WEBVTT - Machines Hate Natural Language 0:00:00.160 --> 0:00:07.240 Brought to you by Toyota. Let's go places. Welcome to 0:00:07.400 --> 0:00:15.560 Forward Thinking. Hey there, everyone, and welcome to Forward Thinking, 0:00:15.600 --> 0:00:17.840 the podcast that looks at the future and says eat 0:00:17.960 --> 0:00:22.520 up or. I'm Jonathan Strickland, I'm Laura, and I'm Joe McCormick. 0:00:23.040 --> 0:00:28.360 Hey guys, Hey Joe. I want you to imagine something. Okay, Okay, Lauren, 0:00:28.560 --> 0:00:34.320 imagine you're human. Check. Okay, Jonathan, you imagine you're a computer. 0:00:35.080 --> 0:00:38.760 Already there, and you got you a big microphone, and 0:00:39.000 --> 0:00:41.279 you happen to be sitting right between Lauren and I 0:00:41.360 --> 0:00:45.520 as we discuss uh recent movies that have come into 0:00:45.560 --> 0:00:49.120 the theaters. We frequently do that. Okay, And I say 0:00:49.200 --> 0:00:54.720 to Lauren, Hey, Lauren, did you see her? Interesting? Okay, 0:00:54.760 --> 0:00:58.040 I'm with you so far. Do you understand what the 0:00:58.040 --> 0:01:01.840 heck I am talking about? Jonathan? Know, I never understand 0:01:01.960 --> 0:01:04.720 what you and Lauren are talking about. Ever, you know 0:01:04.840 --> 0:01:08.520 I'm saying in your in your computer imagination world, Oh God, 0:01:08.600 --> 0:01:12.000 got in this scenario? No, No, I don't understand what 0:01:12.000 --> 0:01:16.480 you're saying. That's perfectly simple to anybody who speaks English 0:01:16.520 --> 0:01:21.160 and in this sort of modern English lingo are familiar. 0:01:21.200 --> 0:01:28.080 What with what movies are recently came out. There's ambiguity 0:01:28.080 --> 0:01:30.920 in that sentence already, even if I weren't a computer, 0:01:31.240 --> 0:01:34.319 because unless I know for a fact you're speaking about 0:01:34.319 --> 0:01:37.040 a movie ahead of time, I just think you're using 0:01:37.080 --> 0:01:40.000 a a pronoun in a way where you could be 0:01:40.240 --> 0:01:45.039 literally talking about anyone, Yeah, anyone who's female. But when 0:01:45.080 --> 0:01:48.760 we have everyday conversations like that, asking if somebody saw 0:01:48.800 --> 0:01:52.240 a movie or anything, that's simple, you don't think about 0:01:52.320 --> 0:01:57.000 how much incredibly complex figuring out you're doing in real 0:01:57.160 --> 0:02:00.760 time to make sense of their words shared so much context, 0:02:00.880 --> 0:02:04.640 the actual literal data is an entirely separate issue. Yeah, 0:02:04.680 --> 0:02:07.720 even if even if we were all to agree upon 0:02:08.400 --> 0:02:10.679 what we were talking about ahead of time, there's still 0:02:10.880 --> 0:02:15.639 so many idiomatic ways of putting information in various languages. Uh, 0:02:15.639 --> 0:02:18.840 there are things like figures of speech, there's metaphors, all 0:02:18.840 --> 0:02:21.800 these other sarcasms. Sarcasm is a great example. There are 0:02:21.840 --> 0:02:25.560 all these ways that we can say sarcastically. Sarcasm is 0:02:25.600 --> 0:02:29.320 a great example. But at any rate, there are all 0:02:29.360 --> 0:02:31.840 these different techniques we can use to communicate with one 0:02:31.840 --> 0:02:35.400 another that as you grow up in and around a language, 0:02:35.560 --> 0:02:37.760 you start to get a grasp on it naturally, right, 0:02:37.800 --> 0:02:41.359 you have natural affinity for that language, and you understand 0:02:41.440 --> 0:02:45.040 what the intent is, even if on the surface level 0:02:45.160 --> 0:02:49.359 it's different than what's the true underlying information. Right, so 0:02:49.600 --> 0:02:52.480 even if you say, great job at opening the pod 0:02:52.520 --> 0:02:56.519 bay doors, how, he's probably not gonna get you. No, 0:02:56.639 --> 0:02:58.960 How's just gonna think, Wow, I thought I did a 0:02:59.080 --> 0:03:02.840 terrible job, but every one else didn't notice. Well, how 0:03:02.840 --> 0:03:06.560 how would probably understand? Because how is really clever like that? Yeah, well, 0:03:06.560 --> 0:03:10.440 there are a lot of these clever uh speech understanding 0:03:10.639 --> 0:03:13.880 robots and computers in science fiction. You've