WEBVTT - Elementary, My Dear Watson 0:00:00.280 --> 0:00:02.840 Brought to you by the reinvented two thousand twelve camera. 0:00:03.160 --> 0:00:08.920 It's ready. Are you get in touch with technology with 0:00:09.039 --> 0:00:17.880 tech Stuff from how stuff works dot com. Hello again, everyone, 0:00:17.920 --> 0:00:20.119 Welcome to tech Stuff. My name is Chris Poulette and 0:00:20.120 --> 0:00:22.320 I am an editor at how stuff works dot Com. 0:00:22.320 --> 0:00:25.760 Sitting across from me, as always, is senior writer Jonathan Strickland. 0:00:25.920 --> 0:00:32.080 The game is afoot Okay. This episode is about a 0:00:32.240 --> 0:00:37.320 system created by IBM as a scientific experiment to determine 0:00:37.360 --> 0:00:41.640 whether a computer can beat a human in a game 0:00:41.680 --> 0:00:47.680 of skill and intelligence. Jonathan, what is Watson? That is correct? 0:00:49.640 --> 0:01:00.040 I like all I would too? And Big Bucks? Are you? 0:01:00.400 --> 0:01:06.080 Are you a giant computer? Sorry's really reaching back now. 0:01:06.120 --> 0:01:08.280 I would like to tell you my sob story about 0:01:08.319 --> 0:01:11.000 my life so I can win a new refrigerator. There. 0:01:11.120 --> 0:01:13.440 That's reaching back, and it's really obscure. If you know 0:01:13.480 --> 0:01:16.200 what I'm referring to with that particular game show, let 0:01:16.280 --> 0:01:18.480 me know, sadly I do. So I'm just gonna stay 0:01:18.480 --> 0:01:20.520 out at this. I'm not eligible to win. I read 0:01:20.520 --> 0:01:24.560 the rules. So we're gonna talk today about the Watson computer. 0:01:24.680 --> 0:01:26.800 We actually had a lot of listeners right in about 0:01:26.800 --> 0:01:30.320 this because The announcement of the Watson computer came shortly 0:01:30.400 --> 0:01:33.800 after we are episode on. Actually, I think it might 0:01:33.840 --> 0:01:36.760 have even been just before our episode about Computers Versus 0:01:36.840 --> 0:01:40.080 Humans published, So of course it looked like we had 0:01:40.080 --> 0:01:46.440 a glaring omission. Yes, but by in our defense, we 0:01:46.480 --> 0:01:49.360 didn't know about it yet. Yes, actually we mentioned one 0:01:49.400 --> 0:01:59.320 of Watson's cousins predecessors, is probably a predecessor of processors. Yeah. Actually, um, 0:01:59.440 --> 0:02:02.480 deep Blue, I'm sorry, Deep Blue, Deep Blue, Big Blue 0:02:02.480 --> 0:02:05.200 would be the company that made it. But the the 0:02:07.640 --> 0:02:10.640 we're talking about IBM, and IBM does this thing occasionally 0:02:10.720 --> 0:02:13.600 where they issue Yeah, well it is a thing, I mean, 0:02:13.639 --> 0:02:16.880 it's it's because it's not just Deep Blue, it's not 0:02:16.960 --> 0:02:21.600 just Watson. They issue what they call grand challenges their 0:02:21.639 --> 0:02:25.160 engineering teams. Yes, they've had a series of these, and 0:02:25.400 --> 0:02:29.320 some of them are are more noticeable to the public, 0:02:29.360 --> 0:02:32.120 I guess, and others. Deep Blue would definitely be one 0:02:32.160 --> 0:02:34.959 of those because that made headlines. In the nineties. Deep 0:02:34.960 --> 0:02:39.040 Blue was of course the computer that challenged Gary Kasparov, 0:02:39.280 --> 0:02:44.600 the chess grand Master um to a series of games. 0:02:44.880 --> 0:02:48.840 In the first series of games, Kasparov was emerged victorious, 0:02:48.840 --> 0:02:52.320 and in the second Deep blue one, and so that 0:02:52.440 --> 0:02:54.240 was one of those things that kind of propelled the 0:02:54.240 --> 0:02:58.120 whole idea of computers being able to outwit humans, to 0:02:58.120 --> 0:03:02.200 be able to outperform humans in certain tasks. But there 0:03:02.200 --> 0:03:05.320 were other tasks that humans were still much much more 0:03:05.440 --> 0:03:10.280 capable of completing than computers. And UM, as it turns out, 0:03:10.280 --> 0:03:13.440 Watson is a grand challenge. To answer one of those, 0:03:14.040 --> 0:03:16.320 so to speak, or maybe question one of those would 0:03:16.320 --> 0:03:17.520 be better because you have to put it in the 0:03:17.520 --> 0:03:20.320 