.600 make sense of it, because a computer does not natively 0:13:16.760 --> 0:13:21.240 possess this ability to understand semantics. And even in that case, 0:13:21.360 --> 0:13:23.120 we would call it the semantic web, but it would 0:13:23.120 --> 0:13:25.840 still be based on program yes, and still be very 0:13:25.920 --> 0:13:28.520 much based on syntax. All right, All the examples that 0:13:28.600 --> 0:13:30.400 we can really come up with are very much based 0:13:30.400 --> 0:13:32.360 in our human world because it's a little bit difficult 0:13:32.360 --> 0:13:36.000 for us to understand, um without metaphor exactly what's going 0:13:36.040 --> 0:13:38.000 on inside of a computer. But it's it's sort of 0:13:38.040 --> 0:13:41.599 talking about the difference between like memorizing versus understanding and 0:13:41.679 --> 0:13:46.880 equation or being tested on memorization versus comprehension. Yeah, okay, 0:13:46.960 --> 0:13:49.839 so well, I want to offer a few of the 0:13:49.920 --> 0:13:52.320 responses I've read about. None of these, of course, are 0:13:52.320 --> 0:13:56.440 original to us. This is something that's been debated over 0:13:56.559 --> 0:13:59.559 and over again. So these are all things that have 0:13:59.640 --> 0:14:02.719 been gust in that philosophy world, and an AI and 0:14:02.800 --> 0:14:05.679 computer science before UM. But I think one of the 0:14:05.760 --> 0:14:08.240 most important things to start with is sort of the 0:14:08.760 --> 0:14:14.040 epistemological question is just going straight back to the Turing skepticism, 0:14:14.480 --> 0:14:18.120 which is that, um, if you can't tell the difference, 0:14:18.400 --> 0:14:21.640 you can't tell the difference. That the idea is, if 0:14:21.760 --> 0:14:25.840 someone appears to you to have understanding, how could you 0:14:25.960 --> 0:14:29.840 say that they don't have understanding? Um? And I think 0:14:29.920 --> 0:14:32.600 that's a really good point to start with. But there 0:14:32.800 --> 0:14:35.240 is also a good response to it, in my mind, 0:14:35.320 --> 0:14:38.240 which is that, um, it still leaves the question of 0:14:38.400 --> 0:14:41.160 what is the basic thing we're talking about? So we 0:14:41.280 --> 0:14:44.240 might not be able to know if other people have understanding, 0:14:44.600 --> 0:14:47.400 but just saying that sentence still acknowledges, well, there is 0:14:47.480 --> 0:14:51.040 such a thing as understanding, and whether or not we 0:14:51.200 --> 0:14:53.840 can know that, the question of whether a program machine 0:14:53.960 --> 0:14:58.720 can in principle attain this state remains. So it's not 0:14:58.880 --> 0:15:02.720 deconstructing the argument, and it's just it's just raising more questions. Yeah, 0:15:02.800 --> 0:15:06.520 and so one point raised by this objection might be that, well, 0:15:06.920 --> 0:15:09.840 what if we can never solve this question, But that's 0:15:09.920 --> 0:15:14.080 not the same as saying that the question is meaningless. Basically, sure, 0:15:14.200 --> 0:15:19.640 another objection or another criticism leveled uses what Searl has 0:15:19.880 --> 0:15:24.360 identified as the system approach. So the system argument is 0:15:24.600 --> 0:15:27.440 about how if you look at the person in the 0:15:27.560 --> 0:15:30.080 room as supposed to you know, the person in the 0:15:30.160 --> 0:15:33.280 room is supposed to represent a computer, then clearly you 0:15:33.320 --> 0:15:37.720 would argue while Joe didn't really learn Chinese, computer being Joe, 0:15:38.920 --> 0:15:41.920 Joe clearly did not learn it. He was following specific 0:15:42.000 --> 0:15:44.280 instructions that were in a