got you've got 0:03:13.919 --> 0:03:16.839 all the cartoon ones like Rosie from the Jetsons, and 0:03:16.919 --> 0:03:19.959 you've got Hall nine thousand, who people just talked to 0:03:20.120 --> 0:03:23.000 and he responds in this calm voice and explains your 0:03:23.000 --> 0:03:27.959 doom sings Daisy right, uh, star Trek Star trek. Yeah, 0:03:28.040 --> 0:03:32.080 the computer's voice, which was voiced by Gene Roddenberry's wife, right, um, 0:03:32.120 --> 0:03:35.440 Margel Barrett yep so uh. She also played as I 0:03:35.480 --> 0:03:38.400 recalled Diana Troy's mother. Yes, in the series. I'm still 0:03:38.400 --> 0:03:40.680 not sure whether it was sweet or creepy that that 0:03:40.760 --> 0:03:43.600 he had her playing. Yeah, let's not get into that. 0:03:43.600 --> 0:03:46.000 That's that's that's a podcast all on its own. I 0:03:46.040 --> 0:03:48.080 don't know, I mean more more the ship's computer. But 0:03:48.080 --> 0:03:50.680 but but anyway, um yeah, and so we have all 0:03:50.720 --> 0:03:54.480 of these terrific examples in science fiction, but it's not 0:03:54.560 --> 0:03:56.960 quite living up to that. Reality is not quite living 0:03:57.000 --> 0:04:01.360 up to that. Yeah, Sirie isn't quite the inter prize computer. No, 0:04:01.520 --> 0:04:05.840 it's it's pretty good series, pretty good at voice recognition 0:04:05.840 --> 0:04:10.920 and responding. Google has its own Google Search that you 0:04:10.960 --> 0:04:13.800 can use with voice. Also can do things like transcribe 0:04:14.160 --> 0:04:17.120 a message you are speaking into a microphone as a 0:04:17.160 --> 0:04:20.480 text message or or whatever, or instant message. I mean, 0:04:20.600 --> 0:04:22.160 if you have an Android phone, you can pretty much 0:04:22.160 --> 0:04:26.600 do everything through voice. Then you've got uh the Google Glass. 0:04:26.760 --> 0:04:29.400 Google Glass has voice control in it, although it also 0:04:29.560 --> 0:04:31.480 is very limited in what you can do. You can't 0:04:31.480 --> 0:04:35.080 just say anything. You have very specific command words. Wouldn't 0:04:35.120 --> 0:04:37.320 it be great if like to talk to people like 0:04:37.360 --> 0:04:41.760 you could not get their attention without saying okay Lauren. Yeah, 0:04:41.800 --> 0:04:44.680 that that's my idea of the perfect feature. Let's enact that, right. 0:04:44.720 --> 0:04:47.920 I actually know some people that that's pretty much true anyway, 0:04:48.040 --> 0:04:51.160 but um yeah, and and I even had a joking 0:04:51.160 --> 0:04:53.120 little reference in here, like even even if you look 0:04:53.160 --> 0:04:55.960 back to Ferbie, Ferbie could respond to to voice commands, 0:04:55.960 --> 0:04:59.599 although Furbie spoken Furbish and would have gradually learned English, 0:04:59.640 --> 0:05:01.960 and it wasn't really responding in the way we're talking 0:05:01.960 --> 0:05:04.080 about here. Yeah, but one thing that I think is 0:05:04.200 --> 0:05:08.760 really important to make the distinction between is natural language 0:05:08.839 --> 0:05:14.120 and voice commands. Right, being able to interpret sound as 0:05:14.480 --> 0:05:19.719 language is one thing, and and interpreting the context of 0:05:19.760 --> 0:05:23.280 words is something else entirely. Let's back up. Let's so 0:05:23.400 --> 0:05:27.840 we're going to talk about computers and robots understanding natural language. 0:05:28.680 --> 0:05:30.960 Let's step back and look at what is the essential 0:05:31.080 --> 0:05:35.039 difference between the language that computers understand and like to 0:05:35.160 --> 0:05:38.760 use in the language that real that human beings used. Okay, 0:05:38.800 --> 0:05:42.240 so human language or natural language as we will refer 0:05:42.320 --> 0:05:44.680 to it in this podcast, and that's you know, generally 0:05:44.720 --> 0:05:48.120 accepted term natural language processing would mean using machines that 0:05:48.120 --> 0:05:51.799 could actually process uh commands are written out in whatever 0:05:52.040 --> 0:05:55.880 native language you happen to be working in human language. 