form of a question, right, That's that's correct. UM. I 0:03:20.320 --> 0:03:23.480 would imagine that Watson does this flawlessly. But we could 0:03:23.480 --> 0:03:26.080 talk about the differences in a human opponent and a 0:03:26.120 --> 0:03:28.760 computer opponent in a little bit. UM. I wanted to 0:03:28.760 --> 0:03:31.040 get into some of the details. Watson is not actually 0:03:31.400 --> 0:03:35.920 a single computer as I typically think about it. UM. 0:03:35.960 --> 0:03:38.880 It is made of ten racks of IBM power, seven 0:03:38.880 --> 0:03:42.560 fifty servers using the Linux operating system. How many cores 0:03:42.720 --> 0:03:46.520 does it have? Two thousand, eight hundred eight processor cores 0:03:46.800 --> 0:03:50.880 wholly free holies? Have you thought your quad core processor 0:03:51.120 --> 0:03:55.040 was the bees knees? I also thought my, uh my 0:03:55.160 --> 0:04:00.360 computers for gigabytes of RAM were pretty much for what 0:04:00.400 --> 0:04:04.520 I'm doing. But Watson has fifteen terabytes of RAM. A 0:04:04.640 --> 0:04:10.000 terabyte is one thousand, twenty four gigabytes, that's right. Also, 0:04:10.880 --> 0:04:14.040 it computes eight at the rate of eighty tarra flops, 0:04:14.120 --> 0:04:19.240 which is eighty trillion calculations per second. And in fact, 0:04:20.200 --> 0:04:23.960 I understand from reading IBM's website about Watson that it 0:04:24.080 --> 0:04:29.120 has somewhere in the neighborhood of two million books essentially. 0:04:29.160 --> 0:04:30.960 I mean, that's it's it's kind of hard to say 0:04:31.000 --> 0:04:34.720 how much information is in a book, but um more 0:04:34.760 --> 0:04:36.640 or less two million books, and it can scan the 0:04:36.839 --> 0:04:42.000 entirety of information on all of those hard drives in 0:04:42.040 --> 0:04:46.760 that machine in roughly two to three second. Right. The 0:04:46.800 --> 0:04:49.679 idea here is that they needed to create a computer. 0:04:49.960 --> 0:04:51.800 You have, the whole the whole challenge here was to 0:04:51.800 --> 0:04:54.200 create a computer that could compete in a game of 0:04:54.279 --> 0:04:57.800 Jeopardy and compete on a championship level. Yeah. And as 0:04:57.839 --> 0:04:59.960 a matter of fact, when we talked about the computer 0:05:00.120 --> 0:05:06.000 to versus person challenge in that podcast, we were discussing how, 0:05:06.440 --> 0:05:09.040 you know, computers do some things really really well and 0:05:09.120 --> 0:05:11.160 some things they don't do so well. And ib AM 0:05:11.200 --> 0:05:14.360 freely admitted that this was a real toughie. Yeah, because 0:05:14.400 --> 0:05:16.200 as it turns out one of the things computers do 0:05:16.240 --> 0:05:19.200 really well. They do well with things like like logical problems, 0:05:19.800 --> 0:05:23.480 you know, because you follow a very set a series 0:05:23.480 --> 0:05:28.440 of steps, things that that obey specific rules. The English 0:05:28.560 --> 0:05:35.480 language does not obey rules as strictly as a mathematical formula. Yes, 0:05:36.160 --> 0:05:38.400 as a matter of fact, we we sort of go 0:05:38.560 --> 0:05:42.200 with with things that might be tricky for computers to 0:05:42.240 --> 0:05:45.000 understand all the time because we constantly on this show 0:05:45.040 --> 0:05:49.360 do wordplay and puns, um, and computers may not necessarily 0:05:49.440 --> 0:05:52.680 understand the nuances of such things, or or slang, or 0:05:52.760 --> 0:05:57.120 metaphors or metaphors. Um. There's a lot of elements to 0:05:57.200 --> 0:06:00.640 human speech that we naturally understand as we develop our 0:06:00.720 --> 0:06:03.599 language skills. Right speak for yourself, I have no idea 0:06:03.640 --> 0:06:07.279 how this thing works, okay, but most of us figure 0:06:07.279 --> 0:06:11.280 out how to determine what someone is talking about based 0:06:11.360 --> 0:06:15.280 on contextual clues and our knowledge of things like wordplay 0:06:15.320 --> 0:06:18.400 and metaphors. So as we build our vocabulary, as we 0:06:18.440 --> 0:06:23.040 build our ability to create sentences, as we understand