WEBVTT - Passing the Turing Test 0:00:04.240 --> 0:00:10.640 Get in test with technology with text stuff from dot com. 0:00:12.039 --> 0:00:15.600 Say everyone, and welcome to text Uff. I'm Jonathan Strickland, 0:00:16.960 --> 0:00:20.439 and uh, you know, something hit the news recently. You 0:00:20.480 --> 0:00:22.160 may have seen it, in fact, you may have seen 0:00:22.200 --> 0:00:26.880 it in multiple places about a program, a computer program 0:00:26.960 --> 0:00:30.920 passing the Turing test, possibly the first such program to 0:00:31.000 --> 0:00:34.080 ever pass the Turing test ever. And we had planned 0:00:34.120 --> 0:00:37.000 on doing an episode on this, and then on top 0:00:37.040 --> 0:00:40.720 of that, one of our listeners, Nick on Twitter said, 0:00:40.880 --> 0:00:43.280 just saw this headline on Google News. A computer just 0:00:43.400 --> 0:00:46.840 passed the Turing test in landmark trial. So we knew 0:00:46.840 --> 0:00:49.879 that the timing was perfect. Thank you Nick for writing in. 0:00:50.640 --> 0:00:52.800 We are really ready to talk about this, but in 0:00:52.880 --> 0:00:55.880 order to do that, it's important that we, you know, 0:00:56.000 --> 0:00:59.120 kind of lay some groundwork. So first of all, the 0:00:59.120 --> 0:01:03.040 Touring test is named after Alan Turing, and we've done 0:01:03.040 --> 0:01:06.160 a full episode on Alan Turing way back when, back 0:01:06.160 --> 0:01:12.479 in November of Yeah. Phenomenal person, amazing thinker. Yeah, one 0:01:12.520 --> 0:01:17.440 of the like the grandfather of computer science. Also tragic 0:01:17.520 --> 0:01:21.160 life story which we went into detail back in that story. 0:01:21.240 --> 0:01:22.800 And so you might wonder why, right, So what does 0:01:22.840 --> 0:01:25.679 that have to do? Was the guy who was the 0:01:25.760 --> 0:01:28.720 essentially the father of computer science or grandfather of computer science. 0:01:29.080 --> 0:01:30.720 What does he have to do with the story about 0:01:31.160 --> 0:01:35.080 a computer program in June two thousand fourteen, a computer 0:01:35.120 --> 0:01:40.360 program with an interesting name, Eugene Goostman passing the Turing test. 0:01:41.080 --> 0:01:42.920 What does that all have to do with each other? Well? 0:01:43.280 --> 0:01:45.800 To answer that, we have to ask what is a 0:01:45.880 --> 0:01:49.760 Turing test? Well, as it turns out Away, back in 0:01:49.880 --> 0:01:55.800 the nineteen fifties, he started envisioning uh I thought experiment. Yeah, 0:01:55.840 --> 0:01:59.600 he published a paper in a journal called Mind in 0:01:59.680 --> 0:02:03.520 ninety fifty called Computing, Machinery and Intelligence, and he sort 0:02:03.560 --> 0:02:06.680 of laid out his thought experiment there. And it's interesting 0:02:06.680 --> 0:02:08.840 because what he did was he took this idea of 0:02:08.880 --> 0:02:12.320 a party game and then adapted it for computers. Right. 0:02:12.480 --> 0:02:16.800 The party game that it's based on would have three participants, 0:02:16.840 --> 0:02:20.480 an interrogator, a man, and a woman um all situated 0:02:20.480 --> 0:02:23.160 so that they can't see one another, and the interrogator 0:02:23.200 --> 0:02:26.080 is supposed to ask questions to try to figure out 0:02:26.480 --> 0:02:29.400 which of the participants is the dude and which is 0:02:29.400 --> 0:02:32.600 the lady exactly, And it's the dude's job to try 0:02:32.600 --> 0:02:37.200 and mislead the interrogator to believe that that he, in fact, 0:02:37.240 --> 0:02:38.760 is the lady and the other one is the man. 0:02:39.160 --> 0:02:42.160 It's the lady's job to say, hey, I want you 0:02:42.200 --> 0:02:44.440 to get this right. I'm the lady. That other guy 0:02:44.480 --> 0:02:47.760 that's the dude. And so the interrogator has to ask questions. Now, 0:02:48.480 --> 0:02:51.120 obviously they can't see each other. Because they could see 0:02:51.120 --> 0:02:53.920 each other, then that would probably give things away. Probably, 0:02:54.200 --> 0:02:56.800 they really shouldn't be able to hear each other because 0:02:56.800 --> 0:02:59.880 that could also give things away. Sure, and if