0:05:56.320 --> 0:05:59.520 They include things like syntax, grammar. They have these rules 0:05:59.640 --> 0:06:03.479 that structure that give you the framework you need in 0:06:03.520 --> 0:06:06.000 order to communicate an idea to someone else who speaks 0:06:06.040 --> 0:06:09.280 that language, and then a vocabulary of words for conveying 0:06:09.320 --> 0:06:13.000 specific ideas. Right, So, if you didn't have a vocabulary, 0:06:13.040 --> 0:06:15.840 you literally have nothing to say. If you didn't have 0:06:15.920 --> 0:06:18.760 the grammar and syntax, you wouldn't have a way of 0:06:18.760 --> 0:06:21.440 putting it together that would make sense to someone else. 0:06:21.480 --> 0:06:23.760 It would be a jumble of words with no no 0:06:23.960 --> 0:06:27.200 real way of being sure what the underlying meaning is. Yeah, 0:06:27.279 --> 0:06:30.240 I think things like grammar and syntax give you a 0:06:30.320 --> 0:06:34.279 more precise grasp on meaning. So you could probably communicate 0:06:34.400 --> 0:06:38.159 some meaning with just words but no syntax, But syntax 0:06:38.200 --> 0:06:40.800 helps you put together complex site. We could all turn 0:06:40.839 --> 0:06:44.400 into Yoda and more or less understand what the other 0:06:44.440 --> 0:06:48.599 person is saying. But it gets increasingly difficult to uh 0:06:48.640 --> 0:06:51.280 to express more complex thoughts like you were saying. It 0:06:51.320 --> 0:06:54.520 gets challenging to do that if you're not following a 0:06:55.560 --> 0:06:59.560 essentially a previously agreed upon framework. That's really what language is. 0:06:59.600 --> 0:07:01.680 But how or even though it's previously agreed upon, that 0:07:01.760 --> 0:07:04.560 does not mean it's static. Right, It changes all the time, 0:07:04.600 --> 0:07:06.679 which I think is one of the amazing and beautiful 0:07:06.720 --> 0:07:10.200 things about language that that new vocabulary enters our our 0:07:10.280 --> 0:07:15.080 syntax all the time. Yeah, and the new structures, new 0:07:15.160 --> 0:07:18.560 structures enter enter our our understanding of language all the 0:07:18.600 --> 0:07:23.240 time because internet, right and text messaging. Uh. You know 0:07:23.320 --> 0:07:26.000 it's it's funny because you'll hear a lot of of 0:07:26.440 --> 0:07:31.800 linguistic purists kind of uh dismiss or or look down 0:07:31.920 --> 0:07:35.360 upon any kind of change to the languages. If language 0:07:35.400 --> 0:07:37.600 itself is a sacred thing that should never change. But 0:07:37.640 --> 0:07:42.080 that's if. If it doesn't change, it it stagnates. Well, 0:07:42.120 --> 0:07:45.280 it becomes more difficult to communicate. I can understand the feeling, 0:07:45.320 --> 0:07:48.440 but it's just pure personal stubbornness. I mean, the language 0:07:48.560 --> 0:07:52.240 changed huge amounts before those people got to it in 0:07:52.280 --> 0:07:55.239 the form that they wanted to stick. Sure, we didn't 0:07:55.240 --> 0:07:59.280 even have grammatical rules or spelling rules before the printing 0:07:59.280 --> 0:08:01.640 press was invent to it was all kind of willy nilly, 0:08:01.760 --> 0:08:05.080 and it changes all the time. And the point we're 0:08:05.120 --> 0:08:07.800 getting at is that language is very complex. It's something 0:08:07.880 --> 0:08:10.120 that as you grow up and as you are exposed 0:08:10.160 --> 0:08:13.080 to language and as you understand it better from either 0:08:13.240 --> 0:08:15.840 a practical point of view just because you're using it, 0:08:15.960 --> 0:08:20.320 or from a formally educated, you know, perspective. It's one 0:08:20.360 --> 0:08:23.119 of those things that you get more adept at over time. 0:08:23.520 --> 0:08:26.840 Machines have a very different way of communicating. It doesn't 0:08:26.880 --> 0:08:31.600 have this this kind of level of linguistic complexity. It's 0:08:31.680 --> 0:08:35.400 really when you boil it down to the very basic level, 0:08:36.120 --> 0:08:39.480 a series of zeros and ones. It's binary code. It's 0:08:39.559 --> 0:08:44.520 essentially saying off or on, and these these little zeros 0:08:44.559 --> 0:08:47.240 and ones are instructions more or less. I mean, you 0:08:47.280 --> 0:08:50.400 can have it represent something. Those zeros and ones could 0:08:50.440 --> 0:08:55.560 represent a once translated a a letter in a natural language, 0:08:56.400 --> 0:09:00.200 or it could represent a numeric figure, and then you 0:09:00.240 --> 0:09:03.079 would have other zeros and ones that essentially give instructions 0:09:03.080 --> 0:09:06.760 to a processor on how to push that other information through, 0:09:07.240 --> 0:09:10.199 and then you get an operation performed. Or you can 0:09:10.320 --> 0:09:14.520 use them to represent a photograph or a piece of music. Yeah. Yeah, 0:09:14.559 --> 0:09:17.319 it really doesn't matter what the data is. It just 0:09:17.360 --> 0:09:19.520 means that you have to be able to represent it 0:09:19.600 --> 0:09:23.400 somehow in a quantifiable sense, to reduce it to that 0:09:23.520 --> 0:09:26.640 numerical component. I think one way that's interesting to think 0:09:26.679 --> 0:09:30.120 about the difference between natural language and machine languages. That 0:09:30.640 --> 0:09:35.280 machine language is very good at getting it right. Natural 0:09:35.360 --> 0:09:39.120 language is very good at making it work. If you 0:09:39.120 --> 0:09:42.760 think about that, so machine language. Um, it's going to 0:09:42.800 --> 0:09:46.720 be very precise, but if a single thing is wrong 0:09:46.840 --> 0:09:49.720 in a long string, it's likely to come up the 0:09:49.760 --> 0:09:54.040 whole well, so that you know your entire meaning is lost. 0:09:54.600 --> 0:09:57.680 You can misspeak a sentence and people can still understand 0:09:57.720 --> 0:09:59.720 what you're saying. Luckily for us, because we do that 0:09:59.760 --> 0:10:02.920 all time. Of course, if you're pedantic like I am, 0:10:02.960 --> 0:10:05.800 you cannot help but point out when other people make 0:10:05.840 --> 0:10:08.080 these little misteps and say I think you mean blah 0:10:08.120 --> 0:10:10.640 blah blah, which, by the way, it makes you sound 0:10:10.679 --> 0:10:12.960 just as obnoxious as I indicated, and I do it 0:10:13.000 --> 0:10:15.320 all the time. Well, it makes you as obnoxious as 0:10:15.320 --> 0:10:18.160 a computer because it seems like you're suggesting that you 0:10:18.240 --> 0:10:21.920 have a perfect grasp of meaning, which nobody who speaks 0:10:21.960 --> 0:10:26.559 language does. Language is sort of natural. Language is necessarily approximate, 0:10:27.000 --> 0:10:31.240 and it's sort of adaptive and elastic as you use it. 0:10:31.400 --> 0:10:33.600 I might be as obnoxious as a computer, Joe, but 0:10:33.920 --> 0:10:37.560 I have feelings and words can hurt. Uh No, alright, 0:10:37.600 --> 0:10:41.760 So this brings us to programming language. So programming language 0:10:41.800 --> 0:10:44.440 is what a lot of maybe not a lot of 0:10:44.480 --> 0:10:47.400 some people who aren't really familiar with computers, they know 0:10:47.480 --> 0:10:49.959 about computers. They know in general what computers do, that 0:10:50.040 --> 0:10:51.840 kind of thing. They've used computers, but they haven't ever 0:10:51.920 --> 0:10:54.760 gotten into any kind of programming or anything along those lines. 0:10:54.800 --> 0:10:59.360 Nothing and computer science a programming language to a lay 0:10:59.360 --> 0:11:01.400 person may seem like, oh, this is the stuff that 0:11:01.480 --> 0:11:05.480 computers communicate in, and it's not. A programming language. Is 0:11:05.520 --> 0:11:08.760 really just a set of rules that exist for our 0:11:08.800 --> 0:11:12.920 benefit for humans, benefits for for human computer scientists and 0:11:13.160 --> 0:11:16.800 computer programmers, so that they can follow these rules and 0:11:16.840 --> 0:11:20.120 create a program that a computer can then execute in 0:11:20.280 --> 0:11:24.800 some future operations. Compromise. Yeah, it's it's really just like, 0:11:24.880 --> 0:11:26.600 this is the set of rules you follow, and as 0:11:26.600 --> 0:11:28.560 long as you follow these rules and you're really careful 0:11:28.600 --> 0:11:31.120 and you do them, do the steps properly, then the 0:11:31.120 --> 0:11:33.360 computer is gonna understand what you want it to do 0:11:33.480 --> 0:11:37.199 whenever whatever input comes into that computer. This is all 0:11:37.240 --> 0:11:40.400 dependent on something called a compiler, which is basically a 0:11:40.440 --> 0:11:44.800 translation mechanism from from something that humans can process easily, 0:11:44.880 --> 0:11:48.160 a language that humans can process easily, to the code 0:11:48.160 --> 0:11:50.640 that lets the computer know what it's supposed to do exactly. 