WEBVTT - TechStuff Classic: Passing the Turing Test 0:00:04.400 --> 0:00:07.800 Welcome to tech Stuff, a production from I Heart Radio. 0:00:12.000 --> 0:00:14.640 Hey there, and welcome to tech Stuff. I'm your host, 0:00:14.760 --> 0:00:17.479 Jonathan Strickland. I'm an executive producer with I Heart Radio 0:00:17.520 --> 0:00:21.560 and love all things tech, and this is time for 0:00:21.680 --> 0:00:25.200 a classic tech Stuff episode. This episode originally published on 0:00:25.320 --> 0:00:29.960 June twenty three, two thousand fourteen. It is titled passing 0:00:30.320 --> 0:00:34.840 the Turing Test, something that we frequently associate with artificial 0:00:34.840 --> 0:00:37.839 intelligence and actual you know, like machines thinking or at 0:00:37.880 --> 0:00:40.879 least appearing to think. So let's listen in to this 0:00:41.040 --> 0:00:45.600 classic episode. The Touring Test is named after Alan Turing, 0:00:46.240 --> 0:00:49.040 and we've done a full episode on Alan Turing way 0:00:49.080 --> 0:00:56.120 back when, back in November. Yeah, phenomenal person, amazing thinker. Yeah, 0:00:56.160 --> 0:01:00.320 one of the like the grandfather of computer science. Also 0:01:00.720 --> 0:01:04.319 tragic life story which we went into detail back in 0:01:04.400 --> 0:01:06.320 that story. And so you might wonder why, right, So 0:01:06.360 --> 0:01:08.360 what does that have to do was the the guy 0:01:08.760 --> 0:01:11.280 who was the essentially the father of computer science or 0:01:11.319 --> 0:01:13.680 grandfather of computer science? What does he have to do 0:01:13.720 --> 0:01:16.759 with the story about a computer program in June two 0:01:16.760 --> 0:01:21.560 thousand fourteen, a computer program with an interesting name, Eugene 0:01:21.560 --> 0:01:25.640 Goostman passing the Turing test. What does that all have 0:01:25.680 --> 0:01:27.840 to do with each other? Well? To answer that, we 0:01:27.880 --> 0:01:31.399 have to ask what is a Turing test? Well, as 0:01:31.480 --> 0:01:35.240 it turns out, Away back in the nineteen fifties, he 0:01:35.360 --> 0:01:40.440 started envisioning uh I thought experiment. Yeah, he published a 0:01:40.520 --> 0:01:44.960 paper in a journal called Mind in nineteen fifty called Computing, 0:01:45.080 --> 0:01:48.160 Machinery and Intelligence, and he sort of laid out his 0:01:48.560 --> 0:01:51.200 thought experiment there. It's interesting because what he did was 0:01:51.240 --> 0:01:53.800 he took this idea of a party game and then 0:01:53.840 --> 0:01:58.160 adapted it for computers. Right. The party game that it's 0:01:58.240 --> 0:02:02.080 based on would have three participants, an interrogator, a man, 0:02:02.240 --> 0:02:04.960 and a woman um, all situated so that they can't 0:02:05.000 --> 0:02:07.720 see one another, and the interrogator is supposed to ask 0:02:07.840 --> 0:02:10.960 questions um to try to figure out which of the 0:02:11.000 --> 0:02:14.239 participants is the dude and which is the lady exactly. 0:02:14.600 --> 0:02:17.440 And it's the dude's job to try and mislead the 0:02:17.480 --> 0:02:21.200 interrogator to believe that that he, in fact is the 0:02:21.280 --> 0:02:23.280 lady and the other one is the man. It's the 0:02:23.360 --> 0:02:26.240 lady's job to say, hey, I want you to get 0:02:26.280 --> 0:02:29.000 this right. I'm the lady. That other guy that's the dude. 0:02:29.480 --> 0:02:32.840 And so the interrogator has to ask questions. Now, obviously 0:02:33.440 --> 0:02:35.240 they can't see each other. Because they could see each other, 0:02:35.320 --> 0:02:38.840 then that would probably give things away. Probably, they really 0:02:38.880 --> 0:02:40.960 shouldn't be able to hear each other because that could 0:02:41.000 --> 0:02:45.040 also give things away. Sure, and if you use handwritten notes, 0:02:45.120 --> 0:02:48.520 then you the interrogator might be able to make judgments 0:02:48.520 --> 0:02:51.680 based on the handwriting style. So it should really be typewritten, right, 0:02:51.720 --> 0:02:54.600 So you want to you want to remove as many 0:02:54.639 --> 0:02:58.280 easily identifiable traits from this game as possible to make 