you 0:03:00.080 --> 0:03:03.399 use handwritten notes, then you the interrogator might be able 0:03:03.440 --> 0:03:06.440 to make judgments based on the handwriting style. So it 0:03:06.440 --> 0:03:08.880 should really be typewritten, right, So you want to you 0:03:08.919 --> 0:03:13.040 want to remove as many easily identifiable traits from this 0:03:13.080 --> 0:03:15.880 game as possible to make it all about the questions 0:03:15.919 --> 0:03:19.480 and the answers. Now, Touring said, what if we were 0:03:19.480 --> 0:03:23.040 to take the same basic premise, but instead of having 0:03:23.160 --> 0:03:29.120 two human interviewees, replace one of those humans with a machine. Now, 0:03:29.480 --> 0:03:33.200 if that machine can convince the interrogator that the machine 0:03:33.240 --> 0:03:36.640 itself is a human being, it would be a pretty 0:03:36.640 --> 0:03:40.360 phenomenal achievement. Or you know, just generally, if the interrogator 0:03:40.560 --> 0:03:45.560 wasn't sure which of the two interrogate ease, right, which 0:03:45.600 --> 0:03:47.360 one is the human, which one is the machine? Or 0:03:47.440 --> 0:03:50.600 even if there could be a case where you don't know. 0:03:50.680 --> 0:03:54.120 It may be that you have two humans that you're interrogating, 0:03:54.200 --> 0:03:56.040 and it may be one of those things where you 0:03:56.120 --> 0:03:57.720 have to you know, if you don't know for a 0:03:57.800 --> 0:03:59.520 fact that one of them the machine, that makes it 0:03:59.560 --> 0:04:02.520 even hard, right, at least assuming that the computer program 0:04:02.560 --> 0:04:06.800 is sophisticated enough. Now, Touring was saying that we don't 0:04:06.800 --> 0:04:08.640 have any machines right now that can do this, but 0:04:08.680 --> 0:04:12.600 I envision a time when computers will be able to 0:04:12.680 --> 0:04:14.880 do such a thing where if you were to interrogate 0:04:14.960 --> 0:04:19.040 a computer, you would get back responses that would be uh, 0:04:19.320 --> 0:04:21.960 convincing enough for it to make it difficult to determine 0:04:21.960 --> 0:04:24.719 if it were man or machine. And so he said 0:04:25.120 --> 0:04:27.640 that he predicted in fifty years time, which would be 0:04:27.680 --> 0:04:31.200 the year two thousand, that computers and software be sophisticate 0:04:31.320 --> 0:04:34.120 enough that interrogators would only be able to guess correctly 0:04:34.520 --> 0:04:37.400 seventy of the time, meaning they would be fooled by 0:04:37.400 --> 0:04:41.240 the computer of the time. Right, So, uh, this was 0:04:41.279 --> 0:04:43.120 just kind of a thing he was coming up with, 0:04:43.160 --> 0:04:47.240 like an idea, a prediction not necessarily a test, although 0:04:47.800 --> 0:04:51.800 very much like Moore's observation became Moore's law, this became 0:04:51.839 --> 0:04:54.600 what is known as the touring test. People talk about 0:04:54.640 --> 0:04:59.240 a touring test as a machine capable of fooling people 0:04:59.320 --> 0:05:02.479 into thinking it's another person at least thirty of the 0:05:02.520 --> 0:05:05.960 time after five minutes of conversation. Very good point. Yes, 0:05:06.040 --> 0:05:07.960 it needs to be five minutes of conversation. If you 0:05:08.000 --> 0:05:10.960 are just getting maybe two or three responses, that might 0:05:11.000 --> 0:05:12.480 not be enough for you to be able to draw 0:05:12.600 --> 0:05:16.440 a conclusion that you feel good about. If after five 0:05:16.520 --> 0:05:21.440 minutes you still are not entirely, entirely certain, then that 0:05:21.560 --> 0:05:24.960 might say that this machine, in fact has passed the 0:05:25.000 --> 0:05:29.440 touring test. And this has been extrapolated to mean something 0:05:29.480 --> 0:05:35.240 about machine intelligence because Turing himself tied the idea of 0:05:35.240 --> 0:05:39.600 of how we perceive a machines intelligence directly to artificial 0:05:39.640 --> 0:05:43.760 intelligence and and and even how we perceive human intelligence. 