0:11:50.679 --> 0:11:54.400 So a compiler is that that intermediary step. It's what 0:11:54.520 --> 0:11:57.480 takes that that code, which to to a lay person 0:11:57.520 --> 0:12:01.200 again doesn't look simple. Right someone who is never coded ever, 0:12:01.360 --> 0:12:03.320 if they were to look at just a page of 0:12:03.480 --> 0:12:06.520 raw code that someone, some programmer has created, to them, 0:12:06.559 --> 0:12:10.640 it looks like gobbledygook. It's just incomprehensible. But or you know, 0:12:10.720 --> 0:12:12.920 you might recognize some words here and there, but there's 0:12:12.960 --> 0:12:14.440 gonna be a lot of stuff where you're like, I 0:12:14.480 --> 0:12:18.080 don't know what that's a string of symbols. Yeah, but 0:12:18.080 --> 0:12:23.160 but that's poetry. It's it's still it's still a format 0:12:23.200 --> 0:12:26.520 that humans find much easier to understand than just a 0:12:26.520 --> 0:12:28.520 bunch of zeros and ones, which if you were to 0:12:28.520 --> 0:12:30.800 look at a full page of zeros and ones, that 0:12:30.840 --> 0:12:33.880 would seem to be meaningless to to most people. I 0:12:33.920 --> 0:12:36.000 would say, there might be some people out there who 0:12:36.000 --> 0:12:38.040 could look at a page of zeros and ones and 0:12:38.360 --> 0:12:40.560 because their brains are wired a different way than mine, 0:12:40.640 --> 0:12:43.280 is it totally makes sense. But for most of us, 0:12:43.440 --> 0:12:46.920 beyond like Neo, it's it's pretty difficult to so in 0:12:46.960 --> 0:12:50.800 that that sense, this compiler changes that programming languages into 0:12:50.880 --> 0:12:53.599 that that object code, those zeros and ones for the 0:12:53.920 --> 0:12:56.800 computer to understand what. Then it goes and runs whatever 0:12:56.840 --> 0:12:59.920 the operations are in the program. This technology, by the way, 0:13:00.160 --> 0:13:02.319 was pioneered in the nineteen fifties by someone that you 0:13:02.400 --> 0:13:05.840 might have heard about, a Rear Admiral Dr. Grace Hopper, 0:13:05.960 --> 0:13:09.000 who was a mathematician and computer scientist who who was 0:13:09.080 --> 0:13:11.319 working for a number of decades, like the forties through 0:13:11.320 --> 0:13:16.040 the nineties, and largely on compilers and computer languages. One 0:13:16.080 --> 0:13:19.160 of her nicknames, in fact, was Grandma Coble because she 0:13:19.160 --> 0:13:23.280 she helped create that particular language, not the planet and 0:13:23.360 --> 0:13:27.600 not the character in Mortal Kombat, but the common business 0:13:27.640 --> 0:13:30.400 oriented language. And and one one little piece of trivia 0:13:30.440 --> 0:13:33.080 before we move on. She's also known for creating the 0:13:33.200 --> 0:13:37.480 term computer bug because, as the story goes, a moth 0:13:37.600 --> 0:13:40.480 flew into a circuit and short circuited the machine, and 0:13:40.559 --> 0:13:43.599 she referred to it as a bug in the system, 0:13:43.679 --> 0:13:47.520 and thus we have computer bug. There you go. Before 0:13:47.559 --> 0:13:51.160 this kind of work was was being done with compilers, 0:13:51.480 --> 0:13:53.920 computers were very different, and she she happened to work 0:13:53.960 --> 0:13:56.040 on one of those as well. She was part of 0:13:56.040 --> 0:13:59.720 the original team working on the Harvard IBM Mark one. Uh. 0:14:00.360 --> 0:14:03.280 The kind of computer that that was was it was 0:14:03.320 --> 0:14:06.200 a punch tape computer, right and even even the later 0:14:06.280 --> 0:14:09.240 punch card computers used compilers. Here's the steps. I mean, 0:14:09.280 --> 0:14:11.840 it's crazy if you think about what what computer programs 0:14:11.840 --> 0:14:14.319 had to go through. If you think that those lines 0:14:14.360 --> 0:14:17.560 of code are difficult, if you think that's laborious, just 0:14:17.760 --> 0:14:20.840 imagine what would happen if you were programming computers in 0:14:20.840 --> 0:14:24.720 the fifties and sixties. So you would get a probably 0:14:24.720 --> 0:14:27.000 you'd start with with some sheets of paper that are 0:14:27.080 --> 0:14:30.600 lined up so that you can actually write your program 0:14:30.680 --> 0:14:33.360 by hand using a pencil, and these sheets of paper 0:14:33.360 --> 0:14:35.480 you might even have to erase stuff over and over again. 