Programming Throwdown - Theoretical Artificial Intelligence

Episode Date: May 31, 2013

This show covers Theoretical Artificial Intelligence, a deep dive into what AI is all about. Tools of the show: TypeScript and Ridiciulous Fishing. Books of the show: Dungeon (Paperback: http://amzn.to/11Iares) and Hitchhiker's Guide to the Galaxy (Kindle: http://amzn.to/117nNv3) (Hardcover: http://amzn.to/11IaEyj) ★ Support this podcast on Patreon ★

Hosting provided by Host Tornado. They offer website hosting packages, dedicated servers, and VPS solutions. HostT.net. Programming Throwdown, Episode 27, Artificial Intelligence Theory. Take it away, Jason. Hey, so we're back from vacation. Pretty excited about that. Yeah, we took a break.

Sorry, that was my fault. No, it wasn't your fault. Well, it wasn't really my fault. It was the end of the season. Oh, yes. And now it's the beginning of another. We're starting a new season. We're pretty excited, though, to be back.

It works for TV shows, so why doesn't it work for us yeah i mean they have to go celebrate on their private island that they bought in the bahamas with all their millions as actors and just like we did yeah that's that's exactly what the show earned us our first million dollars and so we took a couple weeks off yeah right wait that's not what happened to you wait you made oh, I owe you your share. I'm sorry. I'm sorry. I spent it all already.

No, I'm Margaritas. We'll just have to do another 26 episodes. Yep. Yeah, pretty exciting to do our 27th episode. And yeah, I have the first news article. It's pretty interesting. I posted this on G+. So some of you who are

following the programming third-round community got sort of a sneak peek on this article. But yeah, I thought it was pretty funny. Basically, an indie game company made a game dev tycoon, which is sort of like one of these sim, you know, there's these like SimCity kind of games where you have some complex adaptive system, So just some sort of system that's running around. And your job is to sort of tease apart these variables that you don't really know. Like, for example, in SimCity, you know, there's, like, crime is created from residential zones. And you put police stations to reduce the crime. But that takes money and space.

And so you have to make these tradeoffs, things like that. And you start to... I think you just sucked all the fun out of SimCity. You start to learn the dynamics. Half of the fun is learning the dynamics of the sim and all that. These guys made a game dev sim where you're like a few guys in a basement.

You make choices about what you want to spend your money on about making a game. Hacking on games and you can choose. Do I want to buy like the ps3 dev kit which costs a hundred thousand dollars but has a lot but neither of us have played this game so this is speculation oh you did oh you played oh never mind jason has played this game yeah um but uh you know or just make a pc game on the web so it's pretty fun um it's kind of a cute theme cute story so they did something pretty fun where they

released a quote-unquote cracked version of their own game as soon as they released the game and put it on pirate bay um what they did was if you pirated the game using you know their cracked version um about halfway through the game people start pirating your in-game games that you're producing. You must go deeper. And you run out of money and you fail the game. And I thought it was pretty fun.

So the pirates keep stealing enough that you can never win. Yeah, exactly. So you start to just do worse and worse and worse. Yeah. So it's like effectively a not-demo demo. Yeah, yeah. It's basically,

it's the illusion that you're getting the full game, but you're really getting you know the first 30 days of game time or whatever but i mean i think that's fair like yeah i mean people this is what people always say right like oh i download this game so i can see if i like it or not if i like it i'll buy it yeah but if you play it halfway through before this happens or essentially you really liked it you should have probably bought it by then yeah i mean i was super pissed but i know i'm just kidding oh no the truth goes out right but uh it's a pretty funny article some people posted on their forums saying hey

hirers keep you know stealing my game can i research drm in that game and stuff like that so oh wait that'd be really funny actually i want to learn how to drm the game and stuff like that. Oh, wait, that would be really funny, actually. I want to learn how to DRM the game that I cracked. I want to learn to DRM in the game, which, oh, God, this is confusing. Totally inception moment there. Yeah, this has happened before. A couple people have done this, like, where I think one person, like, released a version of the game, but in the game you wear a pirate hat

if you download the version. Like, just, like, to be cutesy, like like remind you or whatever actually but but legally like i guess it's weird right because if you can't get in trouble for downloading a cracked version that the publisher put out uh so like i think yeah not that these people i think these people have no intention of suing the the people stealing but i don't think you ever like right you couldn't sue somebody for downloading a version you released. Yeah, I mean, and also there's an element of entrapment too, like if there's a game and it's

new and it's hot, and you put it out on Pirate Bay immediately, maybe some people couldn't even buy it yet, I mean, I don't know exactly when they put it out and when they made the game available in different countries and whatnot, but you might get someone to pirate the game who wouldn't have if you hadn't done that, right? So, yeah, I don't think you can go after them. But actually, there

was one game, I think it was Squaresoft. They made a game for Android. This is pretty cruel. Oh, no, no, it was Earthbound. Remember the game Earthbound for Nintendo? Vaguely. Vaguely. So, if you copied that game, like if you tried to dump the ROM, they had something in the game. I don't remember specifics, but you'd get to the boss after playing for 30 or 40 hours, and the game would crash, like purposely crash right there. That's pretty devastating.

Well, on a similar thing, talking about Pirates, the much-hyped dude Daft Punk album, Random Access Memories. Pretty awesome. I don't't think as of this recording i don't think they have actually even fully released it yet but what they did was on itunes you can actually go on your computer and stream the whole album so there's a couple interesting caveats one is that like it streams all at once so i went and listened to it but like there's no track so you have to listen like you can pause it and come back and you can move around in the song i think but you can't actually like do like next track next track or shuffle or any of that right you just listen front to back the quality seemed okay i wasn't really like listening to it in an environment

where i could tell if it was a really high quality stream or not but it seemed okay um and people were saying like that i guess a version a version, I didn't quite follow, but a version have, a pirated but lower quality version had been, like, put onto the torrent sites right before this. It's like people were confused, like, did they steal it from the streaming site or something? And it turns out now, like, this happened a couple days ago that they released the streaming version. Somebody did find a way to basically anytime iTunes does this, you can go download it from the stream. Basically, they're just streaming a file.

So you can find a link to that file, basically, and then just download it. Or, of course, you could always just record it. Yeah, just capture the audio. But it's interesting because if you're going to do that hassle, like, presumably you would probably download it somewhere else anyways. But it's interesting to try to maybe get more people to buy it they'll do this i mean i from my impression daft punk has more money than they care to what to do with so like i don't like i don't know that they super care like if people steal their album or not which may be different for small people but this is kind of an interesting move to head off you know people trying to release earlier get the scoop

yeah there was a ted talk about a lady who, you know, going back to your point about big versus, like, small artists, she, from day one, made all of her albums free and did everything off donations and merch sales at her concert. And she did remarkably well. So, like, the studies show that if you release all the music for free, more people will come to your concert, and enough of those people will buy t-shirts and bracelets or especially she kind of you know like she'd make it a point in her concert to be like you know hey if you like it you

know pay money for something you know yeah please support us a little bit of psychology there you have to do kind of the right thing there but uh but yeah it worked really well for her yeah i think it's music industry is probably going to break into pieces where different people earn living in different ways or whatever. But I have heard this story over and over again that, yeah, if you kind of just give it away for free and then charge for your concert and for merchandise, you can do pretty well. But it's increasingly difficult to charge for the music itself.

And people are saying with increasing, with increasing, like, this, you know, RDO and Spotify, and then recently Google announced they were going to be doing something similar. It's rumored Apple's going to do something similar as well, these, like, radio things. But they pay ridiculously low. Like, it's like 100 streams of your song, and you'll get three pennies or something.