0:05:43.880 --> 0:05:47.520 Because here's Touring's idea is a little cheeky, and I 0:05:47.600 --> 0:05:52.000 love the fact that's so cheeky. So Turing kind of said, 0:05:53.200 --> 0:05:55.159 would you say such a machine as intelligent if it 0:05:55.200 --> 0:05:57.960 appears to be intelligent, is it fair to say that 0:05:58.200 --> 0:06:02.440 is intelligent touring. So why not because I'm only able 0:06:02.480 --> 0:06:05.960 to tell that I am intelligent, right, because of my experience. 0:06:05.960 --> 0:06:08.520 I'm only able to have my own personal experience. I 0:06:08.520 --> 0:06:12.000 can't experience what someone else's life is like. All right, 0:06:12.000 --> 0:06:14.400 Sitting across from each other, Jonathan and I can only 0:06:14.520 --> 0:06:17.360 assume that the other one is intelligent, right, And it's 0:06:17.440 --> 0:06:20.960 because the other person is displaying traits that we associate 0:06:21.040 --> 0:06:24.599 with intelligence. They they seem to be able to take 0:06:24.640 --> 0:06:28.279 an information, respond to it, make decisions. And based on 0:06:28.320 --> 0:06:30.919 the fact that we ourselves also do that thing, we 0:06:31.040 --> 0:06:33.159 go ahead and say, all right, well, they clearly have 0:06:33.800 --> 0:06:38.960 the same features that I have, which includes intelligence. Uh. Now, 0:06:39.279 --> 0:06:42.440 he says, why would we not extend that same courtesy 0:06:42.520 --> 0:06:46.320 to a machine if it also appeared to display those 0:06:46.360 --> 0:06:49.839 same uh features? He says, doesn't matter if the computer 0:06:49.920 --> 0:06:53.440 is quote unquote thinking. If it's it can fool you. Yeah, 0:06:53.480 --> 0:06:55.720 if it's able to to simulate it well enough, you 0:06:55.800 --> 0:06:58.119 might as well say it's intelligent, because stimulate is probably 0:06:58.120 --> 0:07:01.120 a kinder phrase for that than fool. Yeah. Yeah, well 0:07:01.279 --> 0:07:03.680 I know it's it is fooling essentially, I mean, because 0:07:04.040 --> 0:07:08.200 ultimately you're talking about a computer programmer, who's making this happen? 0:07:08.640 --> 0:07:11.440 So nowadays we think of this as the Turing test. 0:07:11.560 --> 0:07:15.880 Can a machine of the time are more fool someone 0:07:15.920 --> 0:07:19.600 after five minutes of conversation into thinking it's a human? Now, 0:07:21.440 --> 0:07:24.080 this is really hard to do. This is yeah, this 0:07:24.120 --> 0:07:28.000 is non trivial. I mean it sounds almost simple. Yeah, 0:07:28.120 --> 0:07:33.119 the concept is simple, the execution incredibly difficult. Because here's 0:07:33.120 --> 0:07:39.920 the thing, human language varied. We have unstructured, unpredictable ways 0:07:39.960 --> 0:07:42.680 to say things like if if I were to tell 0:07:42.720 --> 0:07:47.640 you that it's a hundred degrees outside and it is 0:07:49.080 --> 0:07:51.800 humid like the humidities at and you go out there, 0:07:52.160 --> 0:07:55.360 then I'm sure all of our listeners would have slightly 0:07:55.400 --> 0:08:00.480 different ways to express their thoughts on the conditions out side. 0:08:00.880 --> 0:08:05.480 Some of them would probably contain colorful metaphors, probably mine 0:08:05.920 --> 0:08:09.640 very likely when it contain colorful metaphors, particularly if I 0:08:09.640 --> 0:08:11.560 had to be outside for any length of time. But 0:08:11.640 --> 0:08:14.080 that's the point. We would all have different ways of 0:08:14.120 --> 0:08:18.440 saying this. So how do you make a computer program 0:08:18.480 --> 0:08:22.760 able to interpret all the myriad of ways we can 0:08:22.840 --> 0:08:27.440 all express the same thought, let alone any thought. Right, 0:08:27.480 --> 0:08:30.600 this is what's referred to as natural language. Recognition, and 0:08:30.640 --> 0:08:34.120 it's a really huge problem in artificial intelligence and and 0:08:34.120 --> 0:08:38.520 and a lot of other speech related computer programming. Exactly. Yeah, 0:08:38.600 --> 0:08:41.319 this is where a computer program has to be able 0:08:41.360 --> 0:08:46.040 to parse the language so it recognizes things like this 0:08:46.280 --> 0:08:50.400 word is a noun, this word is a verb, this 0:08:50.400 --> 0:08:54.600 this word alters this other word. Uh. And not only 0:08:54.600 --> 