0:14:36.080 --> 0:14:39.360 You then would either hand those that stack of paper 0:14:39.440 --> 0:14:41.960 that you've created that has your program on it to 0:14:42.280 --> 0:14:44.720 someone who is an expert using a key punch machine, 0:14:44.800 --> 0:14:47.640 or you do it yourself. But you would then uh 0:14:48.160 --> 0:14:52.680 coordinate with the key punch machine to key in your program. 0:14:52.720 --> 0:14:55.920 It would punch holes into a series of cards. A 0:14:56.040 --> 0:14:59.600 program might be uh, you know, a few dozen cards, 0:14:59.640 --> 0:15:03.360 amount several hundred, it could be several thousand physical punch cards. 0:15:03.640 --> 0:15:07.600 Generally speaking, programs were about as heavy bell as long 0:15:07.920 --> 0:15:10.400 as the number of punch cards that could fit into 0:15:10.440 --> 0:15:13.640 the hopper of a computer. So if you made a 0:15:13.680 --> 0:15:16.160 program that was longer than that, you pretty much gave 0:15:16.240 --> 0:15:18.760 up because it just it was too hard to feed 0:15:18.800 --> 0:15:21.920 it into the computer. Now that you don't just take 0:15:21.960 --> 0:15:24.760 your program straight to the computer. That's actually not the 0:15:24.800 --> 0:15:26.520 way it works. It's kind of similar to what we're 0:15:26.520 --> 0:15:29.960 talking about with that programming language. This is your programming language, 0:15:30.000 --> 0:15:32.960 your punch cards. You would take that to a compiler. 0:15:33.400 --> 0:15:37.280 So the punch cards you had created are your source deck. 0:15:38.000 --> 0:15:41.040 Starts to sound like I'm playing Magic Big Gatherers. You 0:15:41.120 --> 0:15:44.120 then run it through the compiler. The compiler then compiles 0:15:44.280 --> 0:15:47.760 your program and creates an object deck, which is a 0:15:47.840 --> 0:15:50.880 new deck of cards with holes punched in it, but 0:15:51.040 --> 0:15:53.640 it really kind of compresses down the number of cards 0:15:53.640 --> 0:15:57.600 you need that compiled deck. That object deck is your program, 0:15:57.600 --> 0:16:00.240 which you then can take to the appropriate computer and 0:16:00.360 --> 0:16:02.640 run through to see if you did it correctly or 0:16:02.680 --> 0:16:04.720 if the computer gives you an error. Therefore, you can 0:16:04.760 --> 0:16:08.000 write much longer programs than you know. If it all 0:16:08.040 --> 0:16:09.960 has to fit into that one little stack, well, you 0:16:10.000 --> 0:16:12.960 could definitely, yeah, much you would have. You would have to. 0:16:13.040 --> 0:16:15.800 You would have to break it up into uh into sections, 0:16:15.800 --> 0:16:18.840 which might be somewhat challenging, but uh yeah, And I mean, 0:16:19.240 --> 0:16:21.120 and here's the thing. You need to make sure you 0:16:21.200 --> 0:16:24.600 number those punch cards, y'all, because there's nothing like watching 0:16:24.640 --> 0:16:29.480 a computer engineer just weep uncontrollably after someone accidentally spills 0:16:29.600 --> 0:16:32.360 over a deck of cards that were not numbered and 0:16:32.400 --> 0:16:34.880 you have no idea what order they're supposed to go in. Yeah, 0:16:34.920 --> 0:16:38.200 it still happens a lot today, just kidding. Actually it's 0:16:38.200 --> 0:16:41.680 a lot better today. Today we have these programming languages 0:16:41.960 --> 0:16:45.400 like like See and Java, you know, and and it's 0:16:45.400 --> 0:16:48.480 all and we have monitors and we have story you know, 0:16:48.600 --> 0:16:51.920 like before we're talking about you're you're creating these physical 0:16:51.920 --> 0:16:54.440 programs on physical pieces of paper because there was no 0:16:54.560 --> 0:16:57.400 other way of doing