Latest / Stuff You Should Know / A List Of Games You Would Surely Lose to a Computer

Transcript

A List Of Games You Would Surely Lose to a Computer · Stuff You Should Know

0:01
Speaker 1: Welcome to stuff you should know from how Stuff Works
0:04
Speaker 1: dot com. Hey, and welcome to the podcast. I'm Josh Clark.
0:14
Speaker 1: There's Charles w Chuck Bryant, there's Jerry over there. I'm
0:18
Speaker 1: just gonna come out and tell everybody making fun of
0:20
Speaker 1: me for some weird reason, vaguely weird ways. But I'm
0:26
Speaker 1: all right. So, Chuck, I have a story for you.
0:29
Speaker 1: I'm gonna take us back to the seventeen seventies and
0:33
Speaker 1: the swinging town of Vienna, not Virginia, not Vanna Georgia,
0:40
Speaker 1: but you know that's how they pronounce it, right, Vanna
0:43
Speaker 1: Yanna sausages, right, Vienna, Austria. You ever been there? Vienna, Austria. No,
0:49
Speaker 1: I've been to Brussels. That was pretty close. Vienna's lovely,
0:53
Speaker 1: I'm sure. I think it's a lot like Brussels. Very clean,
0:57
Speaker 1: lovely town. I just remember to be very clean, Yeah,
1:01
Speaker 1: very clean, gorgeous architecture, weird little angled side streets. They're
1:06
Speaker 1: very narrow, very pretty tone. So we're in Vienna and
1:10
Speaker 1: there is a dude skulking about going to the royal
1:15
Speaker 1: Palace in Vienna. His name is Wolfgang von Kimberlin, and
1:21
Speaker 1: he's an inventor. He's an engineer. He's a pretty sharp dude,
1:25
Speaker 1: and he's got with him what would come to be
1:27
Speaker 1: known as the Turk, but he called it the mechanical
1:31
Speaker 1: Turk or the automaton chess player, and that's what it was.
1:36
Speaker 1: It was a It was a wooden figure that moved mechanically,
1:41
Speaker 1: seated at a cabinet and on top of the cabinet
1:44
Speaker 1: was a chessboard. And when he brought it out to
1:47
Speaker 1: show to the Royal Court, he would um, it was cool,
1:53
Speaker 1: kind of but nothing they hadn't seen before, because automata
1:57
Speaker 1: was kind of a hip thing by that. Yeah, people
2:01
Speaker 1: loved building these engineering these automatum machines to do various things.
2:08
Speaker 1: And people are just knocked out by the fact that,
2:11
Speaker 1: you know, you hide these gears and levers behind wood
2:15
Speaker 1: or a cloth and it looks as though there's a
2:18
Speaker 1: real well I'm not real, but you know what I
2:19
Speaker 1: mean is that it's like a real machine. Yeah, but not.
2:23
Speaker 1: They weren't fooled anything like is that a real man?
2:26
Speaker 1: It was, but it was for their time. It was
2:29
Speaker 1: so advanced looking that it's like us seeing ex machina
2:33
Speaker 1: in the movie theater that makes sense. Yeah, no, it
2:37
Speaker 1: does make sense. But imagine seeing like x Makina and
2:39
Speaker 1: being like I've seen this before. This isn't anything special, okay. Yeah,
2:44
Speaker 1: and this thing, to be clear, look like a is
2:47
Speaker 1: it Zoltar or Zoltan from Big Zoltar Sultan? I don't know,
2:53
Speaker 1: it's one of those two, one of those two. Like this,
2:55
Speaker 1: this guy's wearing a turban and it's in a glass
2:58
Speaker 1: case like bust like you know, like a chest up thing. Yeah,
3:03
Speaker 1: he's seated at this this cabinet, so there's no need
3:06
Speaker 1: for legs or anything like that. Yeah. But the thing,
3:08
Speaker 1: this is what was amazing about the Turk. He could
3:12
Speaker 1: play chess, and he could play chess really well. So yeah,
3:16
Speaker 1: he was like an automaton and he moved all herky
3:19
Speaker 1: jerky or whatever. But he could play you in chess,
3:23
Speaker 1: which was a huge, huge advance at the time. Like,
3:29
Speaker 1: this is something that wouldn't come up again until the nineties,
3:33
Speaker 1: more than two hundred years later. This thing, this automaton
3:37
Speaker 1: could play a human being in chess and beat them. Well, yeah,
3:40
Speaker 1: and it looked like when the game started, it would
3:42
Speaker 1: look down at the chessboard and like cock his head
3:45
Speaker 1: like m what should my first move be? And if
3:48
Speaker 1: people I love this part, if people tried to cheat.
3:51
Speaker 1: Apparently Napoleon tried to cheat this thing because this guy
3:55
Speaker 1: he debuted at the Being's court, but then it, you know,
3:58
Speaker 1: it went on a world tour. Yeah, and he was
4:01
Speaker 1: even it was taken over by his successor to the
4:03
Speaker 1: guy who toured with it. Even further, people went nuts
4:06
Speaker 1: for this stuff. They did. They loved it because they
4:08
Speaker 1: were like, this is crazy. I can't believe what I'm seeing.
4:10
Speaker 1: Most people, though, were not taken in by it. There
4:13
Speaker 1: like there's some trick here. But von kempel In and
4:16
Speaker 1: the guy who came after him, I don't remember his name. Um,
4:19
Speaker 1: they would they would demonstrate. You could open this cabinet
4:22
Speaker 1: and you could see all the workings of the mechanical
4:25
Speaker 1: turk inside. Right. So what I was saying is, if
4:29
Speaker 1: this thing's since a cheater like Napoleon supposedly did it,
4:34
Speaker 1: would you know Napoleon would move a piece out of
4:37
Speaker 1: turner illegally or something. This dude, the turk Turk one
4:40
Speaker 1: two would pick up the chess piece move it back
4:43
Speaker 1: as if to say like, no, no, Napoleon, I see
4:45
Speaker 1: what you're doing. And then if the person attempted to
4:47
Speaker 1: move it again, I don't know how many times on
4:49
Speaker 1: you two or three times, eventually it would just go
4:53
Speaker 1: ah and wipe his hand across the board and knock
4:56
Speaker 1: off all the pieces. Game over, which is pretty great,
5:00
Speaker 1: A nice little feature, yeah it is. But it even
5:03
Speaker 1: showed even more that this thing was thinking for itself.
5:07
Speaker 1: That's the key here, right. Chess had been for a
5:10
Speaker 1: very long time viewed as only something that a human
5:16
Speaker 1: would be capable of because it took a human intellect.
5:19
Speaker 1: And there was actually a guy in English um engineer,
5:23
Speaker 1: I think he was a mechanical engineer. His name was
5:27
Speaker 1: Robert Willis. He said that um chess was in quote
5:30
Speaker 1: the province of intellect alone. So the idea that there
5:34
Speaker 1: was this automaton playing chess blew people away. But again
5:38
Speaker 1: people figured out, like, okay, there's something going on here.
5:41
Speaker 1: We think that um von Kempelin is controlling this thing
5:44
Speaker 1: remotely somehow, maybe using magnets or whatever. Other people hit
5:48
Speaker 1: upon the idea that there was a small person inside
5:52
Speaker 1: the cabinet who would hide when the cab when the
5:55
Speaker 1: workings were shown, when the cabinet was open to show
5:57
Speaker 1: the workings, and then when the cabin was closed again
6:00
Speaker 1: and the mechanical turk started playing, the person had crawled
6:03
Speaker 1: back out and it was actually controlling it. This seems
6:06
Speaker 1: to be the case that there was a person controlling it,
6:09
Speaker 1: but the idea that it was it was a machine
6:14
Speaker 1: that could think and beat humans in chess, had like
6:17
Speaker 1: kind of unsettling implications. Yeah, this author Philip Thickness, great name,
6:25
Speaker 1: British author for sure, Philip Thickness. He uh, he said,
6:30
Speaker 1: and you know, people like you said all those more
6:32
Speaker 1: complicated explanations in this article you sent. Astuteley points out
6:37
Speaker 1: that he followed Occam's razor and basically said, he's got
6:40
Speaker 1: a little kid in there. He's got a little a
6:42
Speaker 1: little Bobby Fisher in there that's really good at chess,
6:46
Speaker 1: and that's what's going on. And other people speculated that other,
6:50
Speaker 1: you know, little people might be in there, Um, just
6:52
Speaker 1: adults who would fit in there. But then you know,
6:55
Speaker 1: there's the explanation that he would open it up and
6:57
Speaker 1: shine a candle around and say, you know nothing, just
7:00
Speaker 1: see here everyone. Um, so what should we reveal the
7:07
Speaker 1: real deal? Sure? I think I did already. Well, I
7:11
Speaker 1: don't think you spelled it out is we'll spell it out.