0:02:58.320 --> 0:03:02.640 it all about the questions and the answers. Now, Touring said, 0:03:02.720 --> 0:03:05.119 what if we were to take the same basic premise, 0:03:05.680 --> 0:03:10.520 but instead of having two human interviewees, replace one of 0:03:10.520 --> 0:03:14.720 those humans with a machine. Now, if that machine can 0:03:14.800 --> 0:03:18.840 convince the interrogator that the machine itself is a human being, 0:03:19.400 --> 0:03:22.680 it would be a pretty phenomenal achievement. Or you know, 0:03:22.760 --> 0:03:25.880 just generally, if the interrogator wasn't sure which of the 0:03:25.960 --> 0:03:30.120 two interrogate, right, whether which one is the human, which 0:03:30.120 --> 0:03:32.720 one is the machine, or even if there could be 0:03:32.800 --> 0:03:35.160 a case where you don't know. It may be that 0:03:35.240 --> 0:03:38.600 you have two humans that you're interrogating, and it may 0:03:38.640 --> 0:03:40.520 be one of those things where you have to you know, 0:03:40.560 --> 0:03:42.240 if you don't know for a fact that one of 0:03:42.280 --> 0:03:44.680 them the machine, that makes it even harder, right, at 0:03:44.720 --> 0:03:48.320 least assuming that the computer program is sophisticated enough. Now, 0:03:48.600 --> 0:03:51.480 Touring was saying that we don't have any machines right 0:03:51.480 --> 0:03:53.640 now that can do this, but I envision a time 0:03:54.120 --> 0:03:57.200 when computers will be able to do such a thing 0:03:57.240 --> 0:03:59.960 where if you were to interrogate a computer, you would 0:04:00.040 --> 0:04:04.080 get back responses that would be uh, convincing enough for 0:04:04.120 --> 0:04:06.200 it to make it difficult to determine if it were 0:04:06.280 --> 0:04:09.680 man or machine. And so he said that he predicted 0:04:09.720 --> 0:04:12.360 in fifty years time, which would be the year two thousand, 0:04:13.080 --> 0:04:16.359 that computers and software be sophisticate enough that interrogators would 0:04:16.360 --> 0:04:19.559 only be able to guess correctly seventy percent of the time, 0:04:19.880 --> 0:04:22.679 meaning that they would be fooled by the computer pent 0:04:22.760 --> 0:04:25.440 of the time. Right. So, uh, this was just kind 0:04:25.440 --> 0:04:27.600 of a thing he was coming up with, like an idea, 0:04:28.520 --> 0:04:32.200 a prediction, not necessarily a test. Although very much like 0:04:32.520 --> 0:04:36.160 Moore's observation became Moore's law, this became what is known 0:04:36.200 --> 0:04:39.239 as the Turing test. People talk about a Touring test 0:04:39.720 --> 0:04:43.960 as a machine capable of fooling people into thinking it's 0:04:44.080 --> 0:04:47.200 another person at least thirty percent of the time after 0:04:47.400 --> 0:04:50.400 five minutes of conversation. Very good point. Yes, it needs 0:04:50.440 --> 0:04:52.159 to be five minutes of conversation. If you are just 0:04:52.160 --> 0:04:55.120 getting maybe two or three responses, that might not be 0:04:55.240 --> 0:04:57.160 enough for you to be able to draw a conclusion 0:04:57.160 --> 0:05:01.200 that you feel good about. If after five minutes you 0:05:01.760 --> 0:05:06.480 still are not entirely, entirely certain, then that might say 0:05:06.800 --> 0:05:09.520 that this machine, in fact has passed the Turing test. 0:05:09.880 --> 0:05:14.040 And this has been extrapolated to mean something about machine 0:05:14.080 --> 0:05:19.360 intelligence because Turing himself tied the idea of of how 0:05:19.400 --> 0:05:24.160 we perceive a machines intelligence directly to artificial intelligence and 0:05:24.400 --> 0:05:29.039 and and even how we perceive human intelligence. Because here's 0:05:29.040 --> 0:05:32.200 Touring's idea is a little cheeky, and I love the 0:05:32.240 --> 0:05:35.799 fact that it's so cheeky. So Turing kind of said, 0:05:37.000 --> 0:05:39.000 would you say such a machine as intelligent if it 0:05:39.040 --> 0:05:41.760 appears to be intelligent, is it fair to say that 0:05:42.000 --> 0:05:46.280 is intelligent? Touring? So why not? Because I'm only able 0:05:46.320 --> 0:05:49.800 to tell that I am intelligent, right because of my experience. 