it, right, So these programming languages 0:16:57.440 --> 0:17:00.560 we have today and all the stuff like monitors and 0:17:00.560 --> 0:17:03.440 and everything that makes it a lot easier to write 0:17:03.440 --> 0:17:05.600 programs like this, But it's still a lot of work 0:17:05.720 --> 0:17:08.600 on the front end. Is there a way that we 0:17:08.680 --> 0:17:12.200 could make computers as easy to talk to as our 0:17:12.280 --> 0:17:15.080 coworkers so that when we have an idea of what 0:17:15.160 --> 0:17:17.879 we wanted to do, we can just explain it to 0:17:17.920 --> 0:17:20.840 the computer, right, So this this is sort of the 0:17:20.920 --> 0:17:23.600 idea also behind things like the semantic web, where we're 0:17:23.640 --> 0:17:26.800 able to have the web not only understand exactly what 0:17:26.840 --> 0:17:29.920 we want, but begin to anticipate it based upon that. 0:17:29.920 --> 0:17:32.880 That obviously has other elements of artificial intelligence on top 0:17:32.960 --> 0:17:37.240 of natural language processing, but it's natural language processing as 0:17:37.280 --> 0:17:40.679 a as an integral part of that. So this is 0:17:41.200 --> 0:17:43.359 a really difficult problem, and it's a problem to the 0:17:43.359 --> 0:17:47.000 problem on multiple levels. We probably started just the absolute 0:17:47.040 --> 0:17:51.040 simplest thing, like just getting it to understand regular words 0:17:51.200 --> 0:17:53.879 and stuff. I mean, it's you know, the components. You 0:17:53.880 --> 0:17:57.440 can break it down to include things like understanding sentences, 0:17:57.480 --> 0:18:01.480 so actually understanding the structure and meaning of a sentence, 0:18:01.520 --> 0:18:05.960 which is not a trivial task. It's really complicated. Uh. 0:18:06.240 --> 0:18:09.240 There's also the machine translation, which is the whole idea 0:18:09.359 --> 0:18:13.199 of translating one language into another language, which is really 0:18:13.240 --> 0:18:15.119 what we talked about in the video, is that idea 0:18:15.119 --> 0:18:19.159 of real time translation. Again, very difficult to do if 0:18:19.200 --> 0:18:22.760 you want to preserve the intended meaning of the original message. 0:18:23.040 --> 0:18:26.480 There's parsing and tagging, which is all about examining the 0:18:26.520 --> 0:18:30.040 parts of a statement to determine the relationship to one another. 0:18:30.119 --> 0:18:34.000 So being able to identify what part of a sentence 0:18:34.080 --> 0:18:36.680 is what and how they interact. So if I say 0:18:36.720 --> 0:18:39.600 a simple sentence like Joe, go get me a cup 0:18:39.640 --> 0:18:42.600 of coffee, that it's able to figure out all of 0:18:42.640 --> 0:18:44.880 those things, like what what all of those things mean, 0:18:44.960 --> 0:18:48.720 and what what I'm actually what the command actually is well, 0:18:48.800 --> 0:18:51.520 and even that it's a command, I mean, it needs 0:18:51.560 --> 0:18:55.600 to understand that you are giving an imperative sentence, Joe, 0:18:55.680 --> 0:18:59.560 you do this, not a descriptive sentence some subject called 0:18:59.640 --> 0:19:02.520 Joe should got went and got me a cup of corps, 0:19:02.600 --> 0:19:06.160 right right, some something that already happened. It clearly has 0:19:06.200 --> 0:19:09.080 not happened because I sit here with no cup of 0:19:09.080 --> 0:19:10.920 coffee in front of me. Joe, you just don't take 0:19:10.920 --> 0:19:15.040 a hint anyway. There's also a grammar induction. So this 0:19:15.119 --> 0:19:18.199 is the idea of mapping out the hierarchical structure of 0:19:18.200 --> 0:19:22.240 a natural language statement. Um. And this is the way 0:19:22.240 --> 0:19:24.120 I try and explain to other people is it's like 0:19:24.160 --> 0:19:26.960 your diagramming a sentence, where you're identifying what each part 0:19:27.000 --> 0:19:30.040 of the sentence actually is, what it what its purpose 0:19:30.280 --> 0:19:34.119 is within a sentence, and these as we know, like 0:19:34.200 --> 0:19:36.679 these things can change depending upon the structure of the sentence. 