Yeah, exactly. It's like some ridiculously bad deal. So, like, charging for that is really, really hard. Yep. And it makes sense that the concert, even when people were paying, say, $1 a song on iTunes, that's still nothing compared to $70 a ticket to go to a concert. If you go to a concert, you're probably going to have to bring a friend, right?

Unless you want to go alone. So that's like $140 they're making. But it's not it's not scalable right i mean ideally you would like to be able to sell your album digitally all over the world with basically no cost versus if you do a concert you have to pay to rent the venue you have to go there you can only do it for so that's a good point i saw there's and i forgot the name of it now but there's some artists are doing like concerts online Where like you can pay like five dollars or whatever and like watch them play a couple of songs

I'm like, you know like YouTube like YouTube like as opposed to recording a video or streaming it live though You like people buy to like have you play or whatever. I thought that was kind of interesting, you know, like okay You know, that's another way to do it like a more quote-unquote intimate setting i guess like with only the really diehard fans it's like what udacity is to colleges they're trying to do concerts yes maybe okay i don't know that that point escapes me yep okay but uh yeah another article is pretty interesting. North Carolina has banned. So technically they haven't banned Teslas, but they've effectively banned. Well, Tesla sales. Right, but they've effectively banned Tesla sales. Basically, so apparently to have a car dealership, you have to meet a bunch of criteria. And one of the criteria is you have to be in the state and employed in the state, things like that.

So, you know, there's a lot of, for example, if you go to your neighborhood, you know, Ford or Cadillac, whatever dealership, they have, you know, a general manager for that dealership who lives in that state, etc. And they have a set of certified car salesmen, etc. So Tesla didn't want to do any of that. They wanted to sort of take a page from Apple's textbook and just have a bunch of stores with a couple of cars and then some Tesla geniuses, like the Apple geniuses, but they would really kind of route you through either a phone or an online system

or something like that, and you'd custom order your car, et etc so um they're going that route and uh you know anytime you go against you know the status quo you're going to run up against people with a lot of money and they're for a lot of resistance and so apparently there was uh some lobbyists who gave the senator of North Carolina a bunch of money and he passed a law, pushed through a law saying that, no, if you don't have a certified dealership, you can't sell cars in our state. Yeah, it's a really subtle issue where there are established car dealerships. And I guess the way these things are currently written,

like I think you're alluding to, it's like the people who make cars aren't allowed to sell directly to a consumer. Right. For whatever, you know, probably I'm sure I'm going to trust that was a legitimate reason at one point about why you would want to buy from someone else as opposed to directly from the people who made the car. This whole infrastructure was created. And like I said, I'm sure there are legitimate reasons. I just don't know what they are. I don't know. But Tesla now, right, is trying to skip that.

Like, oh, we don't want middlemen. We just want to sell directly and like help lower costs. In addition to all the other things they're doing with making the electric car and all that. This is one of the ways they're trying to be different. And so, like you said, just have, it's just a showroom, but not actually a dealership kind of thing.

And then people just kind of thing and then people just kind of buy from tesla directly in california right and then you know they just bring the car to them or whatever it's like it's like amazon like you just buy your car in amazon and they just deliver it to your door yeah it just shows up but that you want somewhere to go sit in the car and try it or whatever presumably right but like you said this makes a lot of established people very upset um it's kind of nerve-wracking though that like they're not just trying to hurt tesla but they're also concerned like mayor maybe or maybe not they're concerned about tesla but i think there's also like some other thing where maybe they're concerned about ford saying wow this works really

well for tesla it would work really good for us too. And then Ford starts selling directly in North Carolina. And they bypass the, they can sell cheaper than these other people can because they don't have to hold inventory. You bought your car online or no? No. Okay. Well, I haven't bought my car online. I shopped for it online.

Like I did a lot of shopping around and, you know, emailing people. So ultimately I went somewhere and bought it. It is possible, right, to go online and totally customize the car you want online and then and then it goes to the store i mean like no i think ultimately you have to go through a dealership oh i see like you can't have at least i know of i don't think like you can go on ford's we just keep using ford as an example but yeah go on ford's website like pick out a car and say like you know i want to just go ahead and pay for all of this at this price you know send your check to just go ahead and pay for all of this at this price you know send your check to them or whatever and have it show up to your door okay

or even like have it just show up to somewhere right like i think ultimately you can order it but then i think what happens at least this is my assumption from the little bit of stuff i did you can build the car you want but then ultimately they say here's five dealers in your area which one would you like to get you a quote on what that car will cost? Like how much they'll charge you for it. I see. Okay. Right?

So they kind of have like, here's the listed price, but that price is really high. Like you wouldn't ever really want to pay that price. Oh, I see. So then like they call you and like, oh, we can give you a better deal than that. You know, or maybe just charge you. That just depends. Gotcha. Yeah.

I don't think there's no like but makes sense right like for me i would rather just like pick the car based on reviews and like customize it and then just have somebody drive it on a flatbed to my house right and drop it off one day like so so cars are a little different i think than most things in that you know you you have to sort of negotiate when you're buying a car like i don't negotiate when I'm buying a toaster, right? But yeah, so, but okay, so a house is a huge investment. And also there's limited supply of houses. So there it's a little bit different. But even like take a huge TV.

Let's say you wanna buy a gigantic $6,000 TV. There's some cars that cost 6,000, right? So there's some Kias out there I'm sure that cost, anyway, right so yes there's some theoretically some kios out there i'm sure they cost anyway the point is there's there's things that cost a lot of money that don't involve negotiation right and so i feel like the whole like car negotiating go to the salesman and haggle thing like that can disappear like i don't feel so i feel i i agree i don't like that aspect in fact i you know i almost went there was one dealership we had which just said like look We don't know good like this is our price and they weren't like, you know joking

Like I got the sense like legitimately they had no intention to negotiate with me, but their price is also really really good But then unfortunately as much as I didn't want this to happen, but it ended up better I guess I went and found other people who didn't believe me Until I said like like what this other place was was offering but then they actually did manage to give it to me cheaper and so i was like oh i'd rather give my business to the place that wasn't gonna make me negotiate and was giving me what from all intents and purposes seemed like the best deal but then i did manage to negotiate a better deal somewhere else it's kind of like oh but i do think like i mean cars are one of the most expensive things you would buy

and I also think which increasingly I didn't know this but if you're willing to negotiate you can't actually negotiate on TV purchases you can't do what you can negotiate on TV purchases yeah so like if you're going to buy something that expensive you can well even just like from a low level you could shop around

right like you go to Fry's or Best Buy or oh yeah definitely whatever you could shop around, right? Like you go to Fry's or Best Buy or whatever. You could bring an ad, right? Like, hey, they got it cheaper. Like, can you give me a better deal than this? And then the other thing is cars are typically not just like a standard item. There's a lot of options, right? So if there's like a TV, you could do that, right?

Like if you start to include things like extended warranties and like whether it comes with an installation kit and stuff like you can especially negotiate on those parts yeah actually you just reminded me sorry to interject pro tip if you're getting a new phone you can almost always get a free case so the past two or three phones of paid for I've got free cases just because I just say look this price it's pretty high you know i could probably get it cheaper you know walk out the store come back the next day you get a free case so whether that is like worth it to you or not but yeah to your point you can

negotiate on a lot of things yeah so like i think we maybe it's our generation maybe it's just who we are i don't i have no idea what it is but I personally don't really care for negotiation. Like, I'd rather just be, everybody be up front, which, okay, which never happens. But, like, you know, I don't really like doing that. But I do know that sometimes if you're willing to say something, like this happened to me before. A couple times I've, you know, just mentioned something even. Like, oh, I saw, like, you guys offered this thing or whatever on this other item. Could I get it on this? I'm like, yeah, sure. And they'll give me offer this thing or whatever on this other item. Could I get it on this? And they're like, yeah, sure.