0:05:49.800 --> 0:05:52.320 I'm only able to have my own personal experience. I 0:05:52.360 --> 0:05:55.800 can't experience what someone else's life is like, all right, 0:05:55.839 --> 0:05:58.239 sitting across from each other, Jonathan and I can only 0:05:58.360 --> 0:06:01.200 assume that the other one is intelligent, right, And it's 0:06:01.240 --> 0:06:04.760 because the other person is displaying traits that we associate 0:06:04.839 --> 0:06:09.160 with intelligence. They seem to be able to take an information, 0:06:09.200 --> 0:06:12.520 respond to it, make decisions. And based on the fact 0:06:12.520 --> 0:06:15.240 that we ourselves also do that thing, we go ahead 0:06:15.240 --> 0:06:18.119 and say, all right, well, they clearly have the same 0:06:18.600 --> 0:06:23.559 features that I have, which includes intelligence. Uh. Now, he says, 0:06:23.560 --> 0:06:26.599 why would we not extend that same courtesy to a 0:06:26.640 --> 0:06:31.640 machine if it also appeared to display those same uh features? 0:06:31.640 --> 0:06:35.160 He says, doesn't matter if the computer is quote unquote thinking. 0:06:35.839 --> 0:06:37.840 If it's it can fool you. Yeah, if it's able 0:06:37.920 --> 0:06:40.000 to to simulate it well enough, you might as well 0:06:40.040 --> 0:06:42.400 say it's intelligent, because simulate it is probably a kinder 0:06:42.440 --> 0:06:45.400 phrase for that than fool. Yes. Yeah, well I know 0:06:45.480 --> 0:06:48.720 it's it is fooling essentially, I mean, because ultimately you're 0:06:48.760 --> 0:06:52.680 talking about a computer programmer who's making this happen. So 0:06:53.040 --> 0:06:55.640 nowadays we think of this as the Turing test. Can 0:06:55.680 --> 0:06:59.920 a machine of the time are more fool someone after 0:07:00.080 --> 0:07:03.440 five minutes of conversation into thinking it's a human? Now, 0:07:05.240 --> 0:07:08.719 this is really hard to do. This is is non trivial. 0:07:08.920 --> 0:07:13.040 I mean it sounds almost simple. Yeah, the concept is simple, 0:07:13.240 --> 0:07:20.040 the execution incredibly difficult. Because here's the thing, human language varied. 0:07:20.640 --> 0:07:25.440 We have unstructured, unpredictable ways to say things like if 0:07:25.480 --> 0:07:28.480 if I were to tell you that it's a hundred 0:07:28.520 --> 0:07:34.880 degrees outside and it is humid like the humidities, and 0:07:34.920 --> 0:07:37.480 you go out there, then I'm sure all of our 0:07:37.480 --> 0:07:41.880 listeners would have slightly different ways to express their thoughts 0:07:42.600 --> 0:07:46.640 on the conditions outside. Some of them would probably contain 0:07:46.720 --> 0:07:52.440 colorful metaphors. Mine very likely when it contain colorful metaphors, 0:07:52.480 --> 0:07:54.720 particularly if I had to be outside for any length 0:07:54.720 --> 0:07:57.200 of time. But that's the point. We would all have 0:07:57.240 --> 0:08:00.320 different ways of saying this. So how do you make 0:08:00.400 --> 0:08:05.800 a computer program able to interpret all the myriad of 0:08:05.880 --> 0:08:09.560 ways we can all express the same thought, let alone 0:08:10.200 --> 0:08:12.680 any thought? All right? This is what's referred to as 0:08:12.880 --> 0:08:15.960 natural language recognition, and it's a really huge problem in 0:08:16.080 --> 0:08:19.360 artificial intelligence and and and a lot of other speech 0:08:19.480 --> 0:08:23.880 related computer programming. Exactly. Yeah, this is where a computer 0:08:23.960 --> 0:08:27.280 program has to be able to parse the language so 0:08:27.440 --> 0:08:32.120 it recognizes things like this word is a noun, this 0:08:32.200 --> 0:08:37.720 word is a verb, this this word alters this other word. Uh. 0:08:37.800 --> 0:08:39.480 And not only does it need to be able to 0:08:39.480 --> 0:08:43.160 recognize it, it it needs to be able to respond in kind. Right, 0:08:43.200 --> 0:08:46.120 So if you have appropriately in some way or another share. Yeah, 0:08:46.120 --> 0:08:49.480 you could have a computer program that's literally making up 0:08:49.559 --> 0:08:53.800 sentences or you know, the approximation of a sentence randomly, 0:08:53.960 --> 