September 23, 2015 • 38 mins
What is the future of games? Which games have been "solved?" And are there any games humans will always be better at than computers?
Learn more about your ad-choices at https://www.iheartpodcastnetwork.com
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Brought to you by Toyota. Let's go places. Welcome to
Forward Thinking, either everyone, and welcome to Forward Thinking, the
podcast that looks at the future and says, these foolish
games are tearing me apart. I'm Jonathan Strickland, I'm Laura,
(00:22):
and I'm Joe McCormick. Hey, Joe, do you do you
ever play games? Joe? No, In fact, I don't play games.
Joe is extremely serious. Lauren, do you play games? Lauren? Uh? Sometimes, yes, yes.
You know, this is a question that I don't ask computers.
(00:44):
I get scared to ask computers if they play games
because I was brought up in the eighties when the
the amazing film War Games came out. So I gotta
be honest, I've never seen it, actually me neither. Wow. Okay, guys,
prepare yourselves because this movie is amazing. So what now
We're just gonna watch this movie and the podcast booth
(01:06):
and we don't have time for that. I'm just gonna
give you. I'm gonna sum up. So War Games is
specifically the story of a a uh a hacker. A
young hacker played by Matthew Broderick, who has Fred Savage
not Fred said this is pre Fred Savage. Um, before
Fred Savage, we had Matthew Broderick, right, Yeah, Matthew Broderick
(01:27):
was the proto Fred Savage was. So Matthew Broderick is
playing a part of a of a hacker and he
he likes these particular computer games, so he hacks into
a system in order to get a chance to play
some of these games that have not been released, and
one of them, one of them is Tic Tac Toe,
but another one is Global Thermonuclear War. And it turns
(01:53):
out that the AI developed by the computer scientists who
created the games isn't really really sophisticated, and you can
play these simulated global war scenarios. But spoiler alert, the
program gets installed into an actual military facility. It essentially
infects a military facility like a virus and then starts
(02:16):
to take over actual defense systems and prepares to launch
a full attack on the then Soviet Union. So it
becomes the role of Matthew Broderick's character, along with the
original programmer, to figure out a way to convince the computer, hey,
you don't want to do that because it would be
(02:37):
bad for everybody and you don't need to win this game.
Computer you can let it go. They actually teach that
that there is no way to win. In fact, the
computer says, strange game. The only way to win is
not to play. And the way that the computer figures
this out is by using an analogy. It actually starts first.
It starts with the various simulations of nuclear war uh,
(03:00):
and it runs the simulations over and over and over again.
Hasn't launched anything. It's just simulating. And so you're on
the big like Norad screen you see these things playing
out this which is the tic Tac Toe. The reason
it switches to tic tac Toe is because if you
play Tic tac Toe perfectly on both sides, you will
always end in a draw. And so the lesson is
(03:21):
there is no way to win this game. Playing this
game makes no sense because you cannot win it if
you are doing everything correctly. So I just saved you
the the time of actually watching the movie. But it
is a good film, guys, I actually really like it.
All the joy in any Matthew Broderick film is not
the plot. It's it's Matthew Brodick, right. So at that
(03:43):
point the computer learns what we know that did you
take that back kidding. Matthew Broderick is a Broadway superstar.
Um No that that in mutually assured destruction. Once you
enter a conflict, it's game over and nobody wins. It's
(04:03):
kind of right there in the name. Yeah, So comparing
nuclear war to tic tac toe is absurd but also
kind of awesome. That's able to make this abstract connection
between the two. Uh, and tic tac toe is what
we would call a solved game. And we'll talk more
about what solved versus unsolved means a little bit later. Uh,
(04:25):
And we wanted to talk a little bit more about
games and computers and whether or not we humans are
doomed to always come in second place to computers in
the long run. Will computers get to a point where
they will always be better at whatever games we pick
than we are? Well, maybe the stakes are high for
the computers too. I mean, I'm sure that neither of
you guys ever watched the nineties CG Cartoon Show reboot,
(04:49):
because like five people in the world watched the nineties
cartoon show Reboots. I had a friend who is into it,
and I tried to watch it once. It's real, silly,
it's real, real, silly, I never tried to watch it. Okay, okay,
in reboot, you guys. The premise here is that when
a user runs game software on a computer, all of
the citizens of the computer are that are caught in
(05:10):
the game's path must fight for their lives according to
the rules of the game, and losing the game to
the user means decimation or worse than one tenth destruction
for all of the ones and the zeros and the
sprites who make the computer run. So it's kind of
like Tron. Actually, when you play a game, the user
plays the game, but the people inside the computer live
(05:33):
or die. Yeah, in this case, the people fight against
the users, not for the users. Okay, so that's probably
not really what's happening in in reality, but at any rate,
so we wanted to talk about this, and in order
to do so, we wanted to first mention the concept
of game strategy or game theory. So there are volumes
(05:55):
written on game theory, like textbooks of in formation on
logic and game theory. But ultimately, if you really want
to boil it down to its basics, game theory comes
down to using strategy in order to make decisions to
achieve a goal. That's really what game theory comes down to,
once you get rid of everything else. And not all
(06:17):
games actually rely on game theory. Some of them are
more about random chance and less about strategic decisions made
by the players. Yeah, I mean there are slot machines.