And they'll give me some other thing or a discount or, oh, this coupon expired. Could I go ahead and use it anyways? And in reality, they really shouldn't give it to you because sales are tied to coupons, so they try not to run sales and coupons at the same time and this kind of thing. But if you just ask and then they say okay,

either because they don't know what they're doing or they just really want to make the sale, there's a lot of wiggle room for all all things that was a far off topic i would i really do on the day though where i can just like go online choose what i want get a fair price and like have the car brought to my house yeah that's all it makes and then only have to deal with negotiating and selling my car my previous car yeah i think that we we can get there i mean this is a step in the right direction. Unless it gets banned everywhere. Wait, this is a step in the wrong direction. North Carolina banning it.

Yeah, the article is a step in the wrong direction. But yeah, I mean, eventually, you know, cars will, you know, there's already so much information on the internet, right? That you could go to a dealership. It would be amazing if you could take a car and on the test drive, take it to CarMax and try and sell it. Oh. Yeah. We're dragging this topic out. Have you seen that story from the person who, he said he could beat any chess player

in the world, even a grandmaster, as long as he picks two of them, he can guarantee to beat one of them. And he would just go back and forth playing the move the other one played. Oh, no, I haven't heard this story. Yeah, so let's say there's two grandmasters. Yeah, yeah, yeah. there's two grandmasters. You say, oh yeah, yeah, yeah. You just play them off each other.

As long as one plays white and the other plays black, then yeah, you just play the other move and eventually you beat one of them, right? Yeah, okay. It'd be nice if you could do this with cards. Okay, I didn't, okay, maybe that's a little bit extreme. But I mean, yeah, we'll see. I don't know if we'll ever get there.

It's interesting. And it's also hard because, we're getting off topic, but there's this whole free market aspect, right right but then people want to change what that is like oh tesla is getting you know unfair advantage or whatever we want them to play on the same field people don't want the game to change when they're playing a game and doing good at it right that's true so um i think there's some things it's kind of interesting when you see this and it's like really like this is kind of blatant that like you're just trying to protect your entrenchment and not allowing things to change yeah yeah that's definitely what's going on here i don't know anyways we're that was way off topic

quite a bit so time for tool of the show show show oh i did double there wow reverb tool and i didn't even have to do that in post so all. All right, you're up first. All right, I'm up first. So my tool of the show is TypeScript. So JavaScript. That's where you type out scripts. Why do I always, I'm so compelled every time to make some stupid comment when you say this? It's really annoying.

I'm sorry. It's calligraphy. No. So JavaScript is totally ubiquitous, right? I mean, you know, JavaScript's the only language that works on all major browsers. So if you're going to do something client-side in the browser, you're going to be writing JavaScript. Now JavaScript, so Apple and Google don't quite get along, so neither one wants to support the other one. Like, you know, Apple doesn't support Java, programming Java.

Google doesn't support Objective-C for Android. Windows, I don't even know what you write, and I't support Java, programming Java. Google doesn't support Objective-C for Android. Windows, I don't even know what you write. And I guess C Sharp, who knows. But one thing I do know is that all of them support the JavaScript compiler. So, you know, if you were to try to submit the Java compiler, you know, in C++, like that compiler, as part of your app to the iPhone store, they will just reject it. And this has happened before.

People have tried. They've been rejected over and over. But the JavaScript compiler, totally legit. So JavaScript has taken over mobile for the same reason that it's taken over desktop, which is, you know, it supports all the major phones so with JavaScript becoming so popular people have been trying to sort of you know cut try to get the best of both worlds try to use JavaScript and and get this your universal you know support but then also try to get things like type safety

things like static analysis these things which make life a lot easier if you're doing, say, C++ or Java or something like that. And people have listened to a lot of our episodes know that we talk about this at length. So long story short, TypeScript is a dialect of JavaScript, but it allows you to annotate JavaScript with types. You're allowed to say, hey, this thing should only ever be an integer, and this thing should only ever be a string, et cetera. So then what happens, you run a TypeScript compiler, which takes all of these annotations, and it also supports some cool things like classes and inheritance, things like that. It takes all of your TypeScript files and converts them into JavaScript, and it transpiles them.

And when it does, it doesn't create some horrendous JavaScript that you would expect, like just some complete disaster. But it creates JavaScript that's actually very readable. And along the way, it makes sure that you haven't violated any of your type constraints. So it's pretty cool. I've been using it a little bit.

I haven't made anything with it. But I've been using this Turbulence TypeScript game engine, and I'm really liking it. I'm a big fan. Nice, nice. So is it much different than other things I hear, which are like JavaScript derivative stuff? Yeah, so there's CoffeeScript.

Yeah, CoffeeScript, Dart. There's Dart. So one cool thing about TypeScript is, as I mentioned, kind of an annotation system. So you can take existing JavaScript code and just rename the files to.ts. And then TypeScript will compile them back to JavaScript, which is a no-op in this case. But you can take an existing JavaScript project, and it immediately is a TypeScript project. That was kind of the appeal for me, or actually the appeal for this Turbulence developers when I was reading their blog.

They said, you know, we had a ton of JavaScript code, we were getting all these runtime errors and it's kind of a mess and unmaintainable. And we needed a way to sort of add type safety and other nice things, but we didn't want to rewrite our code from scratch. So, yeah, Dart is pretty awesome. Definitely look into that. Have we talked about Dart? I don't know. I think we did a show on it.

Yeah, we did a show on Dart. Okay. We have terrible memory. Well, it's our 27th show. They say humans can only remember seven things. We'll talk about that in our AI movie. But anyways. Oh, we did. We did talk about that in our AI movie. But anyways. Oh, we did.

We did a whole show on Dart. Man, I feel stupid now. Yes. So there's a bunch of languages. Oh, I remember now, yes. TypeScript is pretty nice because it is JavaScript. At least it can be. So yeah.

Cool. So what is your tool-ish show? That was episode 12 for anybody wanting to play along at home. Oh, yeah. Definitely. Flashback time. You can go listen to us talk about Dart then. No, I remember talking about Dart.

I just was trying to understand the difference because I haven't heard of TypeScript before. Gotcha. To try to relate something new to something I know. Nice. You know, Java is just C++ with garbage collection. That's how I understand Java. Sure.

And batteries. Sorry. That was not a funny joke. In my continued not doing good at sticking to actual things that are useful as tools, I have another game. They're tools for your mind. They're tools for me to spend time. Yeah.