0:08:56.240 does it need to be able to recognize it, it it 0:08:56.360 --> 0:08:59.440 needs to be able to respond in kind. Right, So 0:08:59.559 --> 0:09:02.240 if you appropriately in some way or another share, Yeah, 0:09:02.280 --> 0:09:05.679 you could have a computer program that's literally making up 0:09:05.720 --> 0:09:09.960 sentences or you know, the approximation of a sentence randomly, 0:09:10.120 --> 0:09:13.600 where it's just pulling strings of words and placing them 0:09:13.679 --> 0:09:16.200 in a sequence and then presenting them. But that wouldn't 0:09:16.200 --> 0:09:18.720 be convincing at all. If I were to say, Hello, 0:09:18.760 --> 0:09:21.480 how are you today, and Lauren was to say blue 0:09:21.480 --> 0:09:26.560 panther pickup jump down street, I'd be like what? And 0:09:26.600 --> 0:09:29.120 even that was closer to being a sentence than some 0:09:29.200 --> 0:09:30.760 of the random stuff that you would see if it 0:09:30.840 --> 0:09:35.800 was just truly completely So it also has to be 0:09:35.920 --> 0:09:39.880 able to uh endure a five minute long conversation, like 0:09:39.920 --> 0:09:41.760 we said, in order to pass the Turing test. So 0:09:42.040 --> 0:09:44.520 you can't have too much repetition or that gives it 0:09:44.520 --> 0:09:46.719 away absolutely. If you've ever been playing a video game 0:09:46.760 --> 0:09:48.720 and all of the NPCs say the same thing over 0:09:48.760 --> 0:09:51.200 and over and over again. Yeah, if if all the 0:09:51.320 --> 0:09:55.320 chat bot says is hey listen, then you're clearly playing 0:09:55.400 --> 0:10:00.200 Zelda and you're not actually having a decent conversation. That 0:10:00.280 --> 0:10:02.760 being said, there is someone I know who plays a 0:10:02.840 --> 0:10:05.440 ferry at the Georgia Renaissance Festival, and hey listen is 0:10:05.840 --> 0:10:10.040 heavily represented in a repertoire. Pretty amazing, No, it's pretty awesome, 0:10:10.320 --> 0:10:12.480 but at any rate, Yeah, So, so these are these 0:10:12.520 --> 0:10:16.000 are big problems. You have to build a database of words, 0:10:16.280 --> 0:10:19.000 you have to be able to figure out what kind 0:10:19.040 --> 0:10:22.280 of syntax are you going for. It's a wide open, 0:10:22.360 --> 0:10:27.040 huge problem, and solving this problem can be really beneficial 0:10:27.040 --> 0:10:28.679 in lots of ways. We'll talk a little bit about 0:10:28.679 --> 0:10:31.960 that later. It's it's beyond just making a program that 0:10:32.040 --> 0:10:35.960 seems human, right, That's that's one way of looking at it. 0:10:36.000 --> 0:10:37.440 But there are a lot of other benefits that come 0:10:37.440 --> 0:10:39.200 along with it, which will chat about towards the end 0:10:39.240 --> 0:10:41.439 of the show. But for right now, let's go into 0:10:41.600 --> 0:10:45.520 this story behind Eugene Goostman and whether or not it 0:10:45.600 --> 0:10:49.360 was actually the first chat bought to pass the Turing test. Yeah. So, 0:10:49.520 --> 0:10:52.600 first of all, we've got three programmers in this story 0:10:52.600 --> 0:10:58.040 of Vladimir Vassilov, Eugene Demchinko and Sarage you listen. As 0:10:58.080 --> 0:11:01.079 you may guess from their names, they all hail from 0:11:02.000 --> 0:11:05.880 Russia and the Ukraine as well, and although not all 0:11:05.920 --> 0:11:08.440 the not all of them still live there. But starting 0:11:08.480 --> 0:11:11.760 in two thousand one, they got to work on this, right, Yeah, 0:11:11.800 --> 0:11:14.320 they were trying to design a computer program that would 0:11:14.360 --> 0:11:18.160 pose specifically as a thirteen year old boy from Odessa 0:11:18.240 --> 0:11:21.720 in the Ukraine. Yeah, and that meant that they had 0:11:21.960 --> 0:11:27.319 specific parameters that they could work within. It automatically helped 0:11:27.520 --> 0:11:31.200 reduce some of that unpredictability and that lack of restriction 0:11:31.280 --> 0:11:33.479 that you would have if you were to just say, 0:11:33.520 --> 0:11:38.000 this is a fluent adult speaker of a given language. Yeah, yeah, 0:11:38.160 --> 0:11:41.800 giving him these you know, setting up the expectation from 0:11:41.840 --> 0:11:43.720 the judges that you know, this is a non native 0:11:43.720 --> 0:11:47.560 English speaker. It's it's a kid essentially, Um, you know 0:11:47.600 --> 0:11:50.840 that they'll expect him to have limited knowledge of the 0:11:50.880 --> 0:11:55.880 world and and different subject areas and a limited understanding 0:11:55.920 --> 0:11:59.240 of English. Vocabulary and grammar and all that kind of stuff. 