concepts 0:06:23.080 --> 0:06:28.000 that are not necessarily concrete, then we are able to 0:06:28.080 --> 0:06:33.240 communicate in a more ambiguous way than a computer would 0:06:33.279 --> 0:06:37.640 necessarily be capable of on any normal computer. That is, So, 0:06:37.680 --> 0:06:39.919 what are you trying to say, Johnny get Yeah, what 0:06:39.960 --> 0:06:41.640 I'm trying to say is that I'm trying to say 0:06:41.680 --> 0:06:45.160 is that the depending on the way you word a sentence, Uh, 0:06:45.240 --> 0:06:48.080 a human might be able to determine immediately what the 0:06:48.120 --> 0:06:51.240 significance is of the sentence. You know, what you just said. 0:06:51.240 --> 0:06:53.640 They'd be able to understand it. A computer, depending upon 0:06:53.640 --> 0:06:56.159 the wording, may not be able to interpret it properly 0:06:56.240 --> 0:06:59.839 because you know, you didn't necessarily say like, the ball 0:07:00.120 --> 0:07:03.280 is blue. You know, you might have used a much 0:07:03.320 --> 0:07:06.160 more poetic way of saying it that a computer just 0:07:06.200 --> 0:07:09.479 can't you know, the computer can't equate that as being 0:07:09.560 --> 0:07:11.880 the ball is blue. But any human listener would be 0:07:12.760 --> 0:07:15.040 able to understand what you were getting at and say, oh, 0:07:15.080 --> 0:07:17.400 it's a blue ball. It was just a really fancy, 0:07:17.520 --> 0:07:21.680 flowery way of saying that. Yes, Um, I watched a 0:07:21.760 --> 0:07:24.880 number of videos on the IBM site and some of 0:07:24.880 --> 0:07:29.160 them are quite amusing. Actually, uh, because the early versions 0:07:29.160 --> 0:07:32.600 of Watson just didn't get it. Yeah, they weren't. They 0:07:32.600 --> 0:07:37.080 weren't the most um accurate. And what what's funny about 0:07:37.160 --> 0:07:39.760 is not that the computer didn't get it. But the 0:07:39.880 --> 0:07:43.600 looks on the engineer's faces and as they were going, yeah, okay, no, 0:07:43.760 --> 0:07:45.920 maybe not not so much. We have to go back 0:07:45.960 --> 0:07:50.440 to the drawing board. But Dr Chris Welty was saying 0:07:50.640 --> 0:07:53.600 the point of this exercise is to do the science 0:07:54.120 --> 0:07:57.560 behind this and and they specifically we're looking forward to 0:07:57.640 --> 0:08:01.880 the challenge of Jeopardy and UM. You know, if you 0:08:02.520 --> 0:08:05.040 if you're unfamiliar with the show UM, which some of 0:08:05.080 --> 0:08:08.520 you maybe uh a lot of the questions. Of course, 0:08:08.560 --> 0:08:12.840 the the the answers are presented first. UH. The contestants 0:08:12.840 --> 0:08:15.520 are given the opportunity to choose one of six categories 0:08:15.520 --> 0:08:19.600 that are on the board at different values UH monetary 0:08:19.640 --> 0:08:24.000 values UM. And so you can expect in these categories 0:08:24.040 --> 0:08:27.160 that the the answers UH you are actually supposed to 0:08:27.160 --> 0:08:28.920 give the question if you are contestant on the game. 0:08:29.080 --> 0:08:33.400 The answers can fall within a certain domain of knowledge UM. 0:08:33.440 --> 0:08:38.040 For example, the infamous Potent Potables category UM is about 0:08:38.200 --> 0:08:41.480 alcoholic drinks, and you can expect that if you are 0:08:41.640 --> 0:08:44.560 fairly knowledgeable about different kinds of drinks that you might 0:08:44.840 --> 0:08:47.839 do well or poorly in the category. So you should 0:08:47.840 --> 0:08:52.920 either choose questions or answers from the category or not. Um. Well, 0:08:53.280 --> 0:08:55.240 you know, if no one has bothered to program that 0:08:55.280 --> 0:08:59.440 information into Watson, Uh, then Watson will do poorly in 0:08:59.440 --> 0:09:03.920 that category. But some of the categories on Jeopardy are 0:09:03.920 --> 0:09:05.959 written with a lot of word smithing involved, so you 0:09:06.040 --> 0:09:09.640 might have to supply an answer that rhymes or unscramble 0:09:09.840 --> 0:09:13.120 the war letters to do to form another word. Now, 0:09:13.120 --> 0:09:17.240 the unscrambling thing might come very easy to a computer, um, 0:09:17.280 --> 0:09:20.960 but the rhyming answer, you'd have to go over a 0:09:20.960 --> 0:09:23.360 lot of synonyms in your head to try to find. Okay, well, 0:09:23.400 --> 0:09:25.960 I know the answer to this question, but it obviously 0:09:26.000 --> 0:09:29.600 isn't going to rhyme right. So um. Dr Welty said, 0:09:29.720 --> 0:09:31.000 you know, this is one of the things that we 0:09:31.000 --> 0:09:33.920 were really looking forward to. We wanted, we wanted to challenge. 