And I think actually worse, I think this one is iOS specific, and that is Ridiculous Fishing. Ridiculous. So Ridiculous F fishing is truly ridiculous um the game starts out as you're basically a guy on a fishing boat but it's not like one of those old games if you ever played one like with the little uh handheld fishing rod you like flick it and you like wind or whatever like i don't remember what they were called oh yeah i remember like bass pro yeah or something yeah yeah yeah so it's not like that though it's it's much more ridiculous

so you you cast your lure into the water but there's no casting mechanism you just drop it and it starts to fall through the water column and there are fish but you're really trying to avoid the fish oh because you want to get down because good fish are like lower so you want to get down like not it's like different types of fish versus like each type only has one size but anyway you like navigate your way down trying to avoid fish and the first fish you ran run into you're gonna hook him and then you start coming up but on the way up any fish you hit you also hook so kind of like catamari would I be like getting bigger and bigger clump of fish on your hook oh man but doesn't grow so like you but anyways it does okay you're

so you're sending now through the water column the opposite of however far down you went and that would be pretty ridiculous because it's got like a crazy vector art style and pixel but it doesn't stop there oh okay as soon as you get to the surface instead of having collected those fish you don't actually get the fish until you shoot them so the fish explode into like a a rain of fish and then you have to take your finger on the touch screen and tap the fish to shoot them and some fish take you have to shoot multiple times some fish are negative points you actually don't want to shoot them or you lose points so you shoot all the fish as they're raining

up and then down uh and then any ones you miss fall into the water and you lose them you know it doesn't count so do you get one cast and then you started like you see cast a different game a different instance of the game yeah yeah basically so I get then so once this thing is done then it's over right like you have gotten some money and then you can buy upgrades so like things that allow you to not have to go back up the first time you have fish kind of like a health or whatever you know or like a better lure that allows you to chainsaw through fish and score points that way or like better guns so like maybe like a orbiting laser cannon that blows fish up or

maybe a shotgun or a machine gun and like depending on your style to the bottom is there a yeah so there's different uh fishing holes you go to, and they have different depths, and there are different things at the bottom. And there's a progression and a story, and an end, and I won't spoil it, but it's a ridiculous story that I don't think I really understand.

But then there eventually you get to an endless mode, and then kind of the, you can try to collect all the fish, you try to, anyways, various things. That sounds awesome. It's a pretty cool game but there's a story here which i think is fairly interesting so the developer of this vlam beer i'm probably saying that wrong also you might have heard of the game super crate box okay they maybe not they uh they made this game as well but they had i what was the name of it it was like

radical fishing which was like a web game which was something i i get i haven't played it but something fairly similar and after they released that game that was like a web game, which was something, I haven't played it, but something fairly similar. And after they released that game, that was like 2011, I think, another company came out with something called Ninja Fishing, which was essentially what this Radical Fishing was, but just like for iOS devices. Oh, someone copied it. So someone copied them,

and they were already starting to make this ridiculous fishing. It was like a blank copy. People knew but like it was still i i what i gather like at least somewhat successful wow so what did they do like they just like went and spent more time like refining their game and it took them you know till now a couple years let it die down that thing to pass and then come out with this you know even better game and kind of like show them you know and supposedly it's doing really really well Wow anyways so I would encourage you to check it out three dollars that's right $2.99 it has has 3,000 ratings that's pretty high at least 3,000 people have bought it yeah yeah I don't know so yeah it's I don't think you can get the exact

number of downloads they've had they don't tell you but I think realistically like what is it like it's like at least one percent of people will rate yeah let's say like one in 25 percent of people rate to be to be conservative then multiply by 20 it's 60 000 downloads it's pretty good no that's not that good i think it's way better than that actually all right so i mean i think they've they've sold lots of it oh wow okay anyways but it's regardless of how many they've sold or not sold the destroyed, it's a ridiculous game and you should try it out. Yeah,

it sounds absolutely ridiculous. Hopefully, they'll make an Android version. Yeah. Anything 2D should be pretty straightforward. Should be straightforward. Yeah.

I mean, that's always an interesting thing, right? Like, things which get made for one and not the other for various reasons and various people

have opinions. See if they wrote it in TypeScript. You should mail them a letter, Jason. I know. You should get a petition should mail them a letter jason i know you should get a petition going buying our tools here you should you should get a petition going all right now it's time for the book of the show book of the show all right my book this show is a graphic novel which i always found find really funny like it's graphic like it's describes a lot of gore in the novel i'm

fine i did it again i'm so sorry i'm fine being really immature and calling it comics like it doesn't bother me i don't feel like i'm less of a person anyways so this is a very long comic hundreds of pages but totally awesome and it's called dungeongeon. It's actually written by a couple of French a French author and illustrator but then Joanne Spahr and Louis Trondheim but then it was translated to English. It's pretty awesome basically the story is you know you've played so many video games where you have to go in the dungeon, slay the dragon, save the princess etc right or you've seen so many video games where you have to go in the dungeon, slay the dragon, save the princess, et cetera, right? Or you see in so many books or stories about it, right? This sort of takes that concept and turns it on its head. So the comic is about you are, the protagonist

is a dungeon master. And actually in the first volume, which is what I linked to, which you can buy through our Amazon web referral, which is pretty awesome, is you play as the son of a dungeon, or you follow the son of a dungeon master. And then the feature volumes, he sort of inherits the family business. The business is, you know, getting warriors to come into your dungeon, letting them slay some slimes and building up their confidence and letting them, you know, go back to the town, get some more weapons, come back. And then at some point, killing them and then taking all their armor and weapons and pawning them off. So that is the business of the dungeon.

That's evil. Yeah, so if you ever played like Dungeon Keeper or any of these games, it's kind of like very paradoxical. And it's a pretty awesome awesome book highly recommend it. They have to go to their board meeting where the board of trustees looks in on the dungeon and sees the graphs and they talk about how they don't have hockey stick performance in their dungeon and what can we do

at one point they have a branding problem because the dungeon is too close to another dungeon so they kidnap a princess so that all these warriors start coming to their dungeon instead. And then the enemy dungeon kidnaps their princess back. So it's just pretty funny, pretty awesome, and I highly recommend it.

It's a great read. That does sound interesting. On the notion of lighthearted and less serious thing, my book is Hitchhiker's Guide to the Galaxy. So this was written by Douglas Adams. I believe the order of this is that originally this was a radio show on the BBC. Right. And then he wrote this book. It makes, it's a very, if you've never heard of it before, read it before, a lot of pop culture geek references have their start in this book. Yeah. If there's a lot of jokes that you don't understand it's probably because you haven't read this book that's possible um but it's really it

it's kind of like a comical take on science fiction with a lot of hilarity but it's a really good read there is a story there uh it has a lot of interesting tangents and things that don't always make sense but uh it's funny and it's it is really classic geek material yeah i mean they they interviewed you know there's been tons of interviews with douglas adams but one in particular that i read um he actually hitchhiked um throughout europe and he had a book called hitchhiker's guide to europe um and it was about sort of how to hitchhike through europe with no money and he went through europe completely penniless for months um and so that inspired him to write this hitchhiker's guide to

the galaxy and so it's a phenomenal read so yeah yeah definitely um if you haven't read it before you should definitely check it out yeah totally and it's been it's been out for a while now 25 years at least yeah yeah so um yeah i mean it's it's a pretty reasonable price i think i i haven't bought it in a long time but uh i think i looked it up and it was it's okay that's right it, it's a pretty reasonable price, I think. I haven't bought it in a long time, but I think I looked it up and it was... It's okay. It's all right. It's well-priced and you can get... You probably can get a used copy if you want.

Yeah, $14. Oh, okay. Yeah, not $5. And there's... It's $5. It was... $14. Yeah, I was going to say, so it was supposed to be a trilogy, but it ended up being five

books. What is that called? Quintilogy? I guess. I guess. he has a funny name for i forget i've seen a quote by him about it but i can't recall what it is um but some classic uh characters from that book yeah so any book were not just this isn't too much of a spoiler because it's super early in the book but but any book which the earth gets destroyed on the first

chapter has to be amazing right i mean mean, that sets the stage for awesome. As long as you make sure to bring your towel, you'll be okay. Yeah, definitely. All right. On to theoretical artificial intelligence. Okay. Maybe artificial intelligence theory. Yeah, definitely.