0:08:57.439 where it's just pulling strings of words and placing them 0:08:57.520 --> 0:09:00.000 in a sequence and then presenting them. But that wouldn't 0:09:00.000 --> 0:09:02.520 be convincing at all. If I were to say, hello, 0:09:02.600 --> 0:09:05.320 how are you today, and Lauren was to say blue 0:09:05.320 --> 0:09:10.400 panther pickup jump down street, I'd be like what? And 0:09:10.440 --> 0:09:12.960 even that was closer to being a sentence than some 0:09:13.000 --> 0:09:14.600 of the random stuff that you would see if it 0:09:14.640 --> 0:09:19.719 was just truly completely So it also has to be 0:09:19.760 --> 0:09:23.720 able to uh endure a five minute long conversation, like 0:09:23.760 --> 0:09:25.600 we said, in order to pass the Turing test. So 0:09:25.880 --> 0:09:29.200 you can't have too much repetition or that gives it away. Absolutely. 0:09:29.240 --> 0:09:30.800 If you've ever been playing a video game and all 0:09:30.880 --> 0:09:32.839 of the NPCs say the same thing over and over 0:09:32.880 --> 0:09:35.640 and over again, yeah, if if all the chat bot 0:09:35.720 --> 0:09:40.320 says is hey, listen, then you're clearly playing Zelda and 0:09:40.400 --> 0:09:44.520 you're not actually having a decent conversation. That being said, 0:09:44.559 --> 0:09:47.080 there is someone I know who plays a ferry at 0:09:47.080 --> 0:09:50.559 the Georgia Renaissance Festival, and hey, listen is heavily represented 0:09:50.559 --> 0:09:53.880 in a repertoire. It's pretty amazing. No, it's pretty awesome. 0:09:54.120 --> 0:09:56.319 But at any rate, Yeah, so so these are these 0:09:56.320 --> 0:09:59.800 are big problems. You have to build a database of words, 0:10:00.120 --> 0:10:02.839 you have to be able to figure out what kind 0:10:02.840 --> 0:10:06.120 of syntax are you going for. It's a wide open, 0:10:06.200 --> 0:10:10.840 huge problem, and solving this problem can be really beneficial 0:10:10.880 --> 0:10:12.480 in lots of ways. We'll talk a little bit about 0:10:12.520 --> 0:10:15.760 that later. It's it's beyond just making a program that 0:10:15.880 --> 0:10:19.800 seems human, right, That's that's one way of looking at it. 0:10:19.840 --> 0:10:21.240 But there are a lot of other benefits that come 0:10:21.280 --> 0:10:23.040 along with it, which will chat about towards the end 0:10:23.040 --> 0:10:25.280 of the show. But for right now, let's go into 0:10:25.440 --> 0:10:29.360 this story behind Eugene Goostman and whether or not it 0:10:29.440 --> 0:10:33.200 was actually the first chat bought to pass the Turing test. Yeah. So, 0:10:33.320 --> 0:10:36.400 first of all, we've got three programmers in this story 0:10:36.440 --> 0:10:41.880 of Vladimir Vslov, Eugene Demchinko and Sarage you listen. As 0:10:41.920 --> 0:10:44.920 you may guess from their names, they all hail from 0:10:45.840 --> 0:10:49.720 Russia and the Ukraine as well, and although not all, 0:10:49.960 --> 0:10:52.400 not all of them still live there. But starting in 0:10:52.679 --> 0:10:55.600 two thousand one they got to work on this, right, Yeah, 0:10:55.640 --> 0:10:58.120 they were trying to design a computer program that would 0:10:58.160 --> 0:11:01.400 pose specifically as a third teen year old boy from 0:11:01.400 --> 0:11:05.280 Odessa in the Ukraine. Yeah, and that meant that they 0:11:05.320 --> 0:11:09.880 had specific parameters that they could work within. It automatically 0:11:10.760 --> 0:11:14.360 helped reduce some of that unpredictability and that lack of 0:11:14.440 --> 0:11:17.320 restriction that you would have if you were to just say, 0:11:17.360 --> 0:11:21.800 this is a fluent adult speaker of a given language. Yeah, Yeah, 0:11:22.000 --> 0:11:25.600 giving him these you know, setting up the expectation from 0:11:25.679 --> 0:11:27.520 the judges that you know, this is a non native 0:11:27.559 --> 0:11:31.400 English speaker. It's it's a kid essentially. Um, you know 0:11:31.400 --> 0:11:34.679 that they'll expect him to have limited knowledge of the 0:11:34.720 --> 0:11:39.720 world and and different subject areas, and a limited understanding 0:11:39.760 --> 0:11:43.040 of English vocabulary and grammar and all that kind of stuff. 