0:09:33.920 --> 0:09:36.439 We wanted the computer to be answered able to answer 0:09:36.559 --> 0:09:40.839 questions or question answers that the computer normally wouldn't be 0:09:40.880 --> 0:09:43.200 able to. So they were really looking forward to cracking 0:09:43.200 --> 0:09:46.640 this nut, so to speak. Um. They talked about there 0:09:46.640 --> 0:09:50.400 being five major areas that they had to concentrate on 0:09:50.480 --> 0:09:54.280 in order to make Watson work based upon the way 0:09:54.360 --> 0:09:59.600 Jeopardy works, because again they designed this project with a 0:09:59.679 --> 0:10:02.640 very specific application in mind. It helped give them direction 0:10:02.720 --> 0:10:04.640 as opposed to it just being I just want to 0:10:04.679 --> 0:10:08.520 make a computer that is able to analyze semantics and 0:10:08.520 --> 0:10:11.439 and respond. Um. That's you know, that's a much more 0:10:11.480 --> 0:10:14.240 general approach. By giving them the fact that, okay, well, 0:10:14.280 --> 0:10:16.240 our goal is to be able to create a computer 0:10:16.280 --> 0:10:21.200 that can compete and potentially beat champions in Jeopardy, Uh, 0:10:21.360 --> 0:10:24.480 it provided more focus. So with Jeopardy in mind, they 0:10:24.480 --> 0:10:26.360 said the five things they needed to concentrate on was 0:10:26.360 --> 0:10:30.560 that Jeopardy creates a broad and open domain, which means 0:10:30.600 --> 0:10:34.200 that you don't just get questions about one subject. Yes, 0:10:34.280 --> 0:10:35.760 you're not going to have to know everything there is 0:10:35.800 --> 0:10:38.000 to know about alcoholic drinks and that's the only thing 0:10:38.040 --> 0:10:39.760 you were going to be asked about. Right There might 0:10:39.800 --> 0:10:46.120 be politics, pop culture, sports, literature, all sorts of categories 0:10:46.120 --> 0:10:49.080 that you could potentially come up against. So with that 0:10:49.120 --> 0:10:51.720 in mind, the computer had to be able to answer 0:10:51.760 --> 0:10:56.360 those things. Uh. There were as Chris was saying, there 0:10:56.440 --> 0:11:01.560 was an element of complex language. Jeopardy answers can be tricky. 0:11:02.000 --> 0:11:04.520 They're not necessarily straightforward. It's kind of like the New 0:11:04.559 --> 0:11:07.360 York Times crossword puzzle. If you read the clues to 0:11:07.400 --> 0:11:11.760 that crossword puzzle, they aren't necessarily straightforward. They require you 0:11:11.840 --> 0:11:15.160 to make some You have to bridge some gaps in 0:11:15.240 --> 0:11:18.760 order to get to the right answer yes. And in fact, 0:11:18.840 --> 0:11:22.040 they will ask you even in clues for for that puzzle. 0:11:22.120 --> 0:11:24.680 They will ask you for things in poetic language, and 0:11:24.720 --> 0:11:26.760 you'll have to think about things in a completely different 0:11:26.760 --> 0:11:30.880 way than you might have otherwise. The next area that 0:11:30.920 --> 0:11:33.560 they had to focus on was high precision, so you 0:11:33.600 --> 0:11:37.559 had to be able to narrow down your choices and 0:11:37.640 --> 0:11:41.679 find out which of your potential answers would be the most, 0:11:42.240 --> 0:11:46.160 the most accurate, or the best one to choose. Along 0:11:46.200 --> 0:11:49.000 with that was accurate confidence, which means that the computer 0:11:49.040 --> 0:11:52.000 itself has to be able to determine how likely is 0:11:52.080 --> 0:11:56.000 this answer? How likely is this the right answer? Yes? Right, 0:11:56.400 --> 0:11:58.760 and um. And then the last one was high speed. 