So we're actually doing a two-parter. Two parts. So the first part is... And best of all, they're both free. What? This isn't like one of those Zynga games where we give you the first part and then we charge you $100. Wait, that's a brilliant idea.

No, because then the first part would have to be really good, and then the second part wouldn't be as good, and then people would be mad and want their money back. We're setting up microtransactions. We're just going to go on tangents until you pay us a dollar. Please insert coin here. Yeah. To continue, that would be 25 cents further, please.

Yeah. Oh, man. So you have on here the Turk. No, first we're going to talk about, we will not cover everything in AI because we don't have enough time. Yeah, definitely. So this, I liken this to, we'll do a tour of artificial intelligence.

Yeah. Jason and I will be your tour guides. Definitely, we'll hold hands, hold your hands. What? Stop. We'll hold your hands and walk you through the world of AI. So we're going to go over some highlights. This is by no means comprehensive,

and we may try to shorten some things, and hopefully that's not too bad. If there's anything we miss and you really want to hear about, feel free to email us. Next episode, we'll read some listener feedback. Yeah. And you guys have been doing that, and so keep doing that.

And if you have something you want to hear, you know, please go ahead and write us. Let us know. We can cover something in more detail. I don't know that we could cover something in less detail because we'll have already covered it but we'll do our best so my favorite AI thing when I think about AI

this is kind of bad, I don't know, I think about the Turk so if you don't know what the Turk is this was in the 18th century I guess it started around 1770 was kind of the exact time and this was an build as a mechanical automata chess master right so this was a robot you could go visit who sat in front of a chess table and would play chess very well keep in mind 17 1770 people were amazed they went on tour

for nearly a hundred years um and people would pay to come see it it would saw it was it was just like a you know what in the world like this machine people were amazed to see this machine that could could beat you could play i remember um going to this is really weird but going to north carolina as like uh probably like a third or fourth grader with my family and going somewhere where they had like various arcade machines and they had a chicken that played tic-tac-toe What so you it was it this is really bad Like I don't people people should not have done this but it lived inside like one of those crane machines You know like the cranes right now, but no crane but no crane day but like in like a glass box and

You would put a quarter in and the chicken would go first So I assume what happened they drop a little feed into this box and the chicken would go over and pack and it was labeled like you know input box or something and and the chicken would pack and then you know the screen would show what what uh position it picked and then you would you would go next and then the chicken would go over and pack some more food and then you would you would just back and forth and the chicken was unbeatable nobody could beat the chicken and and so i guess like maybe it's like early childhood memory of not being able to beat a chicken like being embarrassed that i couldn't beat a chicken at tic-tac-toe like like shaped my life um and so the turk was similar this was this machine and

you know i think there's a lot of drawings of you know like kind of the chest was open you could see all these gears you know these gears wearing and spinning and figuring out how to play chess. And for 100 years, people were amazed at this. And then it was finally destroyed in a fire. And they wrote about the secret of the Turk was that there were chess masters inside of him cleverly hiding themselves

and manipulating the arms of the robot to play chess but you know i think and we can go through like history of like all these like things people thinking about and talking about ai but just like the kind of amazement that the average person has with some machine something that's not human but smarter than them yeah yeah definitely that's pretty awesome that's where amazon mechanical turd came from. Yeah, so, yeah, yeah. Nice, clever name. Pretty awesome. So, yeah, I mean, so one of the hallmarks of early AI was Alan Turing's Turing test. And it basically, in a nutshell, says that if a human can't tell whether they're interacting with a computer or another human, then said computer has passed the Turing test.

And so that's been something that sort of AI people have constantly gone back to. Yeah, it kind of goes to what I'm saying, right? Like that people want to know something smarter than themselves. So if you could make, like, how do you tell when a computer is intelligent? If you can't tell the difference between it and a human, it's at least as intelligent as a human. Yeah, yeah, and we'll definitely cover that because there's a lot of interesting areas that can go there. Also, we talked about last episode, I believe, our book was Foundation Series by Isaac Asimov.

And I think that was my book. And Isaac Asimov also has a very popular kind of thing related to artificial intelligence and robotics in particular where he had the three laws of robotics which are a robot may not injure a human being or through inaction allow a human being to come to harm number two a robot must obey orders given to it by human beings except where such orders conflict with the first law and three a robot must protect its own existence as long as such protection does not conflict with the first or second laws so anyways read all that to say that i think it's interesting that people were fascinated and are still fascinated by these laws to think that you could get to the point where it would even matter that you had these laws like that if something was capable of being worrying about whether or not it obeyed by the laws you would you had kind of arrived right and to think

about you know the laws of robotics dealing with safety and being so integrated with safety shows sort of the inherent fear of AI, right? Even today, people are worried that, you know, some AI will go horribly wrong or AI will take over the world. We don't need people. Yeah, it's a classic Terminator thing, right? Yeah, exactly. Yeah.

Yeah. So people are like fascinated and fearful of you know going back to the turk like that this thing can be smarter than them and that they won't have a good way to stop it right right so uh in the early 50s there's a lot of pioneering work there's the first neural network which we'll cover in more detail uh definitely in the second episode but suffice it to say that there's an artificial neural network which is sort of a mathematical approximation of sort of the electrical patterns that go on in a human brain and the plasticity of that in other words how your brain sort of transforms to to capture information and store it there. And we'll cover that in more detail. There's also Arthur Samuel's famous checkers AI,

where some concepts like mini Macs and planning and things like that were introduced. And I think it's interesting, like the relationship of even today games and artificial intelligence, because I think it's a way for like defining a set of rule, almost like the Turing test, right? Like you're defining a world that has a set of rules, the game, and then you're allowing, you can pit yourself against the opponent, the computer. They're like, you can put somebody on the other end.

And as long as if you're both obeying by the same set of rules, you've essentially created a level playing field. Yeah, yeah, definitely. So playing a computer in checkers, like if it can beat you in checkers, you know, in some ways like it's bested you at checkers, right? Right, like in the domain of checkers, it's easy to pass the Turing test, right?

It's much harder to pass the Turing test when, you know, locomotion, speech, all of these other things are involved, right? I mean, once you've done that, then you've made a- Although interestingly, you and I went to a talk by a person who does board game design and i thought he had an interesting point where he was saying that writing a game for the computer it's very sometimes it's very easy to write an ai but playing an ai is very different from playing a human oh that's true that you can make so passing the turn test of checkers is like you can win at checkers but people don't

get the feeling that they're playing a human. You're playing distinctly different and maybe in a way that a human wouldn't normally play. Like he mentioned over-aggressiveness. Often human opponents will try to play well without being mean, but the computer is just going to do whatever is best in its odds. Another thing too is, say take checkers for example so you learn to play checkers from a human learns to play checkers from either somebody or a book that somebody has written so there's sort of this what's called the quote-unquote eternal

conversation um there's an eternal conversation on checkers so the first people who invented checkers said some things about it and how it captures some part of the world of world dynamics. That's probably what inspired them to make checkers. Probably war or something like that. Then there's people who sort of took that, expounded on it, expounded on it. Some people wrote books on strategies. And so your understanding of checkers is actually derived from the people who taught it to you and then subsequently everybody else in that stream, right? Whereas a computer is just going to learn checkers as a mathematical function, right? So it might be exploring checkers in a completely different

vector than you are. Yeah. So almost like if somebody makes an analogy to you about how to determine what to do, then you're likely to make different decisions in the computer, which doesn't understand how to equate playing checkers to moving in a cornfield. Yeah, exactly. I don't know if that's making sense. That's why game AI is so hard. People think, oh, game AI, you know, I'll do all this planning and stuff like that. And they think all these sophisticated methods and things that the Mars Rover would use, right?