7:14
Speaker 1: There was a little person in there, Yeah, not just
7:16
Speaker 1: one little person, but they would travel around and recruit people.
7:20
Speaker 1: I guess people would get tired of being in there,
7:22
Speaker 1: or they forget about them and they'd starve and have
7:24
Speaker 1: to replace them. But it really was a trick. There
7:27
Speaker 1: was a little person in there. They did the same
7:29
Speaker 1: thing as like the magic acts. You know, when they
7:31
Speaker 1: saw a person in half. It's that the lady just
7:35
Speaker 1: gets into a tiny little ball in one section of
7:37
Speaker 1: that box. But my thing is this, like, this is
7:40
Speaker 1: not a satisfying explanation to me, Chuck, I think it's great.
7:43
Speaker 1: How did the person keep up with the board above? Well?
7:48
Speaker 1: I mean some I don't know if they ever proved
7:50
Speaker 1: exactly how it was going. That's what I'm saying, Okay,
7:53
Speaker 1: whether or not I think they that the zoltar or
7:57
Speaker 1: I'm sorry, the turk was just hollowed out and you
8:01
Speaker 1: you would just put your arms through the arm so
8:05
Speaker 1: you would call up into the church. Yeah, you would
8:07
Speaker 1: become the churk. When the churk would fuse, That's what
8:10
Speaker 1: some people thought. I think that's what Edgar Allen Poe
8:13
Speaker 1: thought too. He would Other people thought that there were
8:18
Speaker 1: the the the little person was underneath in the cabinet
8:22
Speaker 1: operating the turk with levers and stuff like that. Whether
8:24
Speaker 1: it could have been a mirror or something, you know,
8:26
Speaker 1: I guess a little telescopic mirror. That's what's getting me
8:29
Speaker 1: is how would they keep up with the game. You
8:32
Speaker 1: could keep track of the game, but how could you
8:34
Speaker 1: see where the other person moved? You would know where
8:36
Speaker 1: you moved, but you wouldn't be able to see where
8:38
Speaker 1: the other person. That's what I don't get. Just mirrors
8:40
Speaker 1: smoking mirrors maybe so. But the point is is it
8:43
Speaker 1: was a fake. It was a fraud, but it raised
8:45
Speaker 1: some really big questions about the idea of a machine
8:50
Speaker 1: beating a person at something like chess. Yeah, and it
8:53
Speaker 1: really peaked the mind of one Charles Babbage. He was
8:59
Speaker 1: he was a kid or or young at least at
9:00
Speaker 1: the time when he saw the turk in person, and
9:04
Speaker 1: a few years afterward he began work on something called
9:06
Speaker 1: the Difference Engine, which was a machine that he designed
9:11
Speaker 1: too to calculate mathematics automatically. So at some point to
9:16
Speaker 1: this is kind of maybe the beginnings of humans trying
9:20
Speaker 1: to create a I well, yeah, with Babbage's differential machine
9:25
Speaker 1: or difference machine, yeah, difference engine. But at the very least,
9:28
Speaker 1: what this is is the first that I know of
9:31
Speaker 1: example of man versus machine, even though it was really
9:35
Speaker 1: man versus man because it was a man in the machine.
9:39
Speaker 1: It was a fraud, yeah, but it was it. It
9:41
Speaker 1: sparked that idea, It definitely did. And that's something that UM,
9:45
Speaker 1: like chess in particular, has always been like this idea
9:49
Speaker 1: of like, if you can teach a machine to play chess,
9:52
Speaker 1: you really achieved a milestone. And there's been you know,
9:55
Speaker 1: plenty of programs, was most notably Deep Blue, which we'll
10:00
Speaker 1: talk about, but there's there's been this idea that like
10:04
Speaker 1: part of AI is chess, teaching it to play chess.
10:09
Speaker 1: But they, the people who develop a I never set
10:13
Speaker 1: out to make a chess playing AI, just to make
10:16
Speaker 1: a machine that can play chess. That's not the point.
10:19
Speaker 1: Chess has always been this way to demonstrate the progress
10:23
Speaker 1: of artificial intelligence because it's a complex game that you
10:27
Speaker 1: can't just program it, like it almost has to learn. Uh, well,
10:32
Speaker 1: it depends on how you come at it at first, right,
10:34
Speaker 1: So initially they did try to program it Okay, there's
10:39
Speaker 1: this for for from basically nineteen fifty two the about
10:44
Speaker 1: the mid like about say sixty years right. That is
10:49
Speaker 1: how they approached AI and chess is you figured out
10:53
Speaker 1: how to break chess down and explain it to a computer. Now,
10:59
Speaker 1: what if you could, ideally you would have this this
11:02
Speaker 1: computer or this AI, this artificial intelligence UM be able
11:07
Speaker 1: to think about the outcome of every possible every possible
11:12
Speaker 1: outcome of a move before making it. That's just not possible.
11:15
Speaker 1: It's still today we don't have computers that can do that, right,
11:19
Speaker 1: So what you have to do is figure out how
11:20
Speaker 1: to create shortcuts for the machine, give it best practices,
11:24
Speaker 1: that kind of thing. And that was actually laid out
11:27
Speaker 1: in by a guy named Claude Shannon, who is the
11:30
Speaker 1: father of information theory. And he wrote a paper with
11:34
Speaker 1: a pretty on the nose title called programming a Computer
11:37
Speaker 1: for Playing Chess. And you have to say it like
11:40
Speaker 1: that when you say the name, it's got a question
11:42
Speaker 1: market in the end, right. But he laid out two
11:44
Speaker 1: big things, um. One is creating a function of the
11:50
Speaker 1: different moves, and then another one is called a mini max.
11:56
Speaker 1: And if those were the two things that Shannon laid
11:58
Speaker 1: out and they established about fifty or sixty years of
12:02
Speaker 1: development in teaching an AI to play chess. Yeah, so
12:06
Speaker 1: this evaluation function is just sort of the base, the
12:10
Speaker 1: very basis of it all kind of where it starts,
12:12
Speaker 1: which is you ev kind of give a number to
12:14
Speaker 1: create a numerical evaluation based on the state of the
12:19
Speaker 1: board at that moment, and assign a real number evaluation
12:24
Speaker 1: to it. So, um, the highest number that you would
12:28
Speaker 1: shoot for is obviously getting checkmate, getting a king and
12:33
Speaker 1: checkmate right, right, So what you've just done now is
12:36
Speaker 1: by assigning a number to a state like the pieces
12:40
Speaker 1: on a board. Um, what you what you've done is
12:43
Speaker 1: to say, like shoot for this number. The higher the number,
12:46
Speaker 1: like you're going to give this a either rule. Now,
12:48
Speaker 1: the higher the number, the more desirable that this move
12:52
Speaker 1: that could lead to that higher number. Function. Evaluation function
12:56
Speaker 1: is what you want to do, right, like capture the
12:58
Speaker 1: night or capture the queen. Kept of the queen would
13:00
Speaker 1: have a higher evaluation number, right exactly. So that's the function.
13:04
Speaker 1: And there's another one called the mini max. Yeah, this
13:07
Speaker 1: is pretty great where you want to minimize the maximum.
13:10
Speaker 1: And this is another shortcut that they taught computers the
13:12
Speaker 1: maximum loss, that is right. So what they what they
13:15
Speaker 1: talk computers do is so you know, computer can look
13:19
Speaker 1: through an entire game every possible outcome. But what you
13:24
Speaker 1: there are computers that can look pretty far down the
13:27
Speaker 1: line at every possible outcome. And what you can say is, okay, um,
13:33
Speaker 1: you want to find the evaluation function that is the
13:36
Speaker 1: worst case scenario, the maximum loss, and then find the
13:42
Speaker 1: move that will minimize the possibility for that outcome. Yeah.
13:46
Speaker 1: By and this is your only limited by your programming power,
13:49
Speaker 1: but by looking not only at the state of the
13:51
Speaker 1: board right now, but if I make this move and
13:54
Speaker 1: I move the the pond to this spot, what are
13:57
Speaker 1: the next like three moves possibly that could happen as
14:00
Speaker 1: a result of this move. And you're only limited, like
14:03
Speaker 1: I said, by programming power. So obviously, the more juice
14:07
Speaker 1: you have, the more moves ahead that you can look exactly.