I guess that counts as a game. There is no
game theory for slot machines. Yeah. No. You you might
think that there's a system to win at a slot machine.
You are incorrect. Uh, it's all based on timing. And
(06:38):
the timing we're talking about is down to like hundreds
or thousands of a second, so there's no way impact
whatsoever on the outcome of the machine. It's pure chance
if you have pushed the button or pulled the lever
at just that right moment, or something like like candy Land,
where you're just rolling dice and moving pieces. Yeah. Yeah,
candy Land is like a progressive slot machine. Yeah. I
(07:02):
wish I could make a candy line joke, but it's
been a long time since I've played that game. Alright. So,
so games that do involve strategy, you should try it sometimes.
It's pretty sweet, nice, nice, all right. So anyway, games
that are have straightforward strategies include stuff like chess. That's
pretty obvious, right, Checkers or drafts if you're English und
(07:24):
draft Yeah, drafts, I've never heard this. Draft Checkers, that
is the pro that's the proper name for Checkers. Actually, yeah,
the the actual older name for Checkers as drafts. Do
they have cracker barrels in the UK? I hope not.
I doubt it. Yeah, I don't. I've never seen one.
(07:45):
Uh if if you work at a cracker barrel in
the UK, give us a shout. So other games that
also involved strategy connect for Tic Tac toe go is
another one. There are lots and lots of these, and
some of the games have enough strict parameters or rules
(08:06):
in play that make them solvable. Right, And so this
is a concept in in game strategy that there are
some games that are solved in other games that are
not solved. And basically, a solved game is a game
where if you know the starting conditions and players play
at optimal performance, meaning they all make the perfect or
(08:29):
best decision every time. Yeah, you can predict exactly what
the game is going to be. Right, this would be
This would be a ken to having two people sit
down at a chessboard and before anyone's even touched a piece,
one person says mate, and twelve moves and the other
guys like, very well done, and they shake hands they leave.
But of course chess is not solved, No, it is not.
(08:50):
Chess is much too complicated to have solved. You can't
know what all of the possible moves in chess are
and not like our current computing powers, right, Uh, yeah,
I mean there's so many possible series of moves in chess.
I kind of doubt that could ever be calculated. You
are not the only person to doubt that. As for
solved games, there are three broad categories of solved games.
(09:15):
There's ultra weekly solved games, weekly being w E a K.
Not not like every week a game this really really
comes out every week. This would be this is really
really not strong. Uh. You can only predict the outcome
accurately from the initial position, meaning before any player has
(09:36):
made a move. And uh, at that point, it would
just say like, all right, well, assuming that everyone plays optimally.
For example, with Connect four, the first player is always
gonna wine if you always you know, if no matter
if both players are playing perfectly, player one always wins
Connect four. Uh. Weekly solved games include not just the
(09:57):
prediction of the outcome, but also a strategy for its
eiaving it from the initial positions. So essentially saying, here
is the the process through which you will always hit
this particular outcome. Uh. Strongly solved games include strategy to
achieve the best possible outcome from any point of the game,
even parts of the game where someone may have made
(10:18):
a mistake earlier on. UH. So this would be where
if you would look at a game that's already in
progress and say, all right, from this point forward, we're
going to have perfect play on both sides and you
could still predict who was going to win. So you
gave the example of tic tac toe in the movie
War Games. But this is a solved game if players,
If both players are playing perfectly, and there are perfect
(10:41):
ways to play, there the optimal moves you can make
in any given game, then it's always going to end
in a draw. That's exactly right, which is why I
say that's why the computer it comes to the conclusion
that there is no point in playing because there's no
way to win. So winning a tic tac toe do
pens on your opponent making a mistake right exactly. You
(11:04):
also mentioned Connect four. Yes, that one was solved in twice.