But it doesn't really work. I mean, think about it, like, even just in Mario, if the Goombas bounced every now and then, that'd probably be harder to jump on them, right? But that doesn't make the game more fun, right? Part of what makes Mario fun is the predictability. You hit the turtle shell, it knocks eight Goombas in a row, you get a one-up, you know, it makes your brain happy, right?

Serotonin's released. So making a computer that can sort of facilitate a fun experience is actually really hard a lot harder than making as you mentioned the perfect player right so leading into the 1950s in the quote-unquote golden age of artificial intelligence yeah so so at this point everyone thought oh we have it we figured out neural networks and we know how the brain you know works and how electricity causes you to store information so that's it we figured out neural networks and we know how the brain you know works and how electricity causes you to store information so that's it we got it figured out we're going to have human level intelligence and they thought that for about 15 years uh 20 years or so and

realize it's not true and we'll get into why that is but um but yeah so so after the golden age which is around mid 70s um we have what's called the AI winter. And this is where people started to lose faith in AI. And the goals that AI set were so lofty that even if AI had made tremendous progress, which it didn't, it still would have gone through this period of remission, right? But it went through a terrible winter where basically people thought AI was kind of dead there's not really a point to it then in the 90s you have this sort of Renaissance of AI a lot of people attribute the Renaissance of AI to you know better faster computers Moore's Law things like that and also just more sophisticated methods and just a lot of infrastructure, both on the research and on the computation side.

In 1996 was when Deep Blue supercomputer and program combo beat Garry Kasparov in chess. And I always find this story interesting because, first of all, I didn't even realize it was that recent. Well, relatively recent, I guess. Yeah, totally. And it's kind of like if you look at their checkers world-class checkers in 51 and it took 45 years to play something that's played on the same board you play chess and checkers on the same or you can play on the same board yeah i think it's even the same size yeah

same size same board right and just different pieces different rules or whatever and like look how drastically different and i mean obviously chess you can even tell simply it's much more complex than checkers. But, like, you know, it took all this time to kind of get there. That's interesting to me. It's also interesting, like, this huge controversy over, like, were people helping it? Like, you know, was it a ploy by IBM to increase their stock price? Like, nobody's ever fully convinced that they i mean today i think people are fairly settled that it's at least my understanding it's pretty hard like chess you can sometimes beat a

chess computer like the world-class ones but yeah now it's at the point where the best chess computers beat the best players and it's it's sort of a given right that's going to remain forever yeah so now but back then like in for you know in the intermeeting years there was a time when it wasn't quite clear and even then right like you would think oh that was the first time that a computer had beat a reigning world chess master and so like okay like that's it right like but yeah maybe there's all these caveats and you know special cases and one of the things i remember that was up for contention was they actually put all of kasparov's games that he's ever played into deep blue's um you know repository so deep blue knew kind of kasparov strategies specifically and so the argument was oh if we took someone who wasn't kasparov you know like would you have to train for each person

or is the system general? At the time, I think some of those were valid. I think that if you had another famous chess player play Deep Blue, you would have to train on that person for a while as well. But now they've gotten to the point where they'll beat anybody.

What is artificial intelligence? Alright, here we're getting some of the meat of the podcast. Alright, let's let me get my soapbox out for me to stand up okay so i'll take a seat so first let's talk about intelligence right because this by itself is pretty loaded right so um when you talk about intelligence you're talking almost certainly about human or animal intelligence right and so lack or lack of it yeah if you're watching real world then you know so so a lot so there's there's two main sources of intelligence

one is neural plasticity so this is you know if I tell you hey my phone number is five four five five five five, whatever. That's not really your number. No, I'm getting phone calls already. Then a week later, you know, I ask you what my phone number is and you recite it to me, right? There's intelligence and just memorization. And at a higher level, if you learn the layout of your house and then the next week you unlock the front door and go straight to the living room because you know where it is, right? So there's that kind of intelligence that's learned after you're born. There's also a tremendous amount of innate intelligence. And so this is actually intelligence that is encoded in your DNA.

So your brain is actually constructed with a certain set of structures. You know structures for language, structures for vision, and a lot of these structures have incredible complexity and what I would call intelligence in them. And so you know things like vision, language understanding, these things are innate. They're not learned. It's not like you know you might be blind, I think babies are blind when they're born, but that's not a mental thing. That's just a physiological aspect. But you actually have a tremendous amount of innate knowledge. Just to put it in perspective, giraffes and other quadrupeds, when they're born, they can see, they can actually run immediately on birth.

In fact, giraffes specifically, they fall, I think, three or four feet out of the womb and immediately get up and run. So that's intelligence. I don't think babies are actually born blind, though. It's something slightly different, but it's okay. Continue on. Oh, is it? Well, yeah, that was a jump. But the point is, I know for a fact that babies are not born without the ability to see on a cognitive level, right? So, okay, so that's intelligence. And intelligence is sort of passed on through the generations.

It's encoded in our DNA, and our DNA is constantly undergoing mutations and things like that to add more intelligence and to make it more amenable, right? So this is a very slow process. It's not like your child is going to be born knowing geometry or something like that. It doesn't really work that way. But the fact is that people born now, or animals, let's say, born now are much more intelligent than protosomes born millions of years ago, right? So now let's talk about artificial intelligence. So we want to emulate human intelligence. Like we want computers, as we mentioned in the Turing test, to have conversations with us, to see, to recognize objects, to

do all these things. But we don't have an earth simulator, right? And even if we did have a complete earth simulator that could really, really quickly simulate the entire process on earth and evolution and all these different speciation and all these things, genetic drift, et cetera, it wouldn't probably come out the same, right? Unless we had an exact copy of Earth as a seed, we would come up with some other intelligent hypothetical species, right? So we can't go back and recreate human intelligence. Like, it's just not possible. So artificial intelligence is our way

of saying knowing we can't do that how can we still make something that passes this Turing test how can we make computers that sort of you know live in our world don't really know our history but are still able to relate to you know to the way that we the way that we behave and interact so that's that's kind of my that's my that's my jason gauti's soapbox opinion of uh intelligence and ai so i always find it interesting the discussion about artificial intelligence like you know these kinds of things about there seems to be different camps like some people who deeply believe in needing to understand how the brain works other people saying it doesn't matter we don't we can do something else and still be fine you know or like you alluded to like maybe

it's important to figure out what you think happened and how we got here other people not not so much we don't need to do that but people can still make progress and you know create useful things regardless of which one of these camps they kind of fall into or which which path they're considering. Yeah, definitely. Yeah. I mean, I think you need you definitely need to do both. So there are some concessions you have to make. I mean, the reality is human intelligence is almost certainly not perfect.