0:11:43.120 --> 0:11:46.400 So you're already managing expectations. That's going to come into 0:11:46.400 --> 0:11:49.800 play when we talk about some of the criticisms about this. 0:11:49.920 --> 0:11:51.880 Although although I do think it's a very clever way 0:11:51.920 --> 0:11:54.640 and and a lot of previous chatbots have have had 0:11:54.679 --> 0:11:57.840 similar yes kind of approaches. Yeah, because like we said, 0:11:57.840 --> 0:12:00.560 if you were to take a quote unquote pure approach 0:12:00.600 --> 0:12:05.199 to this, it's really really challenging. So yeah, by by 0:12:05.320 --> 0:12:09.160 limiting this, the judges have an idea of well, this 0:12:09.160 --> 0:12:10.840 this could be a thirteen year old boy or it 0:12:10.840 --> 0:12:13.360 could be a computer program. It means that the computer 0:12:13.400 --> 0:12:16.200 program doesn't have to be as sophisticated as one that 0:12:16.200 --> 0:12:19.959 would be completely fluent and have you know, an adult's experiences, 0:12:20.880 --> 0:12:24.240 uh and ability to communicate. So that was step one 0:12:24.920 --> 0:12:29.000 and uh, Eugene Goosman took part in the competition. UH 0:12:29.040 --> 0:12:33.000 that had five total chatbots. It was it was one 0:12:33.040 --> 0:12:36.960 of five that took place on the sixtieth anniversary of 0:12:37.040 --> 0:12:41.400 Turing's death, and the program managed to full thirty three 0:12:41.559 --> 0:12:45.280 percent of the judges into thinking it was actually a person. Uh, 0:12:45.280 --> 0:12:47.800 and it had there were thirty judges from what I understand, 0:12:48.080 --> 0:12:51.960 So that was where you got all the headlines of 0:12:52.080 --> 0:12:55.959 chat bought. There are computer beats touring tests, which already 0:12:56.120 --> 0:13:01.160 is not accurate. There were some slight um hiccups in 0:13:01.679 --> 0:13:03.920 a little bit of the news reporting that we'll get 0:13:03.960 --> 0:13:07.120 into that part of the story later. But you know, 0:13:07.440 --> 0:13:10.000 the key takeaways I think here are that this was 0:13:10.040 --> 0:13:14.840 a this was a competition that was celebrating touring awesome um. 0:13:14.920 --> 0:13:19.840 And that there were thirty judges, some of whom were celebrities, yes, 0:13:20.160 --> 0:13:24.560 including an actor who had appeared on Red Dwarf right. Yes. 0:13:24.840 --> 0:13:27.679 And then a couple of years before that, because because 0:13:27.679 --> 0:13:29.839 you said they started working on this in two two 0:13:29.840 --> 0:13:32.359 thousand one, this was not the first time that Gustman 0:13:32.480 --> 0:13:37.479 had entered competition. UH. Two years previously, that same software 0:13:37.520 --> 0:13:40.760 had convinced twenty of judges at a similar competition that 0:13:40.840 --> 0:13:43.880 was held at Bletchley Park. Now, but Bletchley Park that's 0:13:43.880 --> 0:13:48.720 where Touring helped crack the Enigma machine, the encoding device 0:13:48.800 --> 0:13:52.000 that the German military was using during World War Two. 0:13:52.559 --> 0:13:54.880 So this was a big celebration. It was at the 0:13:55.120 --> 0:13:59.240 centennial celebrating his birth, and it ended up falling just 0:13:59.400 --> 0:14:05.960 short of passing the Turing test. Uh. The organizer of 0:14:06.000 --> 0:14:09.400 the the more recent event, the one in which Gustman 0:14:09.720 --> 0:14:13.360 ran away with quote unquote beating the Turing test. UH. 0:14:13.400 --> 0:14:16.680 That organizer was Kevin Warwick. That name may sound familiar 0:14:16.720 --> 0:14:18.400 to some of our listeners. If you've ever heard us 0:14:18.440 --> 0:14:22.000 talk about cyborgs. He's the guy who had an r 0:14:22.120 --> 0:14:26.240 f I D chip surgically implanted into him. His wife 0:14:26.240 --> 0:14:29.480 also did at one point. Yes, they could communicate with 0:14:29.520 --> 0:14:33.920 each other through them. Some unflattering news media sometimes refers 0:14:33.960 --> 0:14:37.400 to him as Captain Cyborg. Yeah. There there are some 0:14:37.600 --> 0:14:41.680 critics who say that he, uh, he courts publicity in 0:14:41.720 --> 0:14:47.240 a manner that is unbecoming