It was by two different independent researchers, and it was
weekly solved to show that the first player can always
force a win given perfect play, and in it was
actually strongly solved to the point where at any given
moment within a game, if you took over from that
(11:25):
point forward, you could you could predict who was going
to win based upon that previous positioning. It wouldn't always
be player one in that case, because they may have
made mistakes. So next we have nine Men's Morris. I
don't know what that is. If you played Assassin's Creed
Black Flag, you would because it's a game that exists
within that game and you get achievements for playing it
(11:47):
and winning. So I played a lot of nine Men's Morris.
Nine Men's Morris involves it is a not no, it's
not specifically a pirate game. There's a board game in
which you have spaces where you can place a piece,
and so the first part of the game involves placing
your pieces on the board strategically. The second part of
(12:09):
the game involves moving those pieces along specific pathways that
are available to you and you're trying to line up
three pieces in a line, and if you can line
up three pieces in a line, then you are allowed
to take one of your opponent's pieces off the board. Well,
whereas your opponent is trying to do the same thing
while also preventing you from lining up three three pieces
(12:30):
in a line. So one of the things I've noticed
so far about all of these examples is that they
don't include any element of luck. Yeah, these are all
again very much strategic games. These are games where it
requires the player to make a decision, and then the
player's decision is uh, you know, whatever happens from that.
The consequences of that are all based upon the strategy
(12:52):
of the other player as well. So I'm sure we'll
talk more in later in the episode about games that
are a combination of skill in law. Yes, but but
checkers or or droughts, drafts, Well, we're we're American. We
can say drafts, some of the British douce drafts. But
so you're saying checkers like they play at the cracker barrel?
(13:13):
Was that's a solved game? Solved in two thousand seven.
It had been close to being solved for much longer
than that, but officially solved in two thousand seven, which
was proving that from the initial standpoint standpoint, both players
can play to a draw with optimal play. So so
even if you go second, yeah, you can at least
(13:36):
play to a draw. You may not win, but you
can at least play to a draw. So if you
are player too and you lose, it's your fault. Uh.
It took. It took eighteen years to solve checkers. At
one point they had two hundred computers working on this. Yeah,
that's because there are five hundred billion billion possible arrangements
(13:56):
that could appear on an eight pie eight checkerboard. So
it's real impractical to analyze every single one of those
that this The solving of it was the odyssey of
one Jonathan Schaefer, who's a comp scientist who began with
just sixteen megs of memory on a computer in and
and built out this this solvability issue. Uh it took.
(14:17):
It took really the scaling up of computer power to
make it possible. This is you know this, this is
not simple stuff. Obviously, it's interesting because a lot of
these games are very easy for us to grasp as players.
We understand the rules and the basics and the and
general strategies pretty intuitively. But when it comes to proving
(14:38):
the you know that you have solved the game, that
you know definitively how the outcome will will be based
upon perfect play, that's a lot trickier. Some games are
only partially solved, So chess is partially solved, but only
from an end game standpoint. Yeah. Yeah, So if you say, um,
you know, the only pieces left on the board, or
(15:01):
these five or something, and they're in these positions, there
is a perfect way to play. And as you get
more pieces on the board, it gets way more complicated
because you have more options, more variables, and so I
think when you get up to about seven pieces, it's
really that's really like the limit of how far you
can solve. And not all of those uh permutations are solved,
(15:24):
yet some of them are, because again it depends on
the combination of pieces. Pieces move in different ways depending
upon there that what piece you're talking about, whether it's
upon or a bishop or a rook or whatever. Uh.
And then there's some other ones that have been solved
for smaller versions of the game, like Go Go as
a game, which is really interesting, and that you can
play it on different size boards that determine how many
(15:46):
pieces are in place. So a five by five board
of go has been solved. There's a perfect way to
play it, Yes, but most people play it on something
like a nineteen by nineteen board, which is nowhere close
to being solved. So again, is partially because of just
the amazing complexity of the game at that scale. Sure, Now,
(16:08):
whether or not a game has been solved doesn't necessarily
mean that humans can beat a computer at playing it consistently. Yeah,
as it turns out, computers don't have to have a
solved that they don't have to know quote unquote the
solution in order to still beat the pants off a
human opponent, even a really good human opponent. Yeah. So
(16:29):
chess is not solved, but we've gotten to the point
where computers will always be the best human chess player,
even way back. Yeah, this was the famous case. It's
sort of like the people people likened this to the
story of John Henry and the and the uh steam
engine that was laying down tracks, like the man versus
(16:51):
machine story. But this was the man versus machine story
for the twentieth century. So in u there was actually
a chess rematch between Gary Kasparov and IBM's Deep Blue computer.