0:19:36.680 --> 0:19:39.400 An object can become the subject in the next sentence, right, 0:19:39.720 --> 0:19:42.600 But a computer doesn't know that. So if you were 0:19:42.680 --> 0:19:45.919 to say, take one sentence and you define all the 0:19:46.000 --> 0:19:49.159 different parts of that sentence, for for the the the 0:19:49.240 --> 0:19:51.800 what they're doing within that sentence, and then you create 0:19:51.840 --> 0:19:54.679 a brand new sentence using those same words, the computer 0:19:54.680 --> 0:19:57.080 wouldn't necessarily be able to tell you what was what. 0:19:57.200 --> 0:19:59.719 It may be basing its decisions on what the previous 0:19:59.800 --> 0:20:04.040 s and said, and then everything's wrong and nothing makes sense. Um. 0:20:04.119 --> 0:20:07.840 On top of that, there's the word sense, disambigulation, ambiguation 0:20:07.960 --> 0:20:11.800 rather and it's um, I know, I know, I like 0:20:11.880 --> 0:20:15.680 to add ls into words where they don't belong. Disambiguation, 0:20:15.720 --> 0:20:21.159 I was being ambiguous. Uh yeah, so yeah, these are 0:20:21.280 --> 0:20:25.200 words that can have multiple meetings, sometimes contradictory meetings. If 0:20:25.240 --> 0:20:27.640 I say that something is cool, I could mean one 0:20:27.680 --> 0:20:30.560 of two things, you know, at least one of two things. 0:20:31.000 --> 0:20:34.240 If I'm being really ironic, I could be meaning multiple things. 0:20:34.320 --> 0:20:37.680 So uh, it's it's one of those deals where it's 0:20:38.119 --> 0:20:40.840 again something that we humans can figure out, usually through 0:20:40.840 --> 0:20:45.360 context and tone, but a computer may have real problems 0:20:45.400 --> 0:20:49.200 with that. And none of these are trivial right. All 0:20:49.200 --> 0:20:53.800 of these are are challenging issues and U and it's 0:20:53.840 --> 0:20:58.199 interesting to me how different people have gone about trying 0:20:58.240 --> 0:21:02.800 to address these oallenges. Uh. For example, just trying to 0:21:02.840 --> 0:21:06.159 teach a computer all these pieces, even if you're working 0:21:06.160 --> 0:21:10.040 with a very limited vocabulary. It's exhaustive to try and 0:21:10.240 --> 0:21:13.480 get all the different variations on there of all the 0:21:13.520 --> 0:21:17.000 ways a human could use those uh, those words and 0:21:17.040 --> 0:21:20.359 those commands to mean different things. And just teaching a 0:21:20.359 --> 0:21:24.000 computer that just using brute strength, it requires a lot 0:21:24.040 --> 0:21:26.600 of processing power and it it just takes a lot 0:21:26.640 --> 0:21:29.280 of time. Sure, there's a lot of parallel processing that 0:21:29.359 --> 0:21:31.359 goes on in the human brain and this is this 0:21:31.400 --> 0:21:34.360 is what allows babies and foreign language learners to pick 0:21:34.440 --> 0:21:38.080 up on this kind of contextual contextual stuff that's that's 0:21:38.080 --> 0:21:41.359 going on simultaneously and sorting out all of those different parts. 0:21:41.400 --> 0:21:44.399 But that's much harder to convince computer to do, or 0:21:44.480 --> 0:21:47.159 to allow a computer to right. Yeah, yeah, just programming 0:21:47.560 --> 0:21:50.080 computer to have that same kind of capability is you know, 0:21:50.440 --> 0:21:55.680 a pretty hefty task. So in the sixties, the approach 0:21:55.760 --> 0:21:59.679 was really to create a semantic based understanding to resolve 0:21:59.760 --> 0:22:02.040 lane which ambiguity. This was kind of the brute force 0:22:02.080 --> 0:22:05.679 approach teach a computer everything you can about language so 0:22:05.760 --> 0:22:08.359 it understands all the rules and can follow along and 0:22:08.400 --> 0:22:12.600 be able to interpret commands. But that's pretty tough to do. 0:22:12.760 --> 0:22:15.760 The example I read was was great, talking about using 0:22:15.920 --> 0:22:20.000 using a language uh interpreter where you could type in 0:22:20.080 --> 0:22:23.520 a phrase translated into a different language, and then translated 0:22:23.560 --> 0:22:26.600 back to see how well it did. Um. My dad 0:22:26.720 --> 0:22:30.360 loves to do this every Christmas. He'll take Christmas carols 0:22:30.600 --> 0:22:32.960 and run it through the Google Translator and running through 0:22:32.960 --> 0:22:35.439