of of of a scientist. Yes, 0:14:47.320 --> 0:14:51.520 really of anybody. Yes, yes, uh, those are the critics 0:14:51.560 --> 0:14:53.320 who say that. By the way, I just want to 0:14:53.320 --> 0:14:56.320 make that clear. At any rate, He said that this 0:14:56.360 --> 0:14:58.760 was the first time a chatbot had passed the Turing 0:14:58.880 --> 0:15:01.840 test at an event, or the conversation was open ended, 0:15:02.200 --> 0:15:06.080 meaning that they had not previously uh decided upon a 0:15:06.160 --> 0:15:10.440 specific topic or line of questioning that the judges were 0:15:10.440 --> 0:15:13.000 allowed to say whatever they wanted to the chat bot 0:15:13.000 --> 0:15:15.760 and response, which obviously makes it harder because you have 0:15:15.840 --> 0:15:20.760 to have a much wider breadth of potential responses. Yeah. Yeah, 0:15:20.800 --> 0:15:22.840 because again, if you were to say, all right, this 0:15:22.920 --> 0:15:26.400 chat bot is just going to talk about, um, I 0:15:26.400 --> 0:15:29.760 don't know, sporting events from last year, well, then you 0:15:30.040 --> 0:15:32.720 can prepare pretty well for that. Yeah, exactly. So it's 0:15:32.720 --> 0:15:34.480 one of those again, it's one of those things where 0:15:34.680 --> 0:15:38.840 the unrestricted nature adds in a degree of difficulty. It's 0:15:38.880 --> 0:15:41.080 time for us to take a quick break, but will 0:15:41.120 --> 0:15:51.040 be right back. So why would you need to make 0:15:51.040 --> 0:15:53.480 the qualification that this is an open ended approach and 0:15:53.520 --> 0:15:56.280 this is the first chat bot to manage it. That 0:15:56.320 --> 0:16:00.400 would be because despite what you may have heard, Eugene 0:16:00.480 --> 0:16:03.600 Goosband was not the first program to beat the Turing test, 0:16:04.040 --> 0:16:08.440 not by a long shot. So work on these sort 0:16:08.440 --> 0:16:13.760 of chatbots, these kind of artificial conversationalists. That's real recent, right, 0:16:13.800 --> 0:16:16.000 I mean they just started doing that like what like 0:16:16.000 --> 0:16:18.200 like maybe three or four years ago or two thousand, 0:16:18.200 --> 0:16:19.840 one of the earliest, I mean that's when they started 0:16:19.840 --> 0:16:22.600 with Goose. Yeah. No, in the nineteen sixties and seventies. 0:16:22.640 --> 0:16:28.080 Say what back in the mid nineteen sixties there was Eliza, 0:16:28.680 --> 0:16:32.400 which was written by Joseph Weisenbaum. If you're pronouncing it 0:16:32.440 --> 0:16:35.840 in the correct German, that's correct. Excellent, I'm finally learning. 0:16:35.920 --> 0:16:39.800 I'm sure. I'm sure he pronounces at Wisenbaum. But yeah, 0:16:39.880 --> 0:16:43.240 but no, it would be Weisenbaum at any rate. Um, 0:16:43.280 --> 0:16:46.680 this this was a program um that would respond to 0:16:46.760 --> 0:16:50.880 human conversation in what ideally would be a relevant way. Yes, 0:16:51.240 --> 0:16:54.240 it was obviously an early attempt. It was not meant 0:16:54.360 --> 0:16:58.400 to be a program that takes on the Turing test. 0:16:58.400 --> 0:17:00.200 It was really a thought again, kind of like a 0:17:00.200 --> 0:17:02.760 thought experiment, the idea of what does it take to 0:17:02.880 --> 0:17:06.080 create a piece of software that can react to questions 0:17:06.640 --> 0:17:08.960 and make it make sense. At that point, it was 0:17:09.000 --> 0:17:11.000 more a can we do this? Then let's do this 0:17:11.040 --> 0:17:14.880 for real? Yeah, and these are the sort of foundations 0:17:15.040 --> 0:17:16.879 that you have to lay in order for other things 0:17:16.920 --> 0:17:21.679 like the Eugene Goostman program to be uh successful. So 0:17:22.480 --> 0:17:26.720 he created a language analyzer. Now this specifically would look 0:17:26.760 --> 0:17:29.680 at words that users would put in and then compare 0:17:29.680 --> 0:17:32.359 them against a database of words that were stored in 0:17:32.359 --> 0:17:36.879 the computer's memory. And then also created scripts. Now, in 0:17:36.880 --> 0:17:39.159 this case, the scripts were sets of rules. They're kind 0:17:39.160 --> 0:17:41.280 of like, you know, like a protocol