And Kasparov and Deep Blue had met in ninety six
for a series of six games, which Kasparov won four
to two. But in the nineties seven match, Deep Blue
(17:14):
one it beat Kasparov. I think it's like three and
a half games to two and a half games, the
haves being I believe draws. So they you know, now
we saw a machine beat a chess master for the
first time. By the way, Kasparov actually demanded a rematch,
he had match. Yeah, he had claimed that there was
some hanky panky going on, and IBM declined his request
(17:38):
because they said they essentially had proven, they had proven
what they set out to prove. And since then these
chess programs have become more powerful and sophisticated. But what
they're actually doing isn't evaluating strategy so much as running
through every single possible option at that time. Give then
(18:00):
the board and the peace position. Yeah, like in encryption
or decryption, I should say, this is called brute force. Yeah,
you're essentially throwing everything you can at the system to
find out what works right. So that's that's really what
brute force is all about. So it might look at
the chess board and say, okay, if I move my
queen here, what would be my opponent's best next move
(18:22):
or what would be all of his or her possible
next ye yeah, yeah. So let's say let's say that
you say, all right, let's move my queen. Uh up,
you know, one square we're moving it, moving the queen
forward one square. What are all the possible responses to
that move? And then what are my responses to those?
And then after figuring all that out, all right, well,
(18:43):
what if we move the queen two squares? What are
all the possible responses and what are my And that's
for every single piece that's in a position that can
be moved. And remember chess has certain rules that also
complicate things like there's the rule about castling. Castling adds
another variable to that sort of stuff. There's also timing.
I mean, you can't have the computer takes seventeen hours
(19:07):
to decide what moved moved to make yea. So now
those roote force with computers that are sufficiently they're sufficiently fast,
they can make these decisions relatively quickly. If we humans
played like this, a chess game would last year entire life.
You would never finish the game because you would be
constantly evaluating all these potential moves before finally settling settling
(19:27):
on the one that's waited to be the strongest. But
that's essentially what chess games and other computer games are doing.
They are looking at all the different possibilities and picking
the one that's the most advantageous or at least advantageous
to a certain degree. Because if you set your difficulty,
because a lot of chess programs allow you to set
the difficulty of the thing, what they'll do is they'll say,
(19:49):
all right, well, we'll just look for x amount of
time and use the best one out of all of that,
rather than evaluate the entire board and all the pieces. Yeah,
and so in this case, the strength of the program
would probably be based on something like how many moves
in advance can it look right? And and how quickly
(20:13):
can it execute that, because if it can't execute it
within enough time, it may have to, you know, cut
back on that. Um So, one thing I thought was
really interesting is that there's a new approach to creating
a chess playing computer that does not rely on brute
force and It comes from Matthew Lay of the Imperial
(20:33):
College of London, and what he did was he developed
a learning algorithm and a neural network, which we've talked
about so much on this show, right, neural networks and
learning algorithms to teach a computer rather than to program
a computer. So he taught a computer chess. He uh.
He created a computer program called it Giraffe and started
(20:54):
to teach it how to play chess. Before he even
got started teaching at chess, he tested it against a
a standardized system that is used to evaluate how well
a computer program does in chess, as a top score
fifteen thousand, so before we had even schooled it, where
it knew the it knew the less. It knew the
(21:14):
rules of chess, but didn't hadn't really learned the value
of various gambits. It scored six thousand, which is not bad.
After he schooled it, and by that I mean he
fed on five million different chess positions into the database,
because you need huge data sets for learning algorithms to work.
(21:34):
Then he set Giraffe against itself. It played itself in
a series of games to start learning which positions were
the most valuable, which one's made you vulnerable, which one's
led to victory or defeat. He then tested it again
and this time it scored uh, like nine thousand, seven hundred,
so much higher. Yeah, yeah, And that still means that
(21:55):
it's not perfect, right, but it's it's at a level
that is consider stint with an international master of chess.