0:11:59.320 --> 0:12:02.600 So you're already managing expectations. That's going to come into 0:12:02.600 --> 0:12:06.000 play when we talk about some of the criticisms about this, 0:12:06.080 --> 0:12:08.080 although although I do think it's a very clever way 0:12:08.120 --> 0:12:10.560 and in a lot of previous chat bots have have 0:12:10.640 --> 0:12:13.679 had similar, yes, similar kind of approaches. Yeah, because like 0:12:13.679 --> 0:12:16.160 we said, if you're to take a quote unquote pure 0:12:16.280 --> 0:12:21.080 approach to this, it's really really challenging. So yeah, by 0:12:21.200 --> 0:12:25.000 by limiting this, the judges have an idea of well, 0:12:25.040 --> 0:12:26.920 this this could be a thirteen year old boy or 0:12:26.920 --> 0:12:29.200 it could be a computer program. It means that the 0:12:29.200 --> 0:12:32.280 computer program doesn't have to be as sophisticated as one 0:12:32.320 --> 0:12:34.760 that would be completely fluent and have you know, an 0:12:34.760 --> 0:12:39.880 adult's experiences, uh and ability to communicate. So that was 0:12:39.880 --> 0:12:43.760 step one. And uh. Eugene Goosman took part in the 0:12:43.800 --> 0:12:48.760 competition UH that had five total chatbots. It was it 0:12:48.800 --> 0:12:52.120 was one of five that took place on the sixtieth 0:12:52.200 --> 0:12:56.880 anniversary of Touring's death, and the program managed to full 0:12:57.000 --> 0:12:59.679 thirty three percent of the judges into thinking it was 0:12:59.679 --> 0:13:03.079 actually a person. UH. And it had there were thirty judges. 0:13:03.120 --> 0:13:07.120 From what I understand. So that was where you got 0:13:07.160 --> 0:13:10.320 all the headlines of chat bought. There are computer beats 0:13:10.360 --> 0:13:14.840 touring tests, which already is not accurate. There were some 0:13:15.040 --> 0:13:18.800 slight um hiccups in a little bit of the news 0:13:18.840 --> 0:13:22.319 reporting that We'll get into that part of the story later, 0:13:22.880 --> 0:13:25.360 but you know, the key takeaways I think here are 0:13:25.440 --> 0:13:28.000 that this was a this was a competition that was 0:13:28.040 --> 0:13:33.959 celebrating touring awesome um and that there were thirty judges, 0:13:34.400 --> 0:13:38.160 some of whom were celebrities, yes, including an actor who 0:13:38.280 --> 0:13:41.360 had appeared on Red Dwarf right. Yes. And then a 0:13:41.360 --> 0:13:44.360 couple of years before that, because because you said they 0:13:44.360 --> 0:13:46.559 started working on this on two th two thousand one, 0:13:46.640 --> 0:13:49.719 this was not the first time that Gustman had entered competition. 0:13:50.400 --> 0:13:55.240 Two years previously, that same software had convinced twenty of 0:13:55.320 --> 0:13:58.840 judges at a similar competition that was held at Bletchley Park. Now, 0:13:58.920 --> 0:14:02.640 but Bletchley Park's where touring helped crack the Enigma machine, 0:14:03.440 --> 0:14:07.080 the the encoding device that the German military was using 0:14:07.160 --> 0:14:10.600 during World War Two. So this was a big celebration. 