Right. I mean, there's almost certainly speak for yourself. Hey, those certainly unless we all had Patrick's brain, there's almost certainly a better brain out there, right? Like some other thing. And of course, there are things that computers are much better at than we are, like arithmetic. We'll never match a computer's ability to do that. But on the other hand, we want computers to pass that Turing test, which is fundamentally a human evaluation. It's a human domain that you want computers to succeed in. So you have to sort of do this balance between,

you know, studying the brain and so, you know, sort of like, I guess, hacking the brain, quote unquote. Because we mentioned in image processing, where you encode the blue, I believe it was blue, right? Yeah, you encode the blue channel more lossy than you encode the blue, I believe it was blue, right? Yeah, you encode the blue channel more lossy than you encode the red channel because humans are, you know, less adept to blue. So that's sort of an example of someone coming at it from the biological side, right? But there's plenty of things on the mathematical side too

which can make a big difference. Yeah, and then in my opinion too, I think there's been somewhat of a shift from like artificial intelligence is all about passing the turing test to like you know and people point out things like search engines as this right like they can do a lot of very useful things right which whether or not that's intelligence it's unknown but it's definitely a very useful thing that people that isn't an outcome of passing the Turing test. Right. Getting an AI sufficient to pass the Turing test

doesn't solve the search engine problem. Exactly. So we get things which, you know, and maybe that's machine learning, maybe it's artificial, I mean, whatever you want to call it. But, you know, kind of the advancements of computer, the future was always making a computer which was equivalent of a human.

Yet we kind of ended up developing this other side branch where it was like, oh, making computers do even things that are more useful than what a human would be kind of ended up developing this other side branch where it was like oh making computers do even things that are more useful than what a human would be able to do yeah definitely yeah it's a really good point so uh yeah along these lines uh this was a hot topic a long time ago before either of us were born but it's still really important and that's common sense knowledge so when you hear this term sort of common sense you think of things like if I was to ask a child what's taller a skyscraper or your house or you know things like that they would just they have this a ton of common sense reasoning right and that's because human

beings absorb an incredible amount of data I mean in the image processing episode last episode, we talked about how much data is in video, right? Uncompressed video. And it was enormous. It was like a dictionary every 10th of a second or something like that. You can fit the entire dictionary. So that's just from your eye, right? You have actually a lot of sensation in your ears, your nose, touch, especially your fingertips. And all of those data sources are extremely rich, and there's a ton of data, way more data than any computer can come close to processing even today. So the fact that humans are able to digest and process so much information is another

aspect of AI that we're starting to make progress on. Especially now, there's that open crawl, which has the entire internet, 50 terabytes of the internet in it or something. So with things like search engines, databases, and things like that, we're able to create a small sample of the knowledge that we experience every day. And so we're starting to make progress there. Yeah, I don't know. It's fascinating having a child and watch them, like, what they pick up and learn

and the process of how they grow. But even, like, my daughter will see us do something in the kitchen or whatever, like, you know, do something with the pots and pans. And she has, like, a little play set of pots and pans. But they don't look like our pots and pans, other than roughly resembling a similar shape um she has no ability like we've never really sat there and told her like this is a pot this is also a pot you know like we never went through labeling and said like these things are so but she'll watch us do something is you will see her observe her at her play kitchen washing her hands under a fake faucet

that looks nothing like our faucet and doesn't emit water right and so she recognizes that this thing that i have here is the same as that thing over there and she's able to kind of she doesn't know necessarily why she's doing it but she wants to do the things that we do and she's able to figure that out all on her own and it's like wow to make a computer do that that'd be like really hard like i'm gonna do that no like amazing yeah absolutely amazing and i mean she's only able to do that because think of how much information she's digested. I mean, it's way more than that crawl, open crawl database, you know, 100 times over, right?

Maybe a thousand times over. So yeah, it's amazing. And that's one problem that, you know, has to be cracked still. So we'll just sort of wrap up what is AI. There's some general goals for AI, which I feel are pretty universal. So one is deducing.

So in other words, have you heard of the canonical monkey-banana problem? No. So the idea is there's a banana hanging from the ceiling and there's a chair against the wall and uh and if you just leave a monkey in a room long enough the monkey will start to swing at the banana but he can't reach because it's too far out of his reach um and then he'll start to understand like to take the chair move it in the middle of the room climb on the chair and then get the banana so the idea is the monkey have to sort of envision the future where he's standing on the chair and then get the banana so the idea is the monkey have to sort of envision

the future where he's standing on the chair do some planning etc so that's deducing there's also reasoning so just um reasoning involves a lot of sort of looking at the past and using that to determine the future right so i know i touched a stove it it hurt me i got burned okay i'm not going to touch those right and there's alsosolving which is gets an expert systems and things like that and we'll definitely talk about all those in more detail in the next episode so what are some of the challenge like we talked about maybe some of what we thought AI would be and how we didn't get there so I mean there's obviously some things that got in the way yeah yeah definitely so yeah one thing that got in the way is

this concept of strong AI so people people wanted, you know, people had sort of reverse engineered the brain at a very low level. They hadn't done anything with DNA, which is, you know, as we talked about, you know, a huge folly, right? But they said, oh, if I have two neurons, one, you know, signals the other one like this. And so our brain has, I think it's like a trillion neurons. So if I just do this a trillion times, I'll have a human brain. And that doesn't work. Now I'm sure everyone listening to the show knows why. But their idea of recreating a human really kind of set AI back.

Another problem is whenever you're doing planning or reasoning, you have what's called the combinatorial explosion so let's just take checkers for example so on average in checkers there's I think like 2.5 moves you can make that are reasonable on average so if you're looking at a certain board there's let's say two moves you can make so you're now you're looking at two boards but in each of those boards there's another two moves you can do it so now you know it just explodes right there's also the knowledge representation so this is an incredibly difficult right how is knowledge actually stored one

common thing that you hear of is humans can only remember seven things at a time that's why most phone digits most phone numbers are seven digits, right? But we do this thing called chunking, where you might remember seven things, but each of those seven things is incredibly complex. And so you've sort of like collapsed a bunch of information into one quote-unquote thing that you have to remember, right? And that's what video games are all about. Like in mario it's kind of all about the context like if you're jumping over a pit you know that you have like no room for error but if you're jumping to hit you know a block to get a

mushroom or something you're much more relaxed like you might try and do that and hit a goomba at the same time or something like that so So in video games, you kind of chunk different actions based on their context. So finally, the biggest challenge in AI is what's called the frame problem. And the frame problem is basically you don't know what affects a certain action. So, for example, let's say you're getting out of your car, and you want to know if now is a good time to get out of your car. So clearly, if the car is moving 50 miles an hour, that's probably not a good time to step out of your car, right? Unless you're being held hostage.

Yeah, yeah. Then you could do like a barrel roll or something. But probably not at 50. No, yeah, maybe not. Maybe slow slow to 30 and then barrel roll okay but the the point is like you know the the current temperature in the current temperature in france for everyone who's not in france right now does not matter whether you know if the car is going 50 miles an hour if it's not matters a lot more than the temperature in france for as to whether you should get out of your car or not. But how do you know what should matter and what shouldn't when you're making decisions? That's the crux of the frame problem.

So if you don't know what matters and what doesn't, then you have to consider every possible thing that could exist. And also, the next time you make that decision, you don't really know what made that decision a good or bad idea the first time, right? Because how do you know that the weather in France isn't the reason why last time you got out of your car and you were hit by a bike, right? So knowing sort of what the context is for a particular decision is the frame problem. And that's actually extremely difficult. I guess that's a good thing about AIs for games because it already frames the problem for you. Yeah, that's why games and AI are sort of intertwined. Okay. So we're kind of like giving a little tour, like background, speaking at a very high level.

Hopefully this is interesting to you. We're going to get to some specifics and some more hardcore stuff in the next episode. But kind of wrap up here, we want to talk a little bit about some of the cool things AI is used for. We talked a lot about games, but there are some other things that are kind of classic examples of what AI gives us or attempts to give us. Yeah, you want to go first? Okay, so first one is Maxima, the computer algebra expert system. So it's something that's able to solve algebra.