giant book that were not 0:15:44.440 --> 0:15:47.560 part of Joe's knowledge. All he was doing was taking 0:15:47.640 --> 0:15:50.240 an input and then applying this set of rules and 0:15:50.280 --> 0:15:53.320 then getting output. But the system approach says, well, you're 0:15:53.320 --> 0:15:56.400 looking at this wrong. You're looking at at the Joe 0:15:56.480 --> 0:15:58.840 at being the representation of the computer, whereas really it 0:15:58.880 --> 0:16:01.440 would be the entire system that you have to take 0:16:01.480 --> 0:16:05.240 into consideration rather than just Joe. Joe would be like 0:16:05.400 --> 0:16:09.360 the processor, the CPU, and that that something to do 0:16:09.480 --> 0:16:13.160 with the room, which has created these instructions, understands Chinese, 0:16:13.200 --> 0:16:16.320 and therefore that that intelligence has happened. So when you 0:16:16.400 --> 0:16:18.800 look at the whole thing together, the room, the English 0:16:18.840 --> 0:16:23.360 instruction manual, me the input and output paper slips that 0:16:23.440 --> 0:16:27.000 are coming through the door, the filing system, which represents memory. Right. 0:16:27.160 --> 0:16:29.720 All of that together, you could say, in a way 0:16:29.960 --> 0:16:33.960 understands Chinese. Right, And it may not be directly analogous 0:16:34.000 --> 0:16:36.640 to the way a person understands Chinese, but that really 0:16:36.760 --> 0:16:39.760 doesn't matter in this argument so much. We're talking about 0:16:40.120 --> 0:16:43.040 strong AI versus weak AI, not whether or not the 0:16:43.120 --> 0:16:48.040 intelligence is exactly like human intelligence. Yeah, and so this 0:16:48.280 --> 0:16:51.880 is probably the I think the most popular response to 0:16:51.960 --> 0:16:54.200 the Chinese room argument. There's some others we're gonna talk 0:16:54.200 --> 0:16:56.280 about in a minute, but I think the dominant one 0:16:56.440 --> 0:17:00.520 is this systems thinking approach. And and this also goes with, 0:17:00.880 --> 0:17:02.840 um a lot of the ways people have come up 0:17:02.880 --> 0:17:05.920 with to think about how intelligence comes out of a 0:17:06.000 --> 0:17:10.200 material mind, sorry, a material brain. Um. So you know, 0:17:10.359 --> 0:17:14.159 it's like you're the brain is not just one discrete 0:17:14.240 --> 0:17:19.720 agent doing one kind of thing, but it's this confabulation 0:17:19.840 --> 0:17:24.239 of tons of different types of events that, in an 0:17:24.280 --> 0:17:28.600 emergent way combined to produce what we think of his consciousness. Right. 0:17:28.880 --> 0:17:32.080 It's also related to another response called the virtual mind reply, 0:17:32.280 --> 0:17:35.160 which which specifies that AI is a machine that has 0:17:35.280 --> 0:17:38.680 created a mind that understands Chinese, not that the AI 0:17:38.840 --> 0:17:43.080 itself understands Chinese. Which is getting back into semantics. Yeah, 0:17:43.080 --> 0:17:45.200 it's getting it's also getting a little metaphysical, to the 0:17:45.280 --> 0:17:49.639 point where Searle's responses puffed. I think that's how you 0:17:49.680 --> 0:17:51.960 would sum it up. But essentially he says like, you 0:17:52.000 --> 0:17:54.080 don't need to bring metaphysics into it. This isn't this, 0:17:54.359 --> 0:17:57.680 that's that's irrelevant. Well, if you want to get really 0:17:57.760 --> 0:18:00.400 hard nosed and practical, there are a couple other ways 0:18:00.480 --> 0:18:02.879 we could go with this. One of them is simply 0:18:02.920 --> 0:18:07.000 the criticism that the thought experiment is too impractical to 0:18:07.080 --> 0:18:09.879 be useful, or it's too unlike the real world. So 0:18:10.000 --> 0:18:12.399 think about it like this. Um It asks us to 0:18:12.560 --> 0:18:14.880 imagine that the person in the room would be able 0:18:14.960 --> 0:18:18.960 to use this English instruction manual, which is the equivalent 0:18:19.000 --> 0:18:23.760 to a computer program, fast enough to generate convincing responses. 