0:14:10.600 --> 0:14:14.120 It was at the centennial celebrating his birth, and it 0:14:14.440 --> 0:14:20.440 ended up falling just short of passing the Turing Test. Uh. 0:14:20.520 --> 0:14:24.520 The organizer of the the more recent event, the one 0:14:24.560 --> 0:14:28.000 in which Goostman ran away with quote unquote beating the 0:14:28.000 --> 0:14:31.960 Turing test. Uh. That organizer was Kevin Warwick. That name 0:14:32.000 --> 0:14:34.080 may sound familiar to some of our listeners if you've 0:14:34.120 --> 0:14:37.480 ever heard us talk about cyborgs. He's the guy who 0:14:37.680 --> 0:14:41.560 had an r f I D chip surgically implanted into him. 0:14:42.000 --> 0:14:45.000 His wife also did at one point. Yes, they could 0:14:45.000 --> 0:14:49.200 communicate with each other through them. Some unflattering news media 0:14:49.360 --> 0:14:53.040 sometimes refers to him as Captain Cyborg. Yeah. There there 0:14:53.080 --> 0:14:56.280 are some critics who say that he uh, he courts 0:14:56.440 --> 0:15:02.000 publicity in a manner that is um unbeco umming of 0:15:02.000 --> 0:15:05.360 of of a scientist, yes, really of anybody. Yes, yes, 0:15:05.960 --> 0:15:08.800 uh those are the critics who say that. By the way, 0:15:08.960 --> 0:15:11.560 I just want to make that clear. At any rate, 0:15:11.720 --> 0:15:13.680 He said that this was the first time a chat 0:15:13.680 --> 0:15:16.320 bot had passed the Turing test at an event where 0:15:16.360 --> 0:15:19.280 the conversation was open ended, meaning that they had not 0:15:19.560 --> 0:15:25.240 previously uh decided upon a specific topic or line of questioning. 0:15:25.640 --> 0:15:28.400 That the judges were allowed to say whatever they wanted 0:15:28.440 --> 0:15:30.960 to the chat bot and response, which obviously makes it 0:15:31.000 --> 0:15:33.640 harder because you have to have a much wider breadth 0:15:33.840 --> 0:15:37.880 of potential responses. Yeah. Yeah, because again, if you were 0:15:37.920 --> 0:15:40.160 to say, all right, this chat bot is just going 0:15:40.200 --> 0:15:43.920 to talk about, um, I don't know, sporting events from 0:15:44.000 --> 0:15:48.560 last year, Well, then you can prepare pretty well for that. Yeah, exactly. 0:15:48.600 --> 0:15:50.160 So it's one of those again, it's one of those 0:15:50.160 --> 0:15:54.720 things where the unrestricted nature adds in a degree of difficulty. 0:15:54.800 --> 0:15:57.440 So why would you need to make the qualification that 0:15:57.520 --> 0:15:59.320 this is an open ended approach and this is the 0:15:59.320 --> 0:16:03.840 first chat to manage it. That would be because despite 0:16:03.880 --> 0:16:06.960 what you may have heard, Eugene Goosband was not the 0:16:07.000 --> 0:16:09.960 first program to beat the Turing test, not by a 0:16:10.000 --> 0:16:14.720 long shot. So work on these sort of chatbots, these 0:16:14.800 --> 0:16:19.520 kind of artificial conversationalists. That's real recent, right, I mean 0:16:19.600 --> 0:16:21.880 they just started doing that like what like like maybe 0:16:21.880 --> 0:16:23.840 three or four years ago or two thousand, one of 0:16:23.880 --> 0:16:26.560 the earliest, I mean that's when they started with Goosman. Yeah. No, 0:16:26.640 --> 0:16:31.320 in the nineteen sixties and seventies, say, back in the 0:16:31.440 --> 0:16:35.160 mid nineteen sixties there was Eliza, which was written by 0:16:35.360 --> 0:16:38.800 Joseph Weisenbaum. If you're pronouncing it in the correct German, 0:16:39.000 --> 0:16:42.160 that's correct. Excellent, I'm finally learning. I'm sure, I'm sure 0:16:42.240 --> 0:16:45.920 he pronounces it Wisenbaum, but yeah, but no, it would 0:16:45.920 --> 0:16:49.840 be Weisenbaum at any rate. Um. This this was a 0:16:49.880 --> 0:16:53.400 program UM that would respond to human conversation and what 0:16:53.880 --> 0:16:57.720 ideally would be a relevant way. Yes, it was obviously 0:16:57.760 --> 0:17:01.040 an early attempt. It was not meant to be a 0:17:01.920 --> 0:17:04.320 program that takes on the Turing test. It was really 0:17:04.359 --> 0:17:06.439 a thought again, kind of like a thought experiment, the 0:17:06.440 --> 0:17:09.320 idea of what does it take to create a piece 0:17:09.320 --> 0:17:12.680 of software that can react to questions and make it 0:17:12.840 --> 0:17:15.080 make sense. At that point, it was more a can 0:17:15.160 --> 0:17:17.480 we do this? Then let's do this for real? Yeah, 0:17:17.520 --> 0:17:20.879 and these are the sort of foundations that you have 0:17:20.920 --> 0:17:23.280 to lay in order for other things like the Eugene 0:17:23.320 --> 0:17:28.840 Goostman program to be uh successful. So he created a 0:17:28.920 --> 0:17:33.000 language analyzer. Now this specifically would look at words that 0:17:33.280 --> 0:17:36.000 users would put in and then compare them against a 0:17:36.080 --> 0:17:38.720 database of words that were stored in the computer's memory. 