So this is a computer program that learned chess, and
it looks at the entire board and the piece positions
on the board and then evaluates that based upon their
attacks and defense capabilities. So it's kind of the way
(22:15):
we play chess. Yeah, exactly. I think statistically speaking, it
only picks the quote unquote best move about half of
the time. Yeah, a little less than half, like forty
six percent of the time it picks the the quote
unquote best move, but the best move is it tends
to be in the top three choices around seventy percent
of the time. And that's pretty good when you consider
it's not doing brute force, it's not planning all of
(22:37):
these out. It's it's essentially taking a look and then intuitively,
which is a weird thing to say, but that's as
close as I can get feeling out which one is
the best approach, So about half the time it gets
the absolute best choice, and seventy percent of the time
the best choice is within its top three options. But
(22:57):
this doesn't mean that you cannot beat the pants right
off of a computer in some games. Yeah, you can
totally do that in some games. There are some games
that computers still cannot beat the best human players, And
then there are other games where computers are just horrible,
like they can't even beat average players. Yeah, if you're
really stronge, well, yeah, when you when you play a
(23:21):
game where the rules are made up on the fly,
it's very hard for computers. X k c D is
responsible for that particular example. Uh So go A is
a good example, like I was saying before, where you
have a board of like nineteen by nineteen positions that
you can start to play in. A really strong human
player can beat a computer in that case, because computers
(23:43):
just can't deal with all the variables. And uh, it's
still a challenge. I mean, it's still really hard, but
it's possible. Other games don't rely heavily on strategy, so
then the computers are at a loss because the one
thing they do really well may not play as an
important role in those games, and especially if you have
(24:03):
incorporated randomness into your gameplay in some respect, like whether
it's dice or shuffling up cards, that kind of stuff
where your position, however you wanted to fine, that is
not fully determined just by your strategy, but also by luck. Yeah. Yeah,
And lots of games involve a little bit of luck,
(24:24):
like a poker or scrabble. Yeah, poker is a is
a very interesting example to me because you could argue
about that there are different levels of skill involved in poker.
I mean, there are a lot of it is just luck,
it's the cards you're drawing, but skill, the the effect
of skill on poker emerges over the course of playing
(24:46):
many hands. There there are three things I would argue
that go into determining whether or not you're going to
be successful at any given poker game where you're not
not a hand but a game ease of hands. One
is luck the cards that you draw. That's going to
be part of it too. And keep in mind that
(25:08):
you know, bluffing is still a strategy in poker. If
you're effective at bluffing, you may be able to win
even with lousy cards. Two is an understanding of the
statistical probability of what other people are holding in their hands.
So knowing what you should bet and how you should
play based on the cards you have and the probability
of receiving other cards. So if it's texas hold them
(25:28):
and you're looking at the flop, which are the first
three cards that are laid down face up, and you have,
you're looking at the cards in your hand. You can
then start to think, well, those cards are now eliminated
from my opponent's hands. There's no way they can hold
those cards because I know what they are, I see them.
What are the combinations that could beat the cards I
have combined with the flop. That's the sort of stuff
(25:50):
you have to start thinking about, what are the odds
that someone at this table can beat the hand that
I have. Then there's the third part that computers really
can't handle, just getting a read on the playing style
of your opponents, knowing which ones are aggressive, which ones
are timid, which ones might be playing on tilt. While
it's especially difficult because humans are good at playing against expectations,
(26:14):
or good humans are, I mean, so a computer can
look at the what you've done in previous hands and
say well, okay, this player tends to be raising the
bed every time, and so I think that this player
is probably bluffing sometimes. Like you could design a program
that wouldn't make that analysis, but it's hard to design
a program that understands that that other player is trying
(26:38):
to get you to think that that's how he or
she plays, so he or she can then surprise you
when you're vulnerable the good old check raise or something
along those things lines. I remember seeing poker games where
really really strong poker players, could you know what they
would start talking during the tournament, and like as they're
(26:58):
looking at their opponent with and they're deciding what to do,
and they even figure out what cards the person's holding,
like at least one of them, uh, one of them
in particular, I remember him saying, like, you gotta I
guess you've got to have a queen. And I look
at the cards that have been shown, including the ones
that you know held by the other players, that I
(27:20):
know what those cards are, but he doesn't. He hasn't
seen them because he's not watching the camera where it
reveals where all the cards are. And it was amazing
to me because the guy held a queen and it
was phenomenal that he was able to use deductive reasoning
based upon the guy's behavior and the cards that were
already out and the ones that he held in his hand,
and was able to be that accurate. Now, computers could cheat,
(27:43):
they could know that if it was built into the program.
But I gotta tell you, I've played a lot of
games that are that put you against a computer opponent
in poker. They're lousy, they're not consistent, they don't they
don't behave with any logic. I mean, if you're playing
Texas hold Them and you're playing one of those poker games,
half the time they're going for they're going to see
(28:04):
the entire hand, like the flop and the turn the river.
They want to see the whole thing. And they might
have like a two seven, and you're thinking, no human
apart from a couple of crazy people, would ever stay
in this hand that long. They would have folded immediately.