or an algorithm 0:17:41.280 --> 0:17:45.639 in a way. These rules dictated how Eliza would respond 0:17:45.680 --> 0:17:49.000 to messages in order to cut down on that huge, 0:17:49.320 --> 0:17:51.760 massive number of variables we were talking about, that the 0:17:51.840 --> 0:17:55.800 whole unrestricted, unpredictable thing, And so they would have different 0:17:56.080 --> 0:17:58.840 uh kind of like like overlays. Think of him as 0:17:58.880 --> 0:18:02.000 an overlay that would kind of guide Aliza's responses, And 0:18:02.000 --> 0:18:05.880 the most well known one was called Doctor, which put 0:18:05.880 --> 0:18:11.080 Eliza in the role of a Rogerian psychiatrist. Uh. This 0:18:11.160 --> 0:18:13.959 is the person who responds to everything with a question. 0:18:14.040 --> 0:18:17.760 Al right. It's that it's that passive interview style where 0:18:17.760 --> 0:18:20.719 where you know, you repeat back. You know, if if 0:18:20.920 --> 0:18:23.080 if I go, oh, man, like I'm I'm I'm really 0:18:23.119 --> 0:18:26.120 sad about my cat, Oh, tell what is it about 0:18:26.119 --> 0:18:28.280 your cat that makes you sad? And then you can 0:18:28.359 --> 0:18:30.400 say things like you know, And that's one of those 0:18:30.400 --> 0:18:34.639 things that where as a conversation starts to wind down, 0:18:34.960 --> 0:18:37.600 you then have another line of questions, So tell me 0:18:37.640 --> 0:18:40.639 more about your mother, like that the whole tell me 0:18:40.720 --> 0:18:42.879 about your mother thing. That's really coming back to this 0:18:42.960 --> 0:18:45.200 kind of model of psychiatrists. But yeah, if you've ever 0:18:45.240 --> 0:18:48.520 heard the joke about them responding to anything with another question, 0:18:48.960 --> 0:18:51.000 so I just just taking the last word and turning 0:18:51.040 --> 0:18:53.560 it into a question what am I paying you for? 0:18:53.720 --> 0:18:57.800 Why do you think you're paying for? That kind of thing? Uh? 0:18:57.840 --> 0:19:00.880 So that's what that's how Aliza spotted and in fact, 0:19:00.920 --> 0:19:07.080 you can find examples of the Eliza proscript transcripts, and 0:19:07.119 --> 0:19:10.080 even even the actual there's there. There are ports you 0:19:10.119 --> 0:19:13.119 could say, or people have essentially created their own version 0:19:13.119 --> 0:19:16.199 of Eliza just using the originalize as a guide. You 0:19:16.240 --> 0:19:18.439 can find tons of them on the web, and you 0:19:18.480 --> 0:19:21.640 can attempt to have a conversation with them. It's not 0:19:21.880 --> 0:19:25.520 terribly compelling, but it's it's kind of fun. Usually within 0:19:25.640 --> 0:19:28.840 maybe four or five exchanges, you've already run into something 0:19:28.840 --> 0:19:31.760 where you're like, well, this can't be a person, or 0:19:31.800 --> 0:19:33.880 if it's a person, it's the weirdest person I've ever 0:19:34.040 --> 0:19:36.719 conversed with. But but again, it wasn't really an earnest 0:19:36.720 --> 0:19:41.119 attempt to to create something that would pass the quote 0:19:41.160 --> 0:19:43.359 unquote tearing tests, right, which I don't think was being 0:19:43.400 --> 0:19:45.199 referred to as such at that point yet. Yeah, it 0:19:45.240 --> 0:19:48.199 was kind of people knew about tourings prediction, but it 0:19:48.240 --> 0:19:51.400 wasn't so much called a Turing test. Another chat bought 0:19:51.440 --> 0:19:54.480 that premiered in the early nineteen seventies went even further 0:19:54.560 --> 0:19:58.080 and actually was an attempt to try and pass the 0:19:58.119 --> 0:20:01.160 Turing test in a very specific approach, kind of like 0:20:01.560 --> 0:20:04.000 you know, Eugene Goostman is a very specific approach to 0:20:04.119 --> 0:20:07.120 narrow down that those parameters. In this case, this one 0:20:07.200 --> 0:20:10.920 was called Perry p A R R y A. Right, 0:20:10.960 --> 0:20:15.000 and Kenneth Colby created it too to emulate a patient 0:20:15.119 --> 0:20:18.919 who has paranoid schizophrenia. Yeah, someone who has you know, 0:20:18.960 --> 0:20:23.679 sort of a persecution complex. Uh. They imagined that that 0:20:23.840 --> 0:20:27.520 there are people