Yeah, totally. Okay, that was your point. I thought you were going to go straight to Eliza. Oh, well, I'll talk about Eliza, and then you can talk about Maxima. So mine, of course, sticking with useless things is eliza chatbot so this was you know kind of like the early you may have seen it before i remember seeing it at like a science center when i was a little kid yeah they actually have it um i was at the uh moma that san francisco moma has an eliza

oh okay yeah so it's like a computer it says like i forget what they're greeting like hello what's your name or something you type you know my name is patrick hi patrick how are you doing today i'm doing good i'm glad to hear you're doing good you know or something and then it kind of so it looks for like certain keywords and tries to create a sentence or alter what you said in a certain way to make it get back often get stuck and just kind of says like i don't know what that is or something you know but um know, it kind of gives you this illusion almost for just a second that, whoa, wait, what's happening? Am I really talking to somebody else? Yeah, it takes some of what you say and sort of puts it in its next statement.

Like you might say, I don't like diapers. And it will say, oh, tell me more about diapers. And so you feel like, oh, wow, he's really answering me. Yeah. Did I tell you I actually, and my wife was really embarrassed, but I actually, I know of a way to trick these chatbots because, so if you don't know, I got a PhD in AI and machine learning. So done this stuff for a long time. And I kind of, so I, so here's a pro tip. If you've ever seen Eliza bot at a museum or something, you can say something to the

effect of like what I said I said oh I really don't like flies and Eliza said oh it tell me more about flies I said well I also don't like bananas and she said something I remember what she does like done but then this is the catch you say no fruit flies like a banana and so you think about it there's two ways to take that like no fruit flies like a banana or no fruit flies like a banana and it flipped out and it said this is X so I've done this before and it just said like I don't understand or does not computers on the right but this time it literally said I I joke you not. Did you just wet yourself?

Is what it said. Literally. And, like, everybody around is, like, just busting out. It said something like, I don't understand. Did you wet yourself? Well, I was so embarrassed. It was pretty epic. I kind of want to go try this now.

Yeah. You should definitely go to the moment. Give us a shout. Okay. All right. So, yeah. So, I'll cover Maximo. Yeah, since I definitely go to the moment and give us a shot. Okay, all right. So, yeah, so we'll cover Maxima. Yeah, since I butchered it by just reading what you wrote on the page. So Maxima is an algebra expert system.

We'll cover expert systems more later, but you can think of it as sort of like searching, but instead of searching through moves and checkers, it's searching through things that you can do in algebra, like dividing or canceling this number out or you know removing an X from both sides etc. And so it does really complex integrals and differential equations and things like that. Like if any of our listeners are in high school or university and they're taking I think it's what Calc 2 or something like that where you study integrals and you

have to memorize all these crazy rules like the trig rule, like the derivative of sine is cosine and derivative of sine squared is 1 minus cosine squared. I mean, all these crazy rules. So this system, Maxima, knows about all these crazy rules and then it will try different rules and get in and hone in on the solution. So you can say, give me the integral of this huge equation, and it will sort of crunch the numbers and figure it out for you.

Another useful project which used AI was the DARPA Grand Challenge. So this was, you know, create a self-driving car that'll drive across the desert and then through a mock city. And this was kind of, you know, first year was like a horrible disaster. They did it again was kind of you know first year was like a horrible disaster they did it again and like you know it was much much better and you know even just like the you know five years or whatever i guess between the how many years was between the first

one and the final one that they did you know like there's just like a huge progression and like you know how good they could do and you know just the complexity of not only just steering the vehicle the way you want it to go, but recognizing where you want to go and, you know, what to avoid and, you know, all these things. And it's just, you know, it takes a lot of planning, intelligence, getting to, like you talked about, the strong guy being able to kind of like solve all the problems, you know, and trying to put them together in something. Not that a self-driving car solves all of the problems, but the problems, but it has to solve many different kind of aspects of AI in order to be able to get from point A to point B.

Because it's not just like move from point A to point B. It's like, well, how do I move? What's the best way to get there? How do I avoid damaging myself or damaging others or all these things? Yeah, they did the Grand Challenge, which was out in the desert. And then recently they did the grand challenge which was out in the desert and then recently they did an urban grand challenge which i found out later wasn't really

urban like it was a fake urban setting so no one was actually at risk well there were people driving the other cars yeah but they were yeah they were like yeah yeah yeah still pretty cool so um yeah we'll do one more okay so this is pretty cool. It's RoboCup Soccer. And basically the goal is, I don't remember when this started. Sometime while I was in university. But basically when it started, the goal was to have a fully autonomous humanoid robot soccer team compete in the World Cup in 50 years. So by 2047, to have like full-sized human robots competing against, you know, like Germany for the World Cup finals. So pretty ambitious goal. They have an incredible amount of backing from like ACM and other organizations

a lot of interesting AI another student of mine or another student that was with me in the lab at university was working on this problem and it's just pretty amazing the amount of progress they've done they actually have they sort of split it up into several domains they have the full-sized humanoid problem. And at that point, they're kind of like doing penalty kicks. And they haven't really made too much progress there. But if you think about it, that's very expensive. So that's kind of like the last step.

But now they have sort of a smaller scale competitions with real robots that run around and kick soccer balls and stuff. It's pretty awesome. The Wikipedia link should point to some pretty epic videos of soccer matches. In the show notes, yep. Yeah, if not, we'll add some videos. But it's totally awesome, and I highly recommend checking that out. So that's interesting, though.

We just talked about AI in-game, but here's actually getting a harder problem as opposed to just a— and I think they even do have a simulation, simulation league or whatever. But now you're trying to play, you know, it's, it's simultaneous as opposed to taking turns. So if you take longer to figure out what you're going to do,

you know, like that's bad. Other people run around too. Yeah. You know, and then also like having to, again, getting back to like what the grand challenge thing,

like dealing with it in the real world, not only get to figure out what, what you want to do, but how you're going to do it yeah like how do you actually do this yeah this interesting thing do you think they'll make it um so 2047 2013 in 30 years i think that they i think that they'll have humanoids playing soccer i don't know if they'll compete in the world cup one thing i wasn't sure about is, if a robot slide tackles you, does it just matter to you, right? I mean, how does that work?

But obviously, I mean, I don't think the rules probably permit them to play. They probably don't meet the anti-doping rules. No, something, right? Like, they probably don't meet the rules. So you'd have to create a scenario in order to allow them to play, right? Like, you're not allowed to make contact i don't know

that's not really fair that's the thing is like i don't really know how they would do deal with that part of it you know so i feel like that's where the actual limitation is going to be like the ai will be sufficiently capable but you know just the mechanics of taking the stripping the ball from another player and not destroying them is gonna be the hard part Yeah, you have to you have to have the robots implement as mobs three laws. Yeah Then the players can't do anything Because the humans will just tackle the robots Okay, well I think that wraps it up for for this episode so next episode we're gonna be talking about

You know in depth kind of how do you actually get these AI problems done? Like what are the tools and techniques you're going to use? Yeah. By the end of the next episode, you'll be able to actually solve AI problems on your own. Or at least know what to go look up to work. Yeah. Using other people's source code. Yes. You'll know the google search that will give you the answer yeah um all right well i think that's it for us for now cool have a good one guys

the intro music is axo by biner pilot programming throwdown is distributed under a creative commons attribution share alike 2.0 license you're free free to share, copy, distribute, transmit the work, to remix, adapt the work, but you must provide attribution to Patrick and I and share alike in kind.