0:18:24.119 --> 0:18:27.080 But in reality, it might take a single person in 0:18:27.160 --> 0:18:31.200 a room like billions of years to go through all 0:18:31.359 --> 0:18:35.080 of the instructions in that instruction manual and turn back 0:18:35.119 --> 0:18:38.920 an answer that was satisfactory. A computer could do this 0:18:39.080 --> 0:18:42.240 in less than a second, probably, so Thus there could 0:18:42.280 --> 0:18:45.480 be a fundamental difference between the kind of slow intelligence 0:18:45.560 --> 0:18:49.320 that the thought experiment would display and an actually convincing 0:18:49.440 --> 0:18:52.840 touring ai um, and thus the thought experiment can actually 0:18:52.960 --> 0:18:56.240 comment on what happens with a touring AI. See. I 0:18:57.359 --> 0:19:00.119 I have a problem with this just from a gut level. Obviously, 0:19:00.200 --> 0:19:04.880 I'm not a philosopher nor my scientist. However, I think 0:19:04.960 --> 0:19:08.120 about people who really take their time and answering a question, 0:19:08.800 --> 0:19:11.440 and I think of computers I've used that also really 0:19:11.520 --> 0:19:13.920 take your time and executing a program, And I'm just 0:19:14.000 --> 0:19:17.560 thinking that when it comes down to the time issue, 0:19:17.800 --> 0:19:20.200 I almost see that as irrelevant to me, like that 0:19:20.280 --> 0:19:22.040 doesn't really matter so much to me. But then a 0:19:22.200 --> 0:19:24.560 part of it is also that I'm used to talking 0:19:24.640 --> 0:19:27.879 about stuff that's taking over taking taking place over the 0:19:27.920 --> 0:19:31.639 course of billions of years. So whether it takes you know, 0:19:31.760 --> 0:19:35.800 a month to respond to a request or it takes 0:19:35.960 --> 0:19:40.440 a second, I don't really necessarily see that as an 0:19:40.520 --> 0:19:44.640 indicator or or you know, something that denies the presence 0:19:44.680 --> 0:19:47.560 of intelligence. Yeah, you could certainly say that you know 0:19:47.720 --> 0:19:50.280 that basically no matter how long it takes, the principles 0:19:50.359 --> 0:19:53.240 behind the question in the thought experiment remain the same. 0:19:53.960 --> 0:19:56.760 Um that it's just that's just sort of an irrelevant 0:19:56.880 --> 0:19:59.560 variable that we can exclude for the sake of argument, 0:20:00.119 --> 0:20:03.119 And the question is is it reasonable to exclude it 0:20:03.160 --> 0:20:05.399 for the sake of argument or not? Right, because if 0:20:05.480 --> 0:20:08.560 time is actually a factor, if speed is a factor 0:20:08.720 --> 0:20:11.920 in determining what is is an intelligent then clearly you 0:20:12.000 --> 0:20:14.280 can't just dismiss it. Right, Like if some if for 0:20:14.800 --> 0:20:18.240 some reason, we discover that that the way that we 0:20:18.760 --> 0:20:22.680 exhibit intelligence is dependent partly on the speed at which 0:20:22.800 --> 0:20:26.239 our neural networks work, and then it is. But if 0:20:26.280 --> 0:20:28.399 they did not work at that speed, we would not 0:20:28.560 --> 0:20:30.720 be intelligent, That's what I'm saying. Like, if that were 0:20:30.800 --> 0:20:34.080 the case, then we would you know, well we wouldn't 0:20:34.080 --> 0:20:36.119 be having this argument because we wouldn't be able to. 0:20:36.280 --> 0:20:38.879 But oh, the way you just phrased it made this 0:20:39.480 --> 0:20:42.600 objection suddenly seem more reasonable to me. Actually, Like, what 0:20:42.720 --> 0:20:45.280 if something that did all the same things a human 0:20:45.359 --> 0:20:49.399 brain does, except did them at one one billiont the speed. 