or other entities that are out to 0:20:27.560 --> 0:20:31.720 get them, and in this uh specific case, he kinda 0:20:31.960 --> 0:20:35.919 he kind of really embraced this approach. It reminds me 0:20:35.960 --> 0:20:39.639 of people who create a like a Dungeons and Dragon's character, 0:20:39.680 --> 0:20:42.600 but then give their character an entire backstory. Yeah. Yeah, yeah, 0:20:42.600 --> 0:20:45.720 this this Perry persona was an entire persona. It was 0:20:45.720 --> 0:20:47.560 a twenty eight year old man with a job as 0:20:47.560 --> 0:20:50.200 a post office clerk who was single, had no brothers 0:20:50.200 --> 0:20:54.000 and sisters, and rarely saw his parents. He had specific hobbies. 0:20:54.080 --> 0:20:57.200 He liked to go to the movies and horse racing. Yeah, 0:20:57.200 --> 0:20:59.800 he liked to bet on the horses. Uh. And he 0:21:00.080 --> 0:21:02.280 placed about with bookies in the past, right, and that 0:21:03.080 --> 0:21:07.520 he realized later bookies have an association with the criminal 0:21:07.640 --> 0:21:11.919 underworld and that therefore the mafia knew about him and 0:21:12.040 --> 0:21:14.199 we're out to get him. Now, now all of that 0:21:14.280 --> 0:21:16.760 might sound very ridiculous to you if you never had 0:21:16.840 --> 0:21:20.920 any kind of interaction with someone who suffers from paranoid schizophrenia, 0:21:21.040 --> 0:21:23.880 that can seem like, well, that seems cartoonish. But no, this, 0:21:23.880 --> 0:21:28.000 this kind of thinking is not uncommon, you know, whether 0:21:28.080 --> 0:21:31.399 it's whether it's a criminal organization or the government or 0:21:31.520 --> 0:21:36.160 some other even unnamed entity. Oh absolutely, uh, it's it's 0:21:36.280 --> 0:21:39.440 very much realistic in terms of that kind of diagnosis. 0:21:39.520 --> 0:21:41.520 And and if I was speaking about it with humor 0:21:41.520 --> 0:21:43.399 in my voice a moment ago, it's it's only because 0:21:43.440 --> 0:21:47.359 I am absolutely tickled that the programmers of this of 0:21:47.440 --> 0:21:51.199 this program built it, built it in. Yeah, I mean 0:21:51.440 --> 0:21:54.360 it's it's actually pretty entertaining that they went so far 0:21:54.400 --> 0:21:57.720 as to make this whole, uh, this whole backstory to explain, 0:21:57.760 --> 0:22:01.160 because that's what gives it the believability, and then they 0:22:01.240 --> 0:22:04.160 ended up testing it in conversation with a human. Perry 0:22:04.200 --> 0:22:08.640 would gradually start to introduce his thoughts quote unquote thoughts 0:22:08.680 --> 0:22:12.359 about being persecuted, and would respond sensitively to anything said 0:22:12.359 --> 0:22:16.280 about his appearance, family, or religious beliefs. I've actually seen 0:22:16.400 --> 0:22:19.320 lots of transcripts of conversations with Perry, and sure enough, 0:22:19.560 --> 0:22:21.920 it's one of those things where you know, you might 0:22:22.280 --> 0:22:24.639 have a few exchanges and then Perry ends up saying 0:22:24.760 --> 0:22:28.520 something that seems really odd, but not so odd as 0:22:28.560 --> 0:22:31.680 to seem artificial. It just seems like it's a non sequitur, 0:22:32.240 --> 0:22:34.280 you know, something like, well, that's what they want you 0:22:34.320 --> 0:22:35.760 to think. And if you were to say, who are 0:22:35.800 --> 0:22:38.840 they the mafia? Who did you think I was talking about? 0:22:39.520 --> 0:22:42.760 Like the mafia after you? Of course they're after me. 0:22:43.240 --> 0:22:44.879 They know who I am. That kind of stuff, and 0:22:45.119 --> 0:22:48.040 it's disturbing, like it's you know, when you know what 0:22:48.080 --> 0:22:50.199 it is, it's kind of amusing, but like if you're 0:22:50.200 --> 0:22:55.040 in the middle of a conversation, join, oh this poor person. Yeah. 0:22:55.200 --> 0:22:58.560 And so in order to test it, Kolby did a 0:22:58.560 --> 0:23:01.880 couple of different things. Did one test where eight psychiatrists 0:23:01.920 --> 0:23:06.040 interviewed both Perry and a human patient via teletypewriter. So 0:23:06.359 --> 0:23:10.879 in both cases the the the psychiatrists could not see 0:23:10.920 --> 0:23:12.639