But computer, if you yeah, if you haven't programmed in
(28:26):
the concept of of personal stakes in a game, then
a computer is going to behave radically. It also doesn't
help that in most computer games, you're obviously playing with
fake money, so there's no there's no consequence to losing,
and probably it's not like in Reboot where the computer
losing is going to destroy part of its little interior
computer town. Probably not. Another game that includes both random
(28:49):
chance and strategy would be something like Yachtzi. Pretty simple game, right,
But when you make a role, then you have to
decide what category does that role go into on your sheet?
And every time you put down a category that one is,
it's then inaccessible to you for the rest of that
round of the game. Right, So if I roll three fives,
I could say, all right, well I want this to
(29:10):
go on my fives category, or I might say, no,
I want this to go in my three of a
kind category. Um, so there's a bit of strategy as
well as the random chance. You don't know what your
next role is going to be. Another one I would
think of is Risk, Like there's some amount of strategy,
but there's also a large degree of luck based on
dice rolls. Yeah, here's the way I know how how
(29:30):
the game is gonna go. For Risk, if you ask
me to play, I've already lost. Uh yeah, I would say.
A way of predicting the end of Risk is did
you start playing it will end with people quitting and
rage slipping the table and hating you forever. I just go,
I just go straight to the hate because I I've
only ever played it once and it was one of
those things where because of the way the game went,
(29:53):
I was eliminated before I even had a chance to
do anything. Oh yeah, yeah, things like like Monopoly and
Settle of Catan are other examples of games that are
partial strategy and partial random chance uh largely table flipping.
There are, of course, other games that involve absolutely no strategy,
like Shoots and Ladders or Snakes and Ladders or the
(30:14):
aforementioned candy Land. Yeah. These are games that are more
like distractions than anything else. It's it's you know, you
can kind of get the feeling of hey, I won
or darn I lost, but then eventually you get to
the point where you're like, no, I roll dice and
that's what determined everything wicked. We move one space at
a time, a citizens quote, isn't it all right? And
(30:37):
then there are games that rely heavily on intuition and interpretation.
These are games that are not designed with like cards
or dice. Necessarily, they can incorporate those, but I'm thinking
of stuff like pictionary. Yeah, see, that's great. You can
have definite skill at this game. You can get good
at it. But I'm sure computers are just awful. Well yeah,
(30:57):
because I mean you can you can create abstract representations
of things in these games which humans can understand because
we can have these weird associative things in our brains that, oh,
because you're acting that way, I happen to know that
the thing you're acting out is pulp fiction, you know,
which could could be whether it's, you know, mimicking something
(31:18):
that's going on in the movie and that's how I
know it, or it's because you've acted out the individual words.
However that might be Wait a second, I take that back,
because actually I think a computer could probably destroy people
at pictionary by just Google image searching the term and
then drawing pictures of the results that come up. A
computer is probably better at drawing than I am. I'm
(31:40):
pretty crap at drawing. It would be good at drawing,
It would not necessarily be good at guessing, because it
would then have to look at the abstract nature of
things that we draw and then figure out what concrete
concept is linked to that. Although I'm wondering right now
if something like like shrade, well not shrades, probably because
that would involve full my robot, really lame party. But
(32:02):
it's something like like Google Deep Dream could be successful
at pictionary by kind of reverse engineering, I don't know.
And well, we're seeing some more sophisticated image recognition software
out there where it looks like we could get to
a point where computers could at least make a guess
based upon uh, certain drawings. And there's certain certain shapes
(32:25):
that seem to be more or less universal when you
want to draw something, so like a cat, you know,
you see, besides those who are actually really good at drawing,
for those of us like me who struggle with keeping
within the lines in a coloring book, it's going to
be a circle and a couple of triangles for years.
You know, you know that, oh that's a cat. Um.
There's certain things I think that are almost universal in
(32:47):
that respect, and computers could easily start to recognize those
as well. So the big difference obviously is the role
that rules play and that strategy works within those rules.
If you have very well defined rules and you have
a lot of restrictions on what can actually happen within
the game, and there's not a lot of variability there.
Those are games that are easier to solve and also
(33:09):
easier for computers to totally dominate. Um, the more you
get into variable rules or different styles of play, different
options that you could have during the course of the game. Intelligence,
that's a big one. Humans will do better at those two. Yeah,
And this is partially described the former part more so
(33:33):
by the game theory concept of convergence versus divergence, which
says that games that are more solvable tend to converge
their pieces and possible moves towards the end of the game.