0:20:49.920 --> 0:20:53.040 Would that have consciousness in the same way that we do? Well, 0:20:53.119 --> 0:20:54.800 I don't know that we would all ever argue them 0:20:54.840 --> 0:20:57.600 as sument have consciousness the same way. Well, not exactly, 0:20:57.720 --> 0:21:00.159 but you know what I mean, has this intelligence and 0:21:00.280 --> 0:21:03.000 part of this is also part of a separate but 0:21:03.160 --> 0:21:06.800 related argument that talks about exactly how ego centric we're 0:21:06.800 --> 0:21:10.480 being in our definition of intelligence and understanding, because that's 0:21:10.480 --> 0:21:11.879 a huge part of it. I mean, you know, like 0:21:12.000 --> 0:21:15.200 like if if we do we define intelligence based on 0:21:15.520 --> 0:21:18.560 neural connections, because there are things out there who display 0:21:19.200 --> 0:21:23.800 relatively intelligent behavior that don't have that on our planet Earth. Yeah, 0:21:23.960 --> 0:21:28.320 I just think of them as incredibly lucky. They're just 0:21:28.440 --> 0:21:31.760 on a They're just on a series of making their 0:21:31.800 --> 0:21:35.440 saving throw against stupidity. That's all that's happening. Uh. Yeah. 0:21:35.560 --> 0:21:38.520 Let me offer one more objection that's sort of more 0:21:38.600 --> 0:21:40.920 on the practical edge I think, and that's sort of 0:21:41.080 --> 0:21:46.080 the lack of diverse sensory grounding. So system, if a 0:21:46.200 --> 0:21:49.320 system is only receiving input by pieces of paper being 0:21:49.320 --> 0:21:52.720 shoved through a slot, that's extremely limiting. Yeah. Um, and 0:21:52.840 --> 0:21:55.760 so okay. This is also known as the robot response 0:21:55.800 --> 0:21:59.440 to the argument. Yeah so uh. This argument basically says 0:21:59.520 --> 0:22:02.879 that what prevents the person in the room from understanding 0:22:03.000 --> 0:22:06.160 Chinese is the lack of sensory grounding in the outside 0:22:06.200 --> 0:22:10.320 world through diversity of cues like visual, auditory all that. So, 0:22:10.720 --> 0:22:14.920 perhaps a computer designed only to manipulate symbols based on 0:22:15.040 --> 0:22:20.000 instructions wouldn't understand language or wouldn't understand those symbols. But 0:22:20.359 --> 0:22:25.359 if granted diverse sensory and put like video cameras haptic input, Basically, 0:22:25.440 --> 0:22:28.560 if you made that computer into a robot that could 0:22:28.640 --> 0:22:31.320 sample the world and all of the ways that we can, 0:22:31.760 --> 0:22:34.880 the system actually would be able to understand what those 0:22:34.920 --> 0:22:38.920 symbols mean. It's an interesting argument. Uh. Cyril of course 0:22:38.960 --> 0:22:42.159 has his own counter which says that you could have 0:22:42.400 --> 0:22:46.240 a string of numbers that to a person in that 0:22:46.440 --> 0:22:49.639 room seemed to be meaningless, but are actually the readoubts 0:22:49.720 --> 0:22:52.119 that a result of a video camera that's set up 0:22:52.520 --> 0:22:55.920 pointing outside of the room, and it wouldn't magically make 0:22:56.040 --> 0:22:59.119 things more meaningful to the person inside the room, and 0:22:59.160 --> 0:23:01.479 that it would just create more work for him or her. Right, 0:23:01.640 --> 0:23:04.400 you're just yeah, that's exactly what Sarl says, Like, all 0:23:04.440 --> 0:23:07.480