0:17:39.240 --> 0:17:42.840 And then also created scripts. Now, in this case, the 0:17:42.880 --> 0:17:45.119 scripts were sets of rules. They're kind of like, you know, 0:17:45.160 --> 0:17:48.200 like a protocol or an algorithm in a way. These 0:17:48.280 --> 0:17:52.640 rules dictated how Eliza would respond to messages in order 0:17:52.680 --> 0:17:56.160 to cut down on that huge, massive number of variables 0:17:56.160 --> 0:17:59.719 we were talking about, the whole unrestricted, unpredictable thing, and 0:17:59.800 --> 0:18:03.679 so they would have different uh kind of like like overlays. 0:18:03.720 --> 0:18:05.520 Think of him as an overlay that would kind of 0:18:05.840 --> 0:18:09.359 guide Aliza's responses. And the most well known one was 0:18:09.400 --> 0:18:13.040 called doctor, which put Eliza in the role of a 0:18:13.200 --> 0:18:18.480 Rogerian psychiatrist. Uh. This is the person who responds to 0:18:18.480 --> 0:18:20.600 everything with a question. Al Right. It's that it's that 0:18:20.800 --> 0:18:25.080 passive interview style where where you know, you repeat back. 0:18:25.640 --> 0:18:27.560 You know, if if if I go, oh, man, like 0:18:27.600 --> 0:18:30.440 I'm I'm I'm really sad about my cat, Oh, tell 0:18:30.920 --> 0:18:32.640 what is it about your cat that makes you sad? 0:18:33.320 --> 0:18:35.159 And then you can say things like you know, And 0:18:35.359 --> 0:18:38.919 that's one of those things where as a conversation starts 0:18:38.960 --> 0:18:42.560 to wind down, you then have another line of questions, 0:18:42.600 --> 0:18:45.280 So tell me more about your mother, like that the 0:18:45.560 --> 0:18:47.960 whole tell me about your mother thing. That's really coming 0:18:47.960 --> 0:18:50.160 back to this kind of model of psychiatrists. But yeah, 0:18:50.200 --> 0:18:52.600 if you've ever heard the joke about them responding to 0:18:52.640 --> 0:18:56.040 anything with another question, just just taking the last word 0:18:56.080 --> 0:18:58.600 and turning it into a question, what am I paying 0:18:58.600 --> 0:19:01.800 you for Why do you think your page for that 0:19:01.880 --> 0:19:05.840 kind of thing? Uh So that's what That's how Eliza responded, 0:19:05.880 --> 0:19:10.040 And in fact, you can find examples of the Eliza 0:19:10.400 --> 0:19:14.159 program script transcripts and even even the actual there's there. 0:19:14.160 --> 0:19:17.399 There are ports, you could say, or people have essentially 0:19:17.440 --> 0:19:20.280 created their own version of Eliza just using the original 0:19:20.320 --> 0:19:22.800 Liza as a guide. You can find tons of them 0:19:22.800 --> 0:19:25.000 on the web, and you can attempt to have a 0:19:25.040 --> 0:19:29.200 conversation with them. It's not terribly compelling, but it's it's 0:19:29.280 --> 0:19:33.240 kind of fun. Usually within maybe four or five exchanges, 0:19:33.240 --> 0:19:35.120 you've already run into something where you're like, well, this 0:19:35.160 --> 0:19:38.120 can't be a person, or if it's a person, it's 0:19:38.119 --> 0:19:41.040 the weirdest person I've ever conversed with. But but again, 0:19:41.080 --> 0:19:44.600 it wasn't really an earnest attempt to to create something 0:19:44.640 --> 0:19:48.159 that would pass the quote unquote tearing tests, which I 0:19:48.200 --> 0:19:49.800 don't think was being referred to as such at that 0:19:49.840 --> 0:19:52.240 point yet. Yeah, it was kind of people knew about 0:19:52.280 --> 0:19:55.520 tourings prediction, but it wasn't so much called a turing test. 0:19:56.000 --> 0:19:59.080 Another chat bought that premiered in the early nineties seventies 0:19:59.119 --> 0:20:02.880 went even further and actually was an attempt to try 0:20:03.000 --> 0:20:06.000 and pass the Turing test in a very specific approach, 0:20:06.080 --> 0:20:08.560 kind of like you know, Eugene Goostman, is a very 0:20:08.560 --> 0:20:12.200 specific approach to narrow down that those parameters. In this case, 0:20:12.280 --> 0:20:15.639 this one was called Perry p A R R Y 0:20:15.920 --> 0:20:19.520 A Right, and Kenneth Colby created it too to emulate 0:20:19.800 --> 0:20:24.080 a patient who has paranoid schizophrenia. Yeah, someone who has 0:20:24.200 --> 0:20:28.360 you know, sort of a persecution complex. Uh. They imagined 0:20:28.440 --> 0:20:32.639 that's that there are people or other entities that are 0:20:32.680 --> 0:20:36.000 out to get them. And in this uh specific case, 0:20:36.320 --> 0:20:40.080 he he kinda he kind of really embraced this approach. 