0:11:58.760 --> 0:12:01.200 It had to be a really really fast computer in 0:12:01.280 --> 0:12:04.880 order to compete against people, because if you know something, 0:12:05.120 --> 0:12:07.920 you just you just spout it out right, you know, 0:12:07.960 --> 0:12:11.559 you you buzz and you say, who is Marshall brain? 0:12:12.200 --> 0:12:14.840 You know? And then you've got the answer, who is 0:12:15.280 --> 0:12:18.120 Marshal brain? I think only one person can answer that question, 0:12:18.360 --> 0:12:21.680 and he is not in the studio today. UM. But yeah, 0:12:21.720 --> 0:12:26.559 you have to have computers capable of of accessing all 0:12:26.600 --> 0:12:29.560 this information and picking it out as quickly as a 0:12:29.600 --> 0:12:32.480 human would be able to. UM. In fact, I saw 0:12:32.520 --> 0:12:36.280 on one of these videos that uh, if you had 0:12:36.960 --> 0:12:42.680 a two point six giga Hurts core processor a computer 0:12:42.760 --> 0:12:45.280 running one of those Okay, posably, I do own a 0:12:45.280 --> 0:12:47.719 computer with a two point six gigga Hurts process right, 0:12:47.800 --> 0:12:49.920 so you know, kind of a middle of the road 0:12:50.000 --> 0:12:53.880 computer right now. But but two point six gigga Hurts computer. 0:12:53.920 --> 0:12:57.960 If you were to try and answer one question uh, 0:12:57.960 --> 0:13:02.679 and you were going to go through all of Watson's 0:13:03.000 --> 0:13:06.319 UH data in order to find that question, the answer 0:13:06.360 --> 0:13:08.839 to that question and compare all the answers and come 0:13:08.920 --> 0:13:12.120 up with the best result and then presented, it would 0:13:12.120 --> 0:13:15.959 take you two hours for that one computer. It doesn't 0:13:15.960 --> 0:13:19.199 surprise me much. So that's why you have that two 0:13:19.240 --> 0:13:23.840 thousand eight processor. You know that with all the different 0:13:23.920 --> 0:13:26.800 uh the web servers running, you have to have those 0:13:26.840 --> 0:13:30.280 core processors running so that you can solve these questions 0:13:30.320 --> 0:13:33.719 in parallel. Excuse me, And you probably remember us talking 0:13:33.760 --> 0:13:37.600 about parallel computing and other podcasts. That's the idea that 0:13:37.640 --> 0:13:39.680 you try and solve a problem by working on parts 0:13:39.720 --> 0:13:42.080 of the problem all at the same time. In this case, 0:13:42.640 --> 0:13:47.679 Watson gets the the answer from Jeopardy and then goes 0:13:47.760 --> 0:13:51.520 through and tries to process all the potential questions that 0:13:51.600 --> 0:13:54.640 would be the correct response to that answer, and then 0:13:54.640 --> 0:13:56.920 it has to evaluate them and choose the right one, 0:13:57.360 --> 0:13:59.520 and has to do this in just a couple of seconds. 0:14:00.960 --> 0:14:06.720 It's a pretty cool idea. The the challenges are not trivial, 0:14:09.679 --> 0:14:14.560 the answers are, but not the the challenges um and 0:14:14.600 --> 0:14:16.800 like you were saying, the early tests were very amusing 0:14:16.840 --> 0:14:19.880 because Watson just didn't get it. It would it would 0:14:19.880 --> 0:14:23.560 give answers that were obviously related to the question, or 0:14:23.600 --> 0:14:26.760 at least related to words that were within the question, 0:14:26.800 --> 0:14:29.760 but we're not the right answer. It's kind of like 0:14:29.800 --> 0:14:31.920 if you were ever using a search engine and you 0:14:32.000 --> 0:14:35.360 put in certain terms and the results you're getting back 0:14:35.920 --> 0:14:38.240 are related to the terms you put in, but not 0:14:38.280 --> 0:14:42.080 to the subject matter you wanted, because it's maybe using hominem's, 0:14:42.240 --> 0:14:45.960 or it's using synonyms, or it's or maybe you misspelled 0:14:46.000 --> 0:14:48.400 something or whatever. But anyway, you're getting the wrong kind 0:14:48.400 --> 0:14:53.840 of responses, same sort of thing. Yep. And speaking of trivial, 0:14:54.160 --> 0:14:56.800 I did want to point out to that Dr Kelly, 0:14:56.920 --> 0:14:59.040 Dr John E. Kelly the third He is a senior 0:14:59.120 --> 0:15:01.760 vice president of ib i'm in the director of IBM Research. 