14:11
Speaker 1: And then they just shy away from ones with the
14:14
Speaker 1: higher function number or lower function number, depending on how
14:17
Speaker 1: you've programmed it. But they they they're making these decisions
14:21
Speaker 1: based on these rules, um. And then there's other things
14:23
Speaker 1: you can do, like little shortcuts to say, if if A,
14:27
Speaker 1: if A decision tree leads to a the other players
14:33
Speaker 1: King being in checkmate. Don't even think about that move
14:36
Speaker 1: any further, don't evaluate any longer, just abandon it because
14:39
Speaker 1: we would never want to make that move, right. So
14:41
Speaker 1: there's all the shortcuts you can do. And that's what
14:44
Speaker 1: they did to teach computers. That's what Deep Blue did
14:47
Speaker 1: when it beat Gary Kasparo. It was this huge, massive
14:51
Speaker 1: computer that knew a lot of chess, a lot about chess.
14:56
Speaker 1: It had a lot of rules, a lot of incredibly
14:59
Speaker 1: intricate programming that was extremely sharp, and it actually wanted
15:04
Speaker 1: became the first computer to beat an actual human chess
15:09
Speaker 1: grand master in like regulation match play. Yeah. I mean,
15:13
Speaker 1: and I don't think caspar Off gets enough credit for
15:16
Speaker 1: like willing being willing to do this, because it was
15:20
Speaker 1: a big deal for him to lose. It was in
15:23
Speaker 1: this community and the AI community. It was sent shock waves,
15:27
Speaker 1: and everyone that was alive remembers, even if you didn't
15:31
Speaker 1: know anything about either, one remembers Deep Blue being all
15:34
Speaker 1: over the news. It was a really big deal. And
15:36
Speaker 1: Casperrof put his name on the line and lost. Yeah,
15:39
Speaker 1: And I was wondering, Chuck, how how like you would
15:42
Speaker 1: get somebody to do that. I'm sure the Mountain of Catch.
15:47
Speaker 1: I guess that would probably be part of it. But
15:48
Speaker 1: I mean, I don't know. I bet, I bet that's
15:52
Speaker 1: out there. We just I just didn't look it up.
15:54
Speaker 1: So um, that's possible. It's also possible that they said, look, man,
15:58
Speaker 1: like this is chess, we're talking about what ever, But
16:00
Speaker 1: really what you're doing is helping advance artificial intelligence because
16:05
Speaker 1: we're not at we're not really trying ultimately to win
16:07
Speaker 1: chess games. We're trying to cure cancer. Yeah, we're gonna
16:10
Speaker 1: take your title because we're gonna beat you, or our
16:14
Speaker 1: machine is gonna beat you. But even still, you're gonna
16:16
Speaker 1: be helping with cancer. Think of the cancer Casparrow. That's
16:20
Speaker 1: probably what they said. Should we take a break? Yeah? Uh, what?
16:23
Speaker 1: Well should we tease our special guests? First? Is he okay?
16:27
Speaker 1: I can smell him. I don't think we even said
16:29
Speaker 1: we're gonna have a special guest. Later in the episode,
16:33
Speaker 1: Mr Jonathan Strickling of tech stuff nice. It's just been
16:36
Speaker 1: a long time since, like years since we had stricken.
16:38
Speaker 1: The last time we had stricken was like two thousand
16:41
Speaker 1: nine with the Necronomicon episode what is what where has
16:44
Speaker 1: he been besides sitting in between this every day? It's
16:47
Speaker 1: been a strickling drought is It's yes, the stricklands coming later.
16:50
Speaker 1: But we're gonna come back after this and talk a
16:52
Speaker 1: little bit more about man versus machine. Okay, dude, So
17:18
Speaker 1: what we just described was how AI was taught to
17:22
Speaker 1: play things like chess or to think like you take something,
17:26
Speaker 1: you figure out how to break it down into little
17:29
Speaker 1: rules and and and things that a computer can think of, right,
17:33
Speaker 1: and then follow these kind of rules to to make
17:35
Speaker 1: the best decision. That's how it used to be. The
17:39
Speaker 1: way that it's done now that everybody's doing now is
17:43
Speaker 1: where you are creating a machine that teaches itself. That
17:48
Speaker 1: was the breakthrough. You may have noticed back in about
17:51
Speaker 1: two all of a sudden, things like Siri and alexa
17:58
Speaker 1: um got way better at what they are doing. They
18:02
Speaker 1: got way less confused. You're a navigation app got a
18:07
Speaker 1: lot better. And the reason why is because these these
18:10
Speaker 1: this type, this new type of AI, this new type
18:12
Speaker 1: of machine learning that can teach itself and learn on
18:17
Speaker 1: its own, just hit the scene and they just started exploding.
18:20
Speaker 1: And one of the things that they were first trained
18:22
Speaker 1: on was games. Yeah, and it makes sense. Um. And
18:25
Speaker 1: if you thought chess was complicated and difficult, when it
18:29
Speaker 1: comes to these new AI s that they're teaching to
18:32
Speaker 1: teach themselves game strategy, they said, we might as well
18:36
Speaker 1: dive in to the Chinese strategic game Go because it
18:40
Speaker 1: has been called the most complex game ever devised by humans.
18:45
Speaker 1: And this was actually that was actually quote from Demi Hassabi,
18:51
Speaker 1: neuroscientists and the founder of deep Mind, which was deep Mind.
18:57
Speaker 1: They were purchased by Google or where they are always
19:00
Speaker 1: part of Google. I don't know if they were a
19:02
Speaker 1: spun off branch or where they were purchased, but it's
19:04
Speaker 1: one of Google's AI outfits. Well, they're they're one of
19:08
Speaker 1: the teams, yeah, that they are designing these new programs.
19:10
Speaker 1: And to give you an idea of how complex Go is, uh,
19:14
Speaker 1: it deals with a board with with different stones and
19:17
Speaker 1: there are uh ten how do you even say that
19:22
Speaker 1: tend to the hundred and seventy power So that means
19:25
Speaker 1: a hundred and seventies zeros uh and take that number
19:28
Speaker 1: and that's the number of possible configurations of a Go board. Right, So,
19:32
Speaker 1: like you say, chess is very complex and complicated and
19:36
Speaker 1: it's very difficult to master Go. And I've never played
19:40
Speaker 1: Go of you now, so it's supposedly it's easy to learn, right,
19:44
Speaker 1: but very complicated in its simplicity, right right, exactly, it's
19:47
Speaker 1: extremely difficult to master. And there was a guy in
19:50
Speaker 1: the late nineties, and I'm guessing that that he was
19:55
Speaker 1: saying this after Deep Blue beat casper Off. Um. There
20:00
Speaker 1: was an astrophysicists from Princeton. He said that it would
20:03
Speaker 1: probably be a hundred years before a computer beats the
20:05
Speaker 1: human at Go. To give you an idea of just
20:08
Speaker 1: how complex Go is, that deeply would just be cast
20:12
Speaker 1: prov And this guy is saying it will still be
20:14
Speaker 1: a hundred years before any anyone gets beat at Go
20:17
Speaker 1: by a computer. And he was someone who knew about
20:19
Speaker 1: this stuff, who was an astrophysicist. He wasn't just some
20:22
Speaker 1: schmoe at home and drunk in his recliner just making
20:26
Speaker 1: asinine predictions. Um. So, And again we've we've said this before,
20:31
Speaker 1: but I want to reiterate the people that, uh, I
20:35
Speaker 1: think Alpha Go is the name of this program. The
20:37
Speaker 1: people that created this a deep mind. They wanted to
20:40
Speaker 1: stress that this is a problem solving program. We're just
20:45
Speaker 1: teaching it this game at first, to to make it
20:48
Speaker 1: learn and to see if it can get good at
20:50
Speaker 1: what it does. But they said it is built with
20:54
Speaker 1: the idea that any task that has a lot of
20:56
Speaker 1: data that is unstructured, and you want to find patterns
20:59
Speaker 1: in the data and then decide what to do right,
21:02
Speaker 1: And that's kind of like what we're talking about. It's
21:04
Speaker 1: it crunches down all these possible options aka data to
21:09
Speaker 1: decide what move should I make right, and you can
21:12
Speaker 1: apply that. Ideally, they're gonna apply this to Alzheimer's and
21:15
Speaker 1: cancer and all sorts of things. Right. That's general general
21:18
Speaker 1: purpose thinking, right, yeah, and thinking on the fly to
21:22
Speaker 1: faced with novel stuff. So one of the reasons why
21:26
Speaker 1: it's good to use games like chess or go or whatever.