0:20:40.600 --> 0:20:43.840 It reminds me of people who create a like a 0:20:43.960 --> 0:20:46.080 Dungeons and Dragon's character, but then give their character an 0:20:46.200 --> 0:20:49.960 entire backstory. Yeah. Yeah, yeah, this this Perry persona was 0:20:50.040 --> 0:20:52.040 an entire persona. It was a twenty eight year old 0:20:52.080 --> 0:20:54.400 man with a job as a post office clerk who 0:20:54.440 --> 0:20:57.240 was single, had no brothers and sisters, and rarely saw 0:20:57.359 --> 0:21:00.119 his parents. He had specific hobbies. He liked to of 0:21:00.160 --> 0:21:03.240 the movies and horse racing. Yeah, he liked to bet 0:21:03.240 --> 0:21:06.720 on the horses and placed a bet with bookies in 0:21:06.720 --> 0:21:11.240 the past. Right, and that he realized later bookies have 0:21:11.359 --> 0:21:15.639 an association with the criminal underworld, and that therefore the 0:21:15.800 --> 0:21:19.240 mafia knew about him and we're out to get him. Now. Now, 0:21:19.280 --> 0:21:21.479 all of that might sound very ridiculous to you if 0:21:21.520 --> 0:21:24.400 you never had any kind of interaction with someone who 0:21:24.440 --> 0:21:27.560 suffers from paranoid schizophrenia, that can seem like, well, that 0:21:27.640 --> 0:21:31.600 seems cartoonish. But no, this, this kind of thinking is 0:21:31.680 --> 0:21:34.719 not uncommon, you know, whether it's whether it's a criminal 0:21:34.800 --> 0:21:40.760 organization or the government or some other even unnamed entity. Oh. Absolutely, Uh, 0:21:40.880 --> 0:21:44.000 it's it's very much realistic in terms of that kind 0:21:44.000 --> 0:21:46.560 of diagnosis. And and if I was speaking about it 0:21:46.600 --> 0:21:48.320 with humor in my voice a moment ago, it's it's 0:21:48.320 --> 0:21:51.800 only because I am absolutely tickled that the programmers of 0:21:52.000 --> 0:21:56.200 this of this program built it, built it in. Yeah, 0:21:56.880 --> 0:21:59.800 it's it's actually pretty entertaining that they went so far 0:21:59.840 --> 0:22:03.199 as to make this whole uh, this whole backstory to explain, 0:22:03.240 --> 0:22:06.280 because that's what gives it the believability, right, and then 0:22:06.440 --> 0:22:09.359 they ended up testing it in conversation with a human. 0:22:09.359 --> 0:22:13.080 Perry would gradually start to introduce his thoughts quote unquote 0:22:13.200 --> 0:22:17.560 thoughts about being persecuted, and would respond sensitively to anything 0:22:17.600 --> 0:22:21.480 said about his appearance, family, or religious beliefs. I've actually 0:22:21.520 --> 0:22:24.760 seen lots of transcripts of conversations with Perry, and sure enough, 0:22:25.000 --> 0:22:27.360 it's one of those things where you know, you might 0:22:27.720 --> 0:22:30.080 have a few exchanges and then Perry ends up saying 0:22:30.240 --> 0:22:34.000 something that seems really odd, but not so odd as 0:22:34.040 --> 0:22:37.120 to seem artificial. It just seems like it's a non sequitur, 0:22:37.720 --> 0:22:39.760 you know, something like, well, that's what they want you 0:22:39.800 --> 0:22:41.240 to think. And if you were to say, who are 0:22:41.280 --> 0:22:44.280 they the mafia? Who did you think I was talking about? 0:22:45.000 --> 0:22:48.199 Like the mafia after you? Of course they're after me. 0:22:48.720 --> 0:22:50.320 They know who I am. That kind of stuff, and 0:22:50.560 --> 0:22:53.439 it's disturbing, like it's you know, when you know what 0:22:53.520 --> 0:22:55.680 it is, it's kind of amusing, but like if you're 0:22:55.680 --> 0:23:00.320 in the middle of a conversation, oh, this poor person. Yeah. 0:23:00.680 --> 0:23:03.920