0:15:02.480 --> 0:15:06.480 Um this the project itself, you know, Yes, they're building 0:15:06.520 --> 0:15:10.680 a computer to win a trivia contest, so that might 0:15:10.800 --> 0:15:16.360 seem trivial. Yes, However, the point is, you know, Dr 0:15:16.440 --> 0:15:20.800 Kelly was saying, Look, the amount of information that is 0:15:20.960 --> 0:15:27.840 being created today is rapidly uh, overcoming our ability to 0:15:29.160 --> 0:15:31.960 identify it, process it, makes sense of it, and and 0:15:31.960 --> 0:15:34.840 and derive knowledge from it. Yeah. In fact, I think 0:15:34.880 --> 0:15:38.520 it is a fifteen petabytes of data raw data get 0:15:38.560 --> 0:15:41.240 generated every day, not just not just from people but 0:15:41.280 --> 0:15:44.520 from machines as well. But that's that's an insane amount 0:15:44.520 --> 0:15:47.000 of information. Yes, yes, now, I mean, the human mind 0:15:47.040 --> 0:15:49.200 is a remarkable thing, and if you have systems in place, 0:15:49.240 --> 0:15:53.240 you can help manage that. But at some point, uh, 0:15:53.280 --> 0:15:55.160 you know, even even people can't keep up with that. 0:15:55.200 --> 0:15:59.800 Even there are remarkable computing machines and our skulls. So uh, 0:15:59.840 --> 0:16:03.200 the idea is to build a tool that can actually 0:16:03.400 --> 0:16:06.960 help people. There will be a tool for people to 0:16:06.960 --> 0:16:10.880 help people make sense of this vast amount of information 0:16:11.360 --> 0:16:13.800 and and to overcome that and get get real help 0:16:13.840 --> 0:16:19.720 I guess from machines and and help people understand or 0:16:19.840 --> 0:16:23.920 navigate the world of information that is rapidly creating. UM. 0:16:24.120 --> 0:16:26.400 One of the cooler videos on this site I think 0:16:27.400 --> 0:16:29.520 was the one where they were explaining, look, there there's 0:16:29.560 --> 0:16:33.640 always been this interconnected system of information going on all 0:16:33.680 --> 0:16:36.560 over the world, but we didn't really understand it nearly 0:16:36.600 --> 0:16:40.560 as well. Until the Internet came around. We could actually 0:16:40.640 --> 0:16:43.560 see what was going on, you know, in seconds, rather 0:16:43.600 --> 0:16:47.120 than you know, having it take hours or days or 0:16:47.160 --> 0:16:52.120 weeks or months or even years in many many years past. UM, 0:16:52.440 --> 0:16:56.240 and it's it's just enabled this and is accelerating the problem. 0:16:56.320 --> 0:17:00.400 So UM, the challenge of creating the computer to play 0:17:00.400 --> 0:17:03.440 the game, well, this is basically, I guess an exercise 0:17:03.600 --> 0:17:06.359 to see can we really do this? Can we create 0:17:07.080 --> 0:17:12.600 uh reasonably intelligent computer that can help us, you know, 0:17:12.640 --> 0:17:16.119 figure out what's going on and where the the answers 0:17:16.119 --> 0:17:18.919 are to our questions? Can can we create a computer 0:17:19.000 --> 0:17:23.960 that can understand natural language so that that you challenge it, right, 0:17:24.160 --> 0:17:27.000 It's it's not it's not that you have to tailor 0:17:27.080 --> 0:17:29.480 your language to the computer so that it understands I mean, 0:17:29.560 --> 0:17:31.720 we were familiar with that. You know, we talked about 0:17:31.720 --> 0:17:34.880 Boollyan logic before, about how if you want to do 0:17:35.160 --> 0:17:38.600 really effective search terms, you need to understand how Booleyan 0:17:38.640 --> 0:17:41.280 logic works so that you can. Because search engines don't 0:17:41.359 --> 0:17:45.080 understand natural language, they'll do their best to try and 0:17:45.119 --> 0:17:48.520 match your query with the right result, but they don't 0:17:48.600 --> 0:17:53.040 understand it. They aren't able to analyze the information. One 0:17:53.080 --> 0:17:56.520 of the concepts that it was really important with Watson 0:17:56.920 --> 0:17:58.879 is one that's going to be very important if we 0:17:58.960 --> 0:18:02.160 ever are to have us semantic web, which is the 0:18:02.200 --> 0:18:06.080 idea that you could talk to your computer, whether you're 0:18:06.240 --> 0:18:09.240 actually speaking or typing or whatever. You you can communicate 0:18:09.240 --> 0:18:11.840 with your computer in a natural way, and the computer 0:18:11.880 --> 0:18:14.880 will be able to understand, at least on some level. 