In other words, there are fewer ways to win the
further along in the game that you get. Less solvable
games have more pieces in play and and more possible
(33:53):
moves towards the end, thus making it harder, even with
a very powerful computer, to to list all of the
potential piece positions and moves that can lead to a
winning board layout. So yeah, it's it's what do you
guys think? Do you guys think we're going to eventually
be in a future where no matter what the game is,
apart from you know, kill the computer, the computer is
(34:16):
always going to win. Uh yeah, I think for for
essentially all math based or strictly logic based games that
that involved what would you call them, you know, numerical
quantities or something as the ways of scoring and measuring advantage,
computers are going to dominate. Even even in the ones
where they're not winning now, they will soon be able
(34:38):
to defeat all human experts because it's really just just
a matter of of programming and processing power. Sure well,
I mean I if it comes down to ultimately there
is a best way to do this, computers have the
advantage and that they can evaluate all the other ways. Now,
the ones I don't know about are the ones like,
uh like the games that have social intelligence aspects which
(34:58):
I mentioned, like bluffing in poker, and obviously the randomness
of it also plays a factor in that there can
be quote unquote luck where the computer could end up
with the worst rolls, the worst cards, whatever it may be,
and that is enough to offset its strategic dominance of
the game. Although I do wonder that that even with
(35:19):
with poor luck, because because you can you can have
a poor hand and still when it poker. I wonder
if especially with with digital imaging and recognition. A computer
could be taught to to identify and exploit a person's
tells well, especially better than a human if you could
use a an algorithm similar to the one we talked
(35:39):
about with the sound, where it can actually detect your
heartbeat through your neck. It's looking it's looking at your
pulse through your neck and can tell what it increases,
and thus can map that to how excited you are
about the hand or how much dread you're feeling about
the hand you have. Yeah, you're talking about some some
I mean that's a that's a computer that can really
read in a pope that I'm only kind of joking. Yeah, yeah, no,
(36:03):
I mean that. And like my nude eye dilation temperature changes,
look at it in the in the infrared spectrum laser infrared,
uh voice changes. Sure, yeah, wow, we were I don't
think computers will ever be better than humans at the
most dangerous game hunting humans. I stand to disagree, but
(36:29):
that's a whole documentary series about that. Uh yeah, it's um.
You know, I think there are certain games that are
are inherently human that that I think computers will The
only time computers will really start winning is when humans
and computers. There's no longer a distinction between the two,
where the two have merged at some point and become
one unified thing. And who knows, we might be playing
(36:53):
totally different games at that point. But yeah, this was
a fun thing to think about, just like, what are
the games of the future going to be like, not
just you know, in the video game sphere, but just
games in general. What is play in the future? And um,
I'm sure we're going to see a lot more examples
of people still wanting to test their skills, both against
(37:14):
human and computer opponents. Um, it could be fun, it
could be really frustrating if you're like me and have
an incredible competitive streak paired with an almost comedic inability
to win a game. Um, and as long as I
can beat a computer at Betrayal at the House in
the Hill, I'm happy. So let's just keep that street going, folks.
(37:36):
That's all I'm saying. All right, guys, if you have
any suggestions for future episodes of forward Thinking, I recommend
you write them in and let us know the email
addresses FW thinking at how stuff Works dot com, or
you can drop us a line on Google Plus, Twitter
or Facebook at Google Plus and Twitter. We are FW
thinking over at Facebook. Just search FW thinking, it'll pull
(37:59):
up our profile. You can go in there and you
can leave us a message and we will talk to
you again really soon. For more on this topic and
the future of technology. This is forward sinking dot com,
(38:26):
brought to you by Toyota. Let's Go Places,
Brought to you by Toyota. Let's go places. Welcome to
Forward Thinking, either everyone, and welcome to Forward Thinking, the
podcast that looks at the future and says, these foolish
games are tearing me apart. I'm Jonathan Strickland, I'm Laura,
(00:22):
and I'm Joe McCormick. Hey, Joe, do you do you
ever play games? Joe? No, In fact, I don't play games.
Joe is extremely serious. Lauren, do you play games? Lauren? Uh? Sometimes, yes, yes.
You know, this is a question that I don't ask computers.