21:29
Speaker 1: Those are called perfect information games where both players or
21:33
Speaker 1: anybody watching has all the information that's available on it,
21:37
Speaker 1: there are definite rules their structure. It's a good proving ground.
21:41
Speaker 1: But as we'll see, AI makers are getting further and
21:44
Speaker 1: further away from those structure games as their AI becomes
21:48
Speaker 1: more and more sophisticated, because the structure and the limitations
21:53
Speaker 1: aren't necessarily needed anymore. Because these things are starting to
21:56
Speaker 1: be able to think on their own in a very
21:58
Speaker 1: generalized and even creative way. Yeah, it's really really interesting
22:05
Speaker 1: the way that they're like you said um earlier and
22:08
Speaker 1: before the break that we we don't have computers that
22:10
Speaker 1: can run all the possibilities. So what they teach in
22:13
Speaker 1: the case of AlphaGo, this program teaches itself by playing
22:17
Speaker 1: itself in these in these games and go specifically, and
22:22
Speaker 1: the more it plays itself, the more it learns, and
22:24
Speaker 1: the more ability it has during a game to choose
22:29
Speaker 1: a move by narrowing down possibilities. So instead of like, well,
22:33
Speaker 1: there are twenty million different variations here, by playing itself,
22:37
Speaker 1: it's able to say, well, in this scenario, there really
22:40
Speaker 1: only fifty different moves that I could or should make.
22:44
Speaker 1: Right or that's kind of a simplified way to say it.
22:46
Speaker 1: But right, No, but it's true. But that's that's exactly right.
22:49
Speaker 1: And what they're doing is basically the same thing that
22:51
Speaker 1: a human does. It's and it's going back to its
22:54
Speaker 1: memory banks, yeah, exact experience and saying well, I've I've
22:58
Speaker 1: been faced with something like this before, and this is
23:00
Speaker 1: what I used and it was successful. Forty out of
23:02
Speaker 1: fifty times. I'll do this one. This is a pretty
23:05
Speaker 1: reasonable move. That is what humans do. Yeah, not only
23:09
Speaker 1: I mean, boy, we screwed up the chess episode, but
23:11
Speaker 1: I get the idea that when you're a chess master,
23:14
Speaker 1: you don't just think what do the numbers say and
23:16
Speaker 1: what does the book say? But oh man, I did
23:19
Speaker 1: this move that one time and it didn't go as
23:23
Speaker 1: the book said. So that's now factored into my thinking, right,
23:27
Speaker 1: except imagine being able to learn from scratch and get
23:32
Speaker 1: to that point in eight days or eight hours. So
23:35
Speaker 1: that Go team, the Alpha Go the first the first
23:39
Speaker 1: iteration of Alpha Go. I think they started working on
23:41
Speaker 1: it in two thousand fourteen and in two thousands sixteen,
23:45
Speaker 1: at the end of two thousand and sixteen, they unleashed
23:47
Speaker 1: it um secretly onto uh an Alpha Go website and
23:53
Speaker 1: it started just wiping the floor with everybody. Everybody's like,
23:57
Speaker 1: this thing is pretty good. Oh it's Alpha Go. Well,
23:59
Speaker 1: here's this it. That was the end of two thousand
24:01
Speaker 1: and sixteen. Okay, so chess had already come and gone.
24:05
Speaker 1: Like by this point, you can download a program that's
24:08
Speaker 1: like deep Blue, right, that was that's a great point. Yeah,
24:11
Speaker 1: Like today the stuff you played chess with on your
24:15
Speaker 1: laptop is even more advanced than Deep Blue was in
24:18
Speaker 1: the nineties. And it's just on your laptop. Um, but
24:21
Speaker 1: this is so, this is Go. This is the end
24:23
Speaker 1: of two thousand and sixteen. The end of two thousand seventeen,
24:26
Speaker 1: um Alpha Go was replaced with Alpha Go zero. It
24:32
Speaker 1: learned what Alpha Go had taken two years or three
24:35
Speaker 1: years to learn in forty days by teaching itself, and
24:41
Speaker 1: it beat the master. And finally, in May of two
24:44
Speaker 1: thousand seventeen, Alpha Go took on Key G, the highest
24:50
Speaker 1: ranked Go player in the world. I don't know if
24:53
Speaker 1: he or she still is. No Lisa A. Doll is
24:57
Speaker 1: the current or was until Alpha alpha Go beat him. Man, yeah,
25:03
Speaker 1: do they get knocked off in Alpha Go is the champion? Like,
25:05
Speaker 1: that's that's not fair. I if it's match play and
25:10
Speaker 1: the player, the human players accepted a challenge from the computer.
25:15
Speaker 1: I don't see why it wouldn't be the world champions.
25:18
Speaker 1: Or do they just now say on websites like human
25:22
Speaker 1: champions maybe in italics with like a sneer maybe yeah? Interesting?
25:28
Speaker 1: What do they call that wet wear? Like your brain,
25:32
Speaker 1: your neurons and all that what instead of hardware, it's
25:35
Speaker 1: wet wear. Oh, I don't know about that. I think
25:37
Speaker 1: that's the term for it. What does that mean? Though?
25:39
Speaker 1: It means like you're you have a substrate, right, you're intelligence.
25:43
Speaker 1: You're intellect is based on your neurons and they're firing
25:47
Speaker 1: all that stuff, and it's wet and squishy and meat.
25:50
Speaker 1: And there's hardware that you can do the same thing on,
25:53
Speaker 1: you can build intelligence on, but it's hardware, it's not
25:56
Speaker 1: wet wear. Interesting, so it's probably it. It's the wet
26:00
Speaker 1: champion versus the hardware champion. But wet wears italicized what
26:05
Speaker 1: the sneer? Uh So where things really got interesting because
26:08
Speaker 1: you were talking earlier about um, what what is it
26:12
Speaker 1: with the chess and go? What are they called? What
26:13
Speaker 1: kind of games? Perfect information games? Right? Then you think
26:19
Speaker 1: And my first thought when you said that was well, yeah,
26:22
Speaker 1: then then there's there's games like poker like Texas Hold Him,
26:26
Speaker 1: where there are a set of rules. But poker is
26:29
Speaker 1: not about the set of rules. It is about sitting
26:31
Speaker 1: down in front of whatever five or six people and lying,
26:35
Speaker 1: bluffing and getting away with it in your game face
26:39
Speaker 1: being bluff Like there's so many human emotions and contextual
26:44
Speaker 1: clues and and uh micro expressions and all these things,
26:49
Speaker 1: like surely you could never ever teach a machine to
26:53
Speaker 1: win at Texas Holding poker. Yeah, it'll be a hundred
26:56
Speaker 1: years at least before that happens, I predict. No, they
27:00
Speaker 1: did it, and more than one team has done it. Yeah.
27:04
Speaker 1: I read um there was one from Carnegie Melon called
27:08
Speaker 1: Liberatus Ai go melon heads, Yeah, go the Thornton Melons.
27:15
Speaker 1: Uh yeah, I mean that's was. The University of Alberta
27:19
Speaker 1: has one called deep Stack. That was the one I
27:21
Speaker 1: read about. And it actually here's the thing, like if
27:24
Speaker 1: you read the release on it, you're like, you don't
27:28
Speaker 1: know how this thing works to you? Yeah, And I'm
27:31
Speaker 1: pretty sure they don't fully get it because that's one
27:33
Speaker 1: of the problems. Actually talked about this in the Existential
27:36
Speaker 1: Risks series that's Scared that is to be released there, right,
27:40
Speaker 1: that there is a there is a type of machine
27:43
Speaker 1: learning where the machine teaches itself, but we don't really
27:46
Speaker 1: understand how it's teaching probably the scariest one, right, or
27:50
Speaker 1: what it's learning, But that's the most prevalent one. That's
27:52
Speaker 1: what a lot of this is is like these machines.
27:54
Speaker 1: It's like, here's here's chess, go figure it out and
27:59
Speaker 1: they go, okay, got it. How did you do that?
28:02
Speaker 1: Wouldn't you like to know? So that's the scariest presentation
28:06
Speaker 1: you will see on ai as when someone says, well,
28:09
Speaker 1: how does all this work? And they go, but we
28:12
Speaker 1: just know it can beat the human at poker. But
28:14
Speaker 1: the thing about Deep Stack at the University of Alberta
28:17
Speaker 1: is that it learned somehow some sort of intuition, because
28:22
Speaker 1: that's what's required is not just the perfect information where
28:26
Speaker 1: you have all the information on the board. It's with poker,
28:30
Speaker 1: you don't know what the other person's cards are and
28:32
Speaker 1: you don't know if they're lying or bluffing or what
28:34
Speaker 1: they're doing. Um, so that's an imperfect information game. So
28:39
Speaker 1: that would require intuition. And apparently not one, but two
28:42
Speaker 1: different research groups taught AI too into it. Yeah, Carnegie
28:47
Speaker 1: Mellon came out in uh January of two thousand seventeen
28:50
Speaker 1: with its Liberatus AI and they said they spent twenty
28:54
Speaker 1: days playing a hundred and twenty thou hands of Texas
28:58
Speaker 1: hold them with four professional poker players and one and
29:04
Speaker 1: smoked him. Basically got up to They weren't playing with
29:07
Speaker 1: real money obviously, but they they would have been great funded.
29:12
Speaker 1: Their next project, Liberatis, was up by one point seven million,
29:17
Speaker 1: and one of the quotes from one of the poker
29:19
Speaker 1: players that he made two Wired magazine said, I felt
29:22
Speaker 1: like I was playing against someone who was cheating, like
29:24
Speaker 1: it could see my cards. I'm not accusing it of cheating,
29:27
Speaker 1: it was just that good. So that's a really interesting thing. Man,
29:32
Speaker 1: that they could teach self teach program, or a program
29:37
Speaker 1: could teach itself intuition. That's creepy. I thought this part
29:42
Speaker 1: was interesting. The Atari stuff, Um, this gets pretty fun.
29:46
Speaker 1: Google deep mind. Let it's a i um reek havoc
29:52
Speaker 1: on Atari forty nine different at twenty games, see they
29:56
Speaker 1: could figure out how to win, and apparently the most
30:00
Speaker 1: difficult one was Miss Pacman, which is a tough game.
30:04
Speaker 1: Still man, Miss pac Man. They nailed it. It's still
30:07
Speaker 1: one of the great games. But but there um their
30:11
Speaker 1: game or their Q deep Q network algorithm beat it. Yeah.
30:17
Speaker 1: I think it got the highest score nine hundred points,
30:21
Speaker 1: and no human or machine has ever achieved that high
30:24
Speaker 1: score from what I understand. Amazing and the way this
30:26
Speaker 1: one does it. The hybrid reward architecture that it uses
30:31
Speaker 1: is really interesting. It's it says here, it generates a
30:34
Speaker 1: top agent that's like a senior manager, and then all
30:38
Speaker 1: these other hundred and fifty individual agents. So it's almost
30:41
Speaker 1: like they've devised this artificial structural hierarchy of these little
30:47
Speaker 1: worker agents to go out and collect I guess data
30:51
Speaker 1: and then move it up the chain to this top agent. Right,
30:56
Speaker 1: and then this thing says, um, Okay, you know, I
31:00
Speaker 1: think that you're probably right what these agents are probably doing.
31:05
Speaker 1: And I don't know this is exactly true, but there's
31:08
Speaker 1: there are models out there like this where the agent
31:11
Speaker 1: says this is um, you have a nine chance of
31:14
Speaker 1: success at getting this pellet um if we take this action.
31:19
Speaker 1: Somebody else says, you've got a two percent chance of
31:21
Speaker 1: evading this ghost if we go this way. And then
31:24
Speaker 1: the the top agent, the senior manager, can put all
31:28
Speaker 1: this stuff together and say, well, if I listen to
31:30
Speaker 1: this guy and this guy done, only will I evade
31:32
Speaker 1: this ghost. I'll go get this pellet um. And it's
31:35
Speaker 1: based on what what confidence level that the lower agents
31:40
Speaker 1: have in success in recommending these moves, and then the
31:44
Speaker 1: top agent ways these things. They should give him a
31:47
Speaker 1: little a little cap But all this is happening like that,
31:51
Speaker 1: you know what I'm saying. This isn't like well, hold on,
31:53
Speaker 1: hold on, everybody, what is Harvey? Harvey? What do you
31:55
Speaker 1: have to say? Well, let's get some Chinese in here
31:58
Speaker 1: and and hash it out. And everybody sits there in
32:01
Speaker 1: order some Chinese food, and then you wait for it
32:03
Speaker 1: to come and then you pick up the meeting from
32:05
Speaker 1: that point on. And then finally Harvey gives his idea,
32:09
Speaker 1: but he forgot what he was talking about, so he
32:11
Speaker 1: just sits down and eats his agrol. Well here's a
32:14
Speaker 1: pretty frightening. Uh. Survey. Uh. There was a survey of
32:17
Speaker 1: more than three fifty AI researchers and they had the
32:20
Speaker 1: following things to say. And these are the pros that
32:23
Speaker 1: are doing this for a living. They predicted that within
32:26
Speaker 1: ten years, ay, I will drive better than we do,
32:30
Speaker 1: they will be able to write a best selling novel,
32:33
Speaker 1: A I will generate this, and bye be better at
32:37
Speaker 1: performing surgery than humans are. You know. So again, one
32:42
Speaker 1: of the things that about the field of artificial intelligence
32:47
Speaker 1: a you know a lot about now famous. It is
32:50
Speaker 1: famous for making huge predictions that did not pan out.
32:54
Speaker 1: But you've also seen it's it's also famous for beating
32:58
Speaker 1: predictions that you know been levied against it um. But
33:02
Speaker 1: there is something in there, Chuck, that stands out to me,
33:06
Speaker 1: and that's the idea of an AI writing a novel.
33:09
Speaker 1: Like for a very long time, I thought, well, yeah, okay,
33:12
Speaker 1: you can teach a robot arm to like put a
33:15
Speaker 1: car part or something somewhere if you wanted to just
33:18
Speaker 1: follow these these mechanical things, or it can use in
33:21
Speaker 1: a wish, or it can use logic and reason. But
33:24
Speaker 1: to create that's different, right. That was like the new frontier.
33:27
Speaker 1: It used to be chess and then it used there
33:29
Speaker 1: was go. The next frontier is creativity, and they're starting
33:33
Speaker 1: to bang on that door big time. There's a game
33:36
Speaker 1: designing AI called Angelina out of the University of foul Myth,
33:40
Speaker 1: which I always want to say foul Mouth, but we'll
33:43
Speaker 1: just call it Foulmouth like it's supposed to. And Angelina
33:46
Speaker 1: actually comes up with ideas for new games, not um,
33:53
Speaker 1: like a different level or something like like you should
33:56
Speaker 1: put a purple loincloth on that player. You know that'll
33:59
Speaker 1: look kind of like new games, but whacked out games
34:03
Speaker 1: that humans would never think of. One example I saw
34:06
Speaker 1: is in a Dungeon Battle Royal game, a player controls
34:10
Speaker 1: like ten players at once, and some you have to
34:13
Speaker 1: sacrifice to be killed to save the others. Like the
34:16
Speaker 1: stuff that human wouldn't necessarily think of, this AI is
34:19
Speaker 1: coming up with. Well, I mean, when you think of creatively,
34:22
Speaker 1: especially something like writing a novel or a film, if
34:24
Speaker 1: there are only seven stories, I mean, and that sort
34:28
Speaker 1: of thinking that they're basically every every dramatic story is
34:32
Speaker 1: a variation of one of seven things. Yeah. So I
34:35
Speaker 1: mean you can look at like um AI is scary,
34:39
Speaker 1: and in some ways very much is and can be,
34:43
Speaker 1: but there's also like definitely a level of excitement of
34:46
Speaker 1: the whole thing and the idea that there are artificial
34:50
Speaker 1: minds that are coming online or that have come online now,
34:53
Speaker 1: that are out there that are they'll they'll just naturally
34:56
Speaker 1: by definition, see things differently than we do, and the
35:00
Speaker 1: idea that they can come up with stuff that we've
35:02
Speaker 1: never even thought of that is just gonna knock our
35:03
Speaker 1: socks off, hopefully in good ways. Um that's that's a
35:07
Speaker 1: really cool thing. And so maybe there's just seven as
35:11
Speaker 1: far as humans know, but there's an unlimited amount if
35:14
Speaker 1: if you put computer minds to thinking about these kind
35:17
Speaker 1: of things, that's the premise of it. Right, So the
35:19
Speaker 1: robot would be like you never thought of boy meets
35:22
Speaker 1: girl meets well trilo bite. But see even that's a
35:28
Speaker 1: variation of right, just imagine something that just we've never
35:33
Speaker 1: even thought. Well, Gina, how they should do this? If
35:35
Speaker 1: they do do that is uh is not is just
35:39
Speaker 1: release a book and not tell anyone that was written
35:43
Speaker 1: by an AI program, because if they do that, then
35:46
Speaker 1: it's going to be so under scrutiny. They should secretly
35:49
Speaker 1: release this book and then after it's a New York
35:52
Speaker 1: Times bestseller, Say, meet the Whopper, the author of this.
35:58
Speaker 1: You know his interests are rule skating, playing tic tac toe,
36:01
Speaker 1: and global thermal nuclear war. All right, should we take
36:05
Speaker 1: a break and get strickling in here. Yeah, we're gonna
36:07
Speaker 1: end the strickling drought because it is about to reign
36:09
Speaker 1: strickling in this piece. Gross. Okay, we're back and get this.
36:37
Speaker 1: The scent of strick has permeated our place, and that's
36:41
Speaker 1: a beautiful scent. It smells like a soldering gun and
36:46
Speaker 1: a circuit board. You know, feel the lavender in a
36:48
Speaker 1: protein bar. That's fair. I'm gonna say, Drac o' noir
36:52
Speaker 1: that that would have been a lie. Is that how
36:54
Speaker 1: you said? I always called it Drecar, drac Car. That's
36:58
Speaker 1: that's fair, d Car. I always pronounced it Benetton colors.
37:03
Speaker 1: I think that was what I wore. Oh is that
37:05
Speaker 1: what you wore? Yeah? During my what I call the
37:07
Speaker 1: Year of Cologne, I had a couple of seven. Uh,
37:13
Speaker 1: this is scintillating. Why are you? Why am I here?
37:17
Speaker 1: So we know that you already know because we talked
37:20
Speaker 1: via email about this, but we'll tell everybody else. We
37:23
Speaker 1: have brought you in here because you're the master of tech.
37:26
Speaker 1: And we were talking tech today, which we've talked about
37:29
Speaker 1: without you before. But frankly, Chuck and I and Jerry
37:31
Speaker 1: huddled and we said this is not quite as good
37:33
Speaker 1: without strict so let's try something different. Got you and
37:37
Speaker 1: and we're talking about games and machine versus man and
37:42
Speaker 1: that whole whole evolution and how that's gone super crazy
37:46
Speaker 1: over the last few years. Games without frontiers, Peter Gabriel
37:49
Speaker 1: would say, war without fear. And we've talked, I mean,
37:51
Speaker 1: we've talked a lot about the evolution of machine learning
37:55
Speaker 1: and how now it's starting to take off like a
37:57
Speaker 1: rocket because they can teach themselves. But one thing we
38:00
Speaker 1: haven't really talked about are solved games. And we talked
38:04
Speaker 1: about chess. Yeah, we talked about go right with those
38:08
Speaker 1: constitutes solved games, not really so so a solved game
38:12
Speaker 1: is the concept where if you were to assume perfect
38:15
Speaker 1: play on either sides of the game, you would always
38:19
Speaker 1: know how it was going to end, which we always
38:21
Speaker 1: assume perfect play, right, that's kind of work bags stuff
38:24
Speaker 1: should know motto. So perfect play just means that no
38:27
Speaker 1: one ever makes a mistake, so very much the way
38:29
Speaker 1: I do my work right stuff exactly. So, if you
38:33
Speaker 1: were to take a game like Tic tac toe and
38:36
Speaker 1: you assume perfect play on both sides, it is always
38:38
Speaker 1: going to end in a draw, which is what's in
38:41
Speaker 1: war games. Yes, right, the only way to win is
38:43
Speaker 1: not to play. Yes, so a game, it's a game
38:48
Speaker 1: like Connect four, whoever goes first is always going to win,
38:52
Speaker 1: assuming perfect play sides. Yes, what, I don't think I've
38:57
Speaker 1: played Connect for it. That's where you dropped for a
38:59
Speaker 1: long time, when where you drop the little uh tokens
39:04
Speaker 1: kind of checkers. We did an interstitial playing Connect for Remember,
39:07
Speaker 1: I was speaking it though, and you had perfect place,
39:09
Speaker 1: so I knew it was useless. I was gonna say
39:11
Speaker 1: that I'm so humiliated by all the Connect four games
39:14
Speaker 1: that I've lost starting even Yeah, but but I mean
39:18
Speaker 1: perfect play. That's something that that obviously only the the
39:22
Speaker 1: best players typically achieve with with significantly complex games. Obviously,
39:27
Speaker 1: the simpler the game, the easier it is to play perfectly.
39:31
Speaker 1: Right tic tac toe? If you know, once you've mastered
39:34
Speaker 1: the basics of tic tac toe and the other person has,
39:36
Speaker 1: you're never really going to win unless someone has just
39:39
Speaker 1: made a silly mistake because they weren't paying attention like
39:43
Speaker 1: a star instead of an ex right, which doesn't count
39:46
Speaker 1: automatically disqualified you. One thing I found this very enjoyable
39:50
Speaker 1: is playing with little kids who haven't figured out that
39:52
Speaker 1: tic tac toe is very easy to bring. Yes, smashed
39:55
Speaker 1: their face on the board, rub it in. Yeah, the
39:58
Speaker 1: same reason why I like to join in on little
40:00
Speaker 1: league games, because I can really whale that ball out
40:02
Speaker 1: of the park. Yeah, it really missed me, feel like
40:05
Speaker 1: a man. That's the most tech stuffy thing you've ever said.
40:07
Speaker 1: You really wail that ball out of the park. Well,
40:10
Speaker 1: to be fair, I did just do a tech stuff
40:12
Speaker 1: episode about the technology behind baseball bats so fresh. Actually
40:17
Speaker 1: it's a lot of fun. So there have been a
40:19
Speaker 1: lot of games that have been solved, but Checkers was
40:22
Speaker 1: one that was recently solved back in recently by the
40:25
Speaker 1: early nineties when it was played against a computer called
40:29
Speaker 1: Chinook and uh h I in O okay, yeah, like
40:33
Speaker 1: the helicopter or the winds that blow through Alberta exactly.
40:37
Speaker 1: And so there are certain games that are more easily
40:42
Speaker 1: solved than others. You do it through an algorithm, but
40:45
Speaker 1: other games, like chess, are more complicated because you can
40:48
Speaker 1: in chess you have multiple moves that you can do
40:51
Speaker 1: where you can you can move a piece back the
40:53
Speaker 1: way you win, right, it's not you're not committed to
40:56
Speaker 1: going a specific direction with certain pieces, like with a
40:59
Speaker 1: night you know you can you could go right back
41:01
Speaker 1: to where you started on the next move if you
41:03
Speaker 1: wanted to, and that creates more complexity. So the more
41:08
Speaker 1: complex the game, the more difficult it is to solve.
41:10
Speaker 1: And some games are not solvable simply because you'll never
41:14
Speaker 1: know what the full state of the game is from
41:18
Speaker 1: any given moment. Did you have a chance to talk
41:21
Speaker 1: about the difference between perfect knowledge and imperfect knowledge and
41:25
Speaker 1: a game. Yeah, yeah, we talked about that. Yeah, so
41:28
Speaker 1: so computers obviously, they do really well if they understand
41:33
Speaker 1: the exact state of the game all the way through.
41:35
Speaker 1: If they if they have perfect knowledge, all of the
41:37
Speaker 1: information is there on the board right and and all
41:40
Speaker 1: players can see all information at all times. But games
41:43
Speaker 1: like poker what you guys talked about, obviously you have
41:46
Speaker 1: imperfect information. You only know part of the state of
41:49
Speaker 1: the game. That's why those games have been more difficult,
41:52
Speaker 1: more challenging for computers to get better than humans until
41:57
Speaker 1: relatively recently. And there've been two major ways of doing that.
42:00
Speaker 1: You either throw more processing power at it, like you
42:03
Speaker 1: get a supercomputer, or you create neural networks, artificial neural networks,
42:07
Speaker 1: and you start teaching computers to quote unquote learn the
42:11
Speaker 1: white people do. So we talked about that, and one
42:15
Speaker 1: of the things that we talked about was how there's
42:17
Speaker 1: this idea that they the programmers, especially, say, the people
42:22
Speaker 1: who are making programs that are playing poker and are
42:24
Speaker 1: getting good at poker, aren't exactly sure how the machines
42:28
Speaker 1: are learning to play poker or what they're learning. They're
42:31
Speaker 1: just getting better at poker. Do they know how they're
42:35
Speaker 1: learning poker? They just know that they're learning poker and
42:38
Speaker 1: that they're good at it. Like, where's the intuition? How
42:41
Speaker 1: is that being learned? An excellent question. The way it
42:44
Speaker 1: typically is learned, especially with artificial neural networks, is that
42:47
Speaker 1: you you set up the computer to play millions of
42:52
Speaker 1: hands of poker that are are randomly assigned, so it's
42:56
Speaker 1: it's truly as random as computers can get. That's the
42:59
Speaker 1: whole philosophical discussion that I don't think we're ready to
43:02
Speaker 1: go into right now. But you have games come up
43:06
Speaker 1: where the computer is playing itself millions upon millions of
43:10
Speaker 1: times and learning every single time how the statistics play out,
43:15
Speaker 1: how different betting strategies play out. It's it's sort of
43:19
Speaker 1: partitioning its own mind to play against itself. And through
43:23
Speaker 1: that process, it's as if you, as a human player,
43:27
Speaker 1: we're playing thousands of games with your friends and you
43:30
Speaker 1: start to figure out, oh, when I have these particular
43:33
Speaker 1: cards and they're in my hand, and let's say we're
43:35
Speaker 1: playing Texas, hold them and the community cards are are these?
43:39
Speaker 1: Then I know that generally speaking, maybe three times out
43:43
Speaker 1: of ten I end up winning, maybe I shouldn't bet.
43:45
Speaker 1: And well, the computer is doing that, but on a
43:47
Speaker 1: scale that far dwarfs what any human can do, and
43:51
Speaker 1: in a in a fraction in the amount of time,
43:53
Speaker 1: And so it's it's sort of well, it's intuition in
43:57
Speaker 1: the sense of it's just done it so much, right,
44:00
Speaker 1: But is that? Does that mean it's completely ignoring um
44:05
Speaker 1: micro expressions and facial cues, So that didn't even come
44:08
Speaker 1: into play? Ye, yeah, I was. I was waiting for
44:13
Speaker 1: you've been doing this. I still nod when I do
44:15
Speaker 1: a solo show and I do a lot of expressive dance.
44:18
Speaker 1: What do you think, Jonathan, I don't know, Jonathan. It
44:21
Speaker 1: gets lonely in here, guys. But yes, what you're saying
44:25
Speaker 1: all the tells. It tells that you would use as
44:28
Speaker 1: a human player. The computer does not pick up typically data. Yes, typically,
44:33
Speaker 1: what it would do is it would study the outcomes
44:35
Speaker 1: of the games from a purely statistical expressions. Most of
44:40
Speaker 1: these poker games tend to be computer based poker games.
44:43
Speaker 1: So it's not that it's playing like it's it's not
44:45
Speaker 1: like there's a computer that says push ten more chips
44:48
Speaker 1: into the table. You know, it's right, exactly, a little
44:53
Speaker 1: it's a little winky face emoticon, like I don't have
44:56
Speaker 1: good cards. No, it's it's all usually over of like
45:00
Speaker 1: internet poker, which a lot of the people who play
45:03
Speaker 1: professional poker cut their teeth on, especially you know in
45:06
Speaker 1: the more recent generations of professional poker players. Yeah, they
45:11
Speaker 1: don't know what it's like being a smokey saloon like Moneymaker.
45:15
Speaker 1: In money Maker rose to the top a few years ago,
45:17
Speaker 1: more than like a decade ago. Now he had come
45:20
Speaker 1: from the world of internet poker, and so he was
45:23
Speaker 1: using those same sort of skills in a real world setting.
45:27
Speaker 1: But obviously there are subtle things that we humans do
45:31
Speaker 1: in our expressions that computers do not pick up on,
45:33
Speaker 1: and in fact that leads us sort of into the
45:36
Speaker 1: realm of games where computers don't do as well as humans.
45:41
Speaker 1: Is that list you sent a joke or is it really? No,
45:44
Speaker 1: that's real. It does seem like it's weird, like one
45:46
Speaker 1: of the games on there is Pictionary, for example, or tag. Yeah,
45:50
Speaker 1: but these are some of these are are They sound silly,
45:54
Speaker 1: But when you start to think about them in terms
45:55
Speaker 1: of computation and robotics, you start to realize how incredibly
45:59
Speaker 1: complex it is from a technical perspective, but incredibly easy
46:03
Speaker 1: it is for your average human being. So with humans
46:07
Speaker 1: a game of tag, once you know the basics, it's
46:10
Speaker 1: it's all an instinct. You know what to do. You
46:12
Speaker 1: run after the person you tried to catch up with
46:14
Speaker 1: them and tag them. But you also know push him
46:16
Speaker 1: in the back as hard as you can. Well, if
46:19
Speaker 1: you're Josh, you push him as hard as you can.
46:20
Speaker 1: But most of us we tag and we're not trying
46:22
Speaker 1: to cause harm. Robots, however, robots not so good on
46:29
Speaker 1: you said, and just say Isaac Asimov Isaac Asimov's rules
46:33
Speaker 1: of robotics. Aside, robots are not very good at judging
46:36
Speaker 1: how hard they have to hit something in order to
46:39
Speaker 1: make contact, right. They're not as good at even your
46:43
Speaker 1: bipedal robots that walk around like people, even the ones
46:46
Speaker 1: that can run and do flips and stuff. Have you
46:48
Speaker 1: seen that one the other day that the footage of
46:50
Speaker 1: that thing running and jumping, it's really impressive and in
46:53
Speaker 1: super creepy. Yeah, but even so, that's that's a clip
46:58
Speaker 1: of the best of If you ever if you ever
47:00
Speaker 1: see the clips where they show all the times the
47:02
Speaker 1: robots fallen over, we'll pouring hot coffee in someone's head. Yes,
47:07
Speaker 1: but they always play those clips shows to Yake, yes
47:10
Speaker 1: that this this is true. So so, DARPA had its
47:13
Speaker 1: big robotics challenge a few years ago where they had
47:16
Speaker 1: bipedal robots tried to go through a scenario that was
47:21
Speaker 1: simulating the Fukushima nuclear disaster. So the interesting thing was
47:26
Speaker 1: the robot had to complete a series of tasks that
47:29
Speaker 1: would have been mundane to humans, things like open up
47:32
Speaker 1: a door and walk through it, and pick up a
47:35
Speaker 1: power power tool and use it against a wall. And
47:39
Speaker 1: you can watch the footage of some of these robots
47:42
Speaker 1: doing things like being unable to open the door because
47:46
Speaker 1: they can't tell if they need to pull or push
47:49
Speaker 1: or they open the door, but then immediately fall over
47:51
Speaker 1: the threshold of the door. And when you see that,
47:54
Speaker 1: you realize, as advanced as robotics is, as advanced as
47:58
Speaker 1: machine learning has become, as incredible as our technology has progressed,
48:04
Speaker 1: there are still things that are fundamentally simple to your
48:07
Speaker 1: average humans that are incredibly complicated from a technical standpoint,
48:11
Speaker 1: Like a six year old can play jinga better than
48:13
Speaker 1: a robot, right right, right, Okay, But the thing is
48:16
Speaker 1: we're talking robots here, and is we go more and
48:18
Speaker 1: more and more online in our world, becomes more and
48:21
Speaker 1: more like web based rather than reality based. Doesn't the
48:27
Speaker 1: the fact that a robot can't walk through a door
48:29
Speaker 1: matter less and less. And the idea that that machines
48:34
Speaker 1: are learning intellect and the robotivity, you just blew my
48:39
Speaker 1: mind that that's becoming more and more vital and important
48:42
Speaker 1: and something we should be paying attention. It absolutely is
48:45
Speaker 1: something we should pay attention to. I mean, we have
48:47
Speaker 1: robotic stock traders there trading on thousands of trades per second,
48:53
Speaker 1: right fast, so fast that we have had stock market
48:56
Speaker 1: booms and crashes that last less than a second long
49:00
Speaker 1: due to that. So the robot army that will ultimately
49:03
Speaker 1: defeat us is not something from the terminator. It's invisible.
49:07
Speaker 1: It's online, or it will be online. It's it's it's
49:10
Speaker 1: it's what's determining our retirement, right, yeah, the global economy
49:15
Speaker 1: or um our municipal waters apply or whatever. Yeah. Now
49:19
Speaker 1: there's the fascinating thing to me about this is not
49:24
Speaker 1: just that we're training machine intelligence to learn and to
49:28
Speaker 1: perform at a level better than humans, but that we're
49:31
Speaker 1: putting a lot of trust in those devices and things
49:36
Speaker 1: that have real incredible impact on our lives, the significant
49:41
Speaker 1: enough impact where if things were to go south, it
49:43
Speaker 1: would be really bad for us. Uh, and not in
49:46
Speaker 1: that Terminator respect. Terminator is a terrifying dystopian science fiction story.
49:53
Speaker 1: But then when you realize what could really happen behind
49:56
Speaker 1: the scenes, you think the robots don't have to do
49:59
Speaker 1: any physical harm to us to really mess things up.
50:02
Speaker 1: So there are certainly some cases for us to be
50:06
Speaker 1: very vigilant in the way we deploy right from the
50:11
Speaker 1: outset exactly, and that it depends not necessarily I think.
50:17
Speaker 1: I think it's I don't think it's too late, but
50:21
Speaker 1: I think it's getting to that point of no return
50:25
Speaker 1: very very quickly. December of this year. Yeah, Well, if
50:28
Speaker 1: you're if you're someone like if you're someone like Elon Musk,
50:31
Speaker 1: you'd say, if we don't do something now, we're we're
50:34
Speaker 1: totally going to plummet off the edge of the cliff.
50:37
Speaker 1: But now is a window that is rapidly closing. Yes, yeah, yeah,
50:41
Speaker 1: the now is the now is a time where we've
50:43
Speaker 1: got a deadline. We don't know exactly when that deadline
50:47
Speaker 1: is going to be up, but we know that it's
50:50
Speaker 1: not getting further out it. We're just getting closer to
50:52
Speaker 1: that deadline. So, and a lot of this is is
50:56
Speaker 1: it covered in deep conversations and the artificial and eligience
51:00
Speaker 1: and machine learning fields that has been going on for ages,
51:04
Speaker 1: to the point where you even have bodies like the
51:07
Speaker 1: European Union that have debated on concepts like granting personhood
51:12
Speaker 1: to artificial intelligence. So this is a really fascinating and
51:17
Speaker 1: deep subject that and the games thing is a great
51:21
Speaker 1: entry point into having that conversation. Uh, you know, I'm
51:25
Speaker 1: lucky if I can win a game of chess against
51:26
Speaker 1: another human being. Right, So, I can't even describe chess,
51:31
Speaker 1: did I did. My big thing is I do that
51:33
Speaker 1: night thing I called the night shuffle. I just move
51:35
Speaker 1: them back and forth. I just castle. If I can castle,
51:39
Speaker 1: then I'm I'm I'm so happy. And that's the third
51:43
Speaker 1: tech stuffiest thing to come in threes. Well, strict, thank
51:46
Speaker 1: you for stopping by. Thank you should stick around for
51:49
Speaker 1: a listener mail. I think you should too, love to
51:52
Speaker 1: and and throw out any funny comments that you have
51:55
Speaker 1: I'll throw out comments and then Jerry can decide which
51:58
Speaker 1: ones are funny. Okay, all right, fair, all right, So
52:01
Speaker 1: if you want to know more about AI, go listen
52:03
Speaker 1: to Tech Stuff. Strict does this every week, what days, Monday, Tuesday, Wednesday, Thursday,
52:10
Speaker 1: and Friday. That's amazing, buddy, and wherever you find your
52:14
Speaker 1: podcast Okay, and you've been doing it for years, so
52:18
Speaker 1: if you love this, there's a whole big backlog, nine
52:20
Speaker 1: plus episodes you're celebrating. You're celebrating your ten year as well,
52:24
Speaker 1: right yep, by sure, m I'll be uh, we'll be
52:26
Speaker 1: turning ten and Tech Stuff on June eleven, congratulating. Well,
52:31
Speaker 1: since I said happy anniversary, it means it's time for
52:33
Speaker 1: listener mail. Guys. I'm gonna call this, uh Matt graining
52:39
Speaker 1: and cultural relativism about that. Hey, guys, love your podcast
52:44
Speaker 1: so much. The massive archive makes for endless learning and entertainment.
52:47
Speaker 1: My favorite part is you were such rad guys, including Strickland,
52:51
Speaker 1: and I could totally imagine how did they know? I
52:54
Speaker 1: can totally imagine myself getting a beer with you two,
52:57
Speaker 1: But without Strickland, your Simpsons episodes were absolutely perfect. I
53:01
Speaker 1: used to live in Portland and drove on Flanders and
53:03
Speaker 1: love Joy Streets a lot. Wait, is this Matt Greening,
53:07
Speaker 1: Matt Graenning, Drew bart and the sidewalk cement behind Lincoln
53:11
Speaker 1: High School in downtown Portland's You can google that. I
53:14
Speaker 1: would like to offer one interesting observation though, I've noticed
53:17
Speaker 1: that on several episodes, you guys have said that you
53:19
Speaker 1: are cultural relativists. Is that pronounced right? Yeah? But then
53:23
Speaker 1: in nearly every episode I hear you pass moral judgments
53:27
Speaker 1: on all the messed up stuff that people do, whether
53:29
Speaker 1: it's racism, freak shows, or crematori ums bearing bodies on
53:33
Speaker 1: the sly you guys are never shy to condemn something
53:36
Speaker 1: that deserves to be condemned. Reminds me of something I
53:38
Speaker 1: read from Yale sociologists who Loop Gorsky, who points out
53:43
Speaker 1: that our own relativism is rarely as radical as our
53:47
Speaker 1: theory requires. We can't be complete relativists in our daily lives.
53:52
Speaker 1: He then gives the example of how academic social scientists,
53:55
Speaker 1: where die hard relativists, get furious at uh and moralistic
53:59
Speaker 1: at the data fudging of other researchers. Anyway, love the show, guys,
54:04
Speaker 1: love tech stuff especially, and will forever be indebted to
54:07
Speaker 1: you for your hilarity and knowledgeability. Cheers Jesse Lusco ps
54:13
Speaker 1: go tech stuff. That's that's sweet. So yeah, thanks a lot, Jesse. Um.
54:18
Speaker 1: There was an actual episode, and I don't remember which
54:20
Speaker 1: one it was, where we abandon our cultural relativism, do
54:23
Speaker 1: you remember, Because we used to just be like, no judgment,
54:26
Speaker 1: no judgment, we just can't judge, you know, and then
54:29
Speaker 1: finally we're like, you know what, No, that's not true.
54:32
Speaker 1: We changed our the our philosophy to include the idea
54:35
Speaker 1: that there are moral absolutes that are universal, although sometimes
54:40
Speaker 1: we are just judging even beyond that. Look at us. Yeah, well,
54:44
Speaker 1: if you want to get in touch with us, you
54:46
Speaker 1: can tweet to us. I'm at josh um Clark and uh,
54:50
Speaker 1: John Strickland's at John Strickland. That's correct. On Twitter, Chuck's
54:54
Speaker 1: a movie crush and uh, you have a tech stuff
54:57
Speaker 1: hsw Twitter, right, that's right. And then we also have
54:59
Speaker 1: s y sk podcast. We're on Facebook dot com slash
55:02
Speaker 1: stuff you should know, slash tech stuff and slash movie Crush.
55:07
Speaker 1: Do you have a movie crush page for Facebook? Yeah? Yeah,
55:10
Speaker 1: that's actually where I spend most of my time. Oh,
55:12
Speaker 1: I didn't know that, said it's time. And then there's
55:15
Speaker 1: also a slash stuff you Should Know Facebook page. You
55:17
Speaker 1: guys have a lot of Facebook and to do it
55:19
Speaker 1: so many social medias flying around. You can send us
55:22
Speaker 1: an email to stuff podcast at how stuff Works dot com.
55:25
Speaker 1: You can send John an email to tex stuff at
55:28
Speaker 1: how stuff Works dot com. Nice, and then hang out
55:30
Speaker 1: with us at our home on the web Stuff you
55:32
Speaker 1: Should Know dot com and just go to tech stuff.
55:36
Speaker 1: Just search it in Google. I come up all the time.
55:38
Speaker 1: Fair enough for more on this and thousands of other topics,
55:46
Speaker 1: is it how stuff Works dot com