0:18:15.040 --> 0:18:17.160 It may not be a deep level, but be able 0:18:17.160 --> 0:18:20.480 to interpret what you're saying and give you the right result. 0:18:21.000 --> 0:18:24.600 Uh in response, that's right. It just it depends on 0:18:24.680 --> 0:18:29.159 a system of contexts, and without those contexts, and the 0:18:29.160 --> 0:18:32.080 computer has to be able to interpret that well, um, 0:18:32.800 --> 0:18:36.000 you're you know, it's it's not nearly as effective as 0:18:36.040 --> 0:18:39.360 it could be um, So this is this is definitely 0:18:39.359 --> 0:18:41.920 a step in the right direction. Yeah, I think it's 0:18:41.920 --> 0:18:44.679 pretty fascinating the way it talked about how or the 0:18:44.680 --> 0:18:48.240 way the the engineers talked about how the computer comes 0:18:48.320 --> 0:18:50.600 up with its answers. So what it does is it 0:18:50.600 --> 0:18:54.480 will it comes up with candidate answers. This is part 0:18:54.480 --> 0:18:58.159 of that parallel processing where all the potential answers to 0:18:58.200 --> 0:19:01.159 a question pop up, and then it turns each of 0:19:01.200 --> 0:19:06.160 those answers into a hypothesis and then examines each hypothesis 0:19:06.240 --> 0:19:10.399 to determine how likely that hypothesis is in fact the 0:19:10.520 --> 0:19:13.480 right answer, and if it doesn't meet a certain level 0:19:13.560 --> 0:19:18.159 of confidence, then then Watson won't buzz in. So Watson 0:19:18.240 --> 0:19:20.200 is not going to buzz in on every question because 0:19:20.200 --> 0:19:22.119 occasionally there's gonna be a question it's gonna be worded 0:19:22.119 --> 0:19:24.960 in such a way that Watson is not really able 0:19:25.000 --> 0:19:28.520 to interpret what what the answer is or just doesn't 0:19:28.560 --> 0:19:31.120 have the information and database. That's another thing we should 0:19:31.119 --> 0:19:35.000 point out. Watson is completely self contained. Yes, it is 0:19:35.080 --> 0:19:37.480 not hooked up to the Internet, so lest you think 0:19:37.560 --> 0:19:40.119 it is searching on Google, it is not. Right. So 0:19:40.600 --> 0:19:43.520 all the information that Watson has available to it is 0:19:43.720 --> 0:19:47.320 self contained. It doesn't. It cannot get more information during 0:19:47.320 --> 0:19:51.239 the course of a game. Now, in between games, um, 0:19:51.920 --> 0:19:55.160 the people ib folks at IBM where it would update Watson, 0:19:55.359 --> 0:19:58.400 especially with things like pop culture references, so that pop 0:19:58.480 --> 0:20:01.560 so that Watson would be able to interpret questions that 0:20:01.680 --> 0:20:04.000 revolved around pop culture and be able to respond to 0:20:04.040 --> 0:20:06.840 them U or news items, things that just happened in 0:20:06.880 --> 0:20:09.080 the news that would have they'd have to update Watson 0:20:09.119 --> 0:20:11.760 with that information as well. But yeah, the key was 0:20:11.840 --> 0:20:15.640 to be able to let Watson break down a sentence 0:20:15.680 --> 0:20:18.800 and really understand what the sentence was saying, not just 0:20:19.040 --> 0:20:21.640 you know this this must be the object and this 0:20:21.720 --> 0:20:24.560 is the the subject and this is the verb, but 0:20:24.640 --> 0:20:28.360 to really understand what it was saying because uh, context, 0:20:28.400 --> 0:20:30.840 as you were pointing out, is so important. One of 0:20:30.880 --> 0:20:35.760 the elements that they talked about was temporal reasoning. Temporal 0:20:35.840 --> 0:20:39.320 reasoning meaning that, uh, there are different ways of saying 0:20:39.320 --> 0:20:44.359 the same thing. For instance, I could say, uh that, um, 0:20:44.400 --> 0:20:49.320 I graduated twenty years ago, or I could say I graduated, 0:20:51.280 --> 0:20:53.679 or I could say the twenty high school