(00:44):
I get scared to ask computers if they play games
because I was brought up in the eighties when the
the amazing film War Games came out. So I gotta
be honest, I've never seen it, actually me neither. Wow. Okay, guys,
prepare yourselves because this movie is amazing. So what now
We're just gonna watch this movie and the podcast booth
(01:06):
and we don't have time for that. I'm just gonna
give you. I'm gonna sum up. So War Games is
specifically the story of a a uh a hacker. A
young hacker played by Matthew Broderick, who has Fred Savage
not Fred said this is pre Fred Savage. Um, before
Fred Savage, we had Matthew Broderick, right, Yeah, Matthew Broderick
(01:27):
was the proto Fred Savage was. So Matthew Broderick is
playing a part of a of a hacker and he
he likes these particular computer games, so he hacks into
a system in order to get a chance to play
some of these games that have not been released, and
one of them, one of them is Tic Tac Toe,
but another one is Global Thermonuclear War. And it turns
(01:53):
out that the AI developed by the computer scientists who
created the games isn't really really sophisticated, and you can
play these simulated global war scenarios. But spoiler alert, the
program gets installed into an actual military facility. It essentially
infects a military facility like a virus and then starts
(02:16):
to take over actual defense systems and prepares to launch
a full attack on the then Soviet Union. So it
becomes the role of Matthew Broderick's character, along with the
original programmer, to figure out a way to convince the computer, hey,
you don't want to do that because it would be
(02:37):
bad for everybody and you don't need to win this game.
Computer you can let it go. They actually teach that
that there is no way to win. In fact, the
computer says, strange game. The only way to win is
not to play. And the way that the computer figures
this out is by using an analogy. It actually starts first.
It starts with the various simulations of nuclear war uh,
(03:00):
and it runs the simulations over and over and over again.
Hasn't launched anything. It's just simulating. And so you're on
the big like Norad screen you see these things playing
out this which is the tic Tac Toe. The reason
it switches to tic tac Toe is because if you
play Tic tac Toe perfectly on both sides, you will
always end in a draw. And so the lesson is
(03:21):
there is no way to win this game. Playing this
game makes no sense because you cannot win it if
you are doing everything correctly. So I just saved you
the the time of actually watching the movie. But it
is a good film, guys, I actually really like it.
All the joy in any Matthew Broderick film is not
the plot. It's it's Matthew Brodick, right. So at that
(03:43):
point the computer learns what we know that did you
take that back kidding. Matthew Broderick is a Broadway superstar.
Um No that that in mutually assured destruction. Once you
enter a conflict, it's game over and nobody wins. It's
(04:03):
kind of right there in the name. Yeah, So comparing
nuclear war to tic tac toe is absurd but also
kind of awesome. That's able to make this abstract connection
between the two. Uh, and tic tac toe is what
we would call a solved game. And we'll talk more
about what solved versus unsolved means a little bit later. Uh,
(04:25):
And we wanted to talk a little bit more about
games and computers and whether or not we humans are
doomed to always come in second place to computers in
the long run. Will computers get to a point where
they will always be better at whatever games we pick
than we are? Well, maybe the stakes are high for
the computers too. I mean, I'm sure that neither of
you guys ever watched the nineties CG Cartoon Show reboot,
(04:49):
because like five people in the world watched the nineties
cartoon show Reboots. I had a friend who is into it,
and I tried to watch it once. It's real, silly,
it's real, real, silly, I never tried to watch it. Okay, okay,
in reboot, you guys. The premise here is that when
a user runs game software on a computer, all of
the citizens of the computer are that are caught in
(05:10):
the game's path must fight for their lives according to
the rules of the game, and losing the game to
the user means decimation or worse than one tenth destruction
for all of the ones and the zeros and the
sprites who make the computer run. So it's kind of
like Tron. Actually, when you play a game, the user
plays the game, but the people inside the computer live
(05:33):
or die. Yeah, in this case, the people fight against
the users, not for the users. Okay, so that's probably
not really what's happening in in reality, but at any rate,
so we wanted to talk about this, and in order
to do so, we wanted to first mention the concept
of game strategy or game theory. So there are volumes
(05:55):
written on game theory, like textbooks of in formation on
logic and game theory. But ultimately, if you really want
to boil it down to its basics, game theory comes
down to using strategy in order to make decisions to
achieve a goal. That's really what game theory comes down to,
once you get rid of everything else. And not all
(06:17):
games actually rely on game theory. Some of them are
more about random chance and less about strategic decisions made
by the players. Yeah, I mean there are slot machines.
I guess that counts as a game. There is no
game theory for slot machines. Yeah. No. You you might
think that there's a system to win at a slot machine.
You are incorrect. Uh, it's all based on timing. And
(06:38):
the timing we're talking about is down to like hundreds
or thousands of a second, so there's no way impact
whatsoever on the outcome of the machine. It's pure chance
if you have pushed the button or pulled