reunion is 0:20:53.720 --> 0:20:56.280 coming up for me. All of those things essentially give 0:20:56.320 --> 0:21:00.120 you the same information. By the way I did not graduate. Um. 0:21:00.160 --> 0:21:03.480 But all that all that information, all those those phrases 0:21:03.480 --> 0:21:08.200 give you the same information that I graduated high school. UM, 0:21:08.200 --> 0:21:10.520 but it's different ways of saying it, and a computer 0:21:10.800 --> 0:21:14.600 does not necessarily know that each of those different sentences 0:21:14.640 --> 0:21:17.080 means the same thing. So they had to find a 0:21:17.119 --> 0:21:21.080 way for Watson to learn that, to learn that there 0:21:21.119 --> 0:21:25.120 are many different ways of conveying the same information using 0:21:25.240 --> 0:21:29.000 totally different sentences. And you'll actually be able to see 0:21:29.000 --> 0:21:32.119 that on on February fourteenth, if you tune in to 0:21:32.160 --> 0:21:35.280 watch the show. That's when it's scheduled to air here 0:21:35.280 --> 0:21:38.320 in the United States. Um. And we we know that, 0:21:38.520 --> 0:21:41.480 we know that it performed pretty well already at least, 0:21:41.680 --> 0:21:44.359 let's kind of get into that. Okay, Sorry, No, I 0:21:44.440 --> 0:21:46.560 just figured after after we you know, we could talk 0:21:46.560 --> 0:21:48.960 about the actual show. It's coming up there. I think 0:21:49.000 --> 0:21:52.280 actually the show itself, uh, this particular episode is going 0:21:52.320 --> 0:21:54.399 to be interesting. But well, I was gonna mention that 0:21:54.440 --> 0:21:59.320 a minute, Okay, uh no, basically one of the things 0:21:59.320 --> 0:22:01.639 that I think is really kind of cool. You're not 0:22:01.680 --> 0:22:04.119 going to be just sitting there watching a box and 0:22:04.200 --> 0:22:07.160 to human opponents, they actually made They actually made an 0:22:07.160 --> 0:22:11.160 interface for people to watch, which I think was probably 0:22:11.200 --> 0:22:14.000 key for Jeopardy because I imagine they would actually want 0:22:14.040 --> 0:22:15.560 to see It's like, well, how do we know what 0:22:15.560 --> 0:22:18.600 it's doing? Um, it could be brewing coffee for all 0:22:18.640 --> 0:22:22.959 we know, um, mr coffee. It has an avatar, then 0:22:23.000 --> 0:22:24.639 you'll see it. It looks kind of like a planet 0:22:24.640 --> 0:22:27.119 with a little uh, I don't know, thought wigglies. What 0:22:27.240 --> 0:22:33.159 do you call those? Illustrated I'd call that Doug's hair. Um. Basically, 0:22:33.400 --> 0:22:36.879 if the computer is feeling I put this in quotes, 0:22:36.920 --> 0:22:39.960 if you don't mind confident, the avatar that you see 0:22:40.040 --> 0:22:43.280 is green, so it has it's feeling pretty sure that 0:22:43.400 --> 0:22:47.160 it's got an answer it can use to to buzz in. However, 0:22:47.240 --> 0:22:50.440 if it doesn't have the correct answer, it will be orange, 0:22:51.160 --> 0:22:53.760 so you will be able to see what's going on, 0:22:53.880 --> 0:22:55.520 and you will also be able to see it thinking 0:22:55.720 --> 0:22:58.800 because as the algorithms are processing information to try to 0:22:58.840 --> 0:23:02.679 find an uh A correct question. It's so weird to 0:23:02.680 --> 0:23:06.080 say in this context, um, the avatar is going to flicker, 0:23:06.240 --> 0:23:08.800 so you'll actually be able to see it in the 0:23:08.840 --> 0:23:12.640 process of trying to determine an answer for itself. Um. Now, 0:23:12.680 --> 0:23:15.320 and in two thousand seven, they started building Watson, which, 0:23:15.320 --> 0:23:18.440 by the way, we didn't mention, I don't think uh uh, 0:23:18.440 --> 0:23:21.160 this is named after IBMS founder Thomas J. Watson nine 0:23:21.160 --> 0:23:26.000 after the h Sir Arthur Arthur Conan Doyle character. Right, 0:23:26.119 --> 0:23:31.440 he's not a doctor who who served in India. Um. 0:23:31.480 --> 0:23:34.439 But yeah, that they actually started working on this problem 0:23:34.520 --> 0:23:37.720 and our project in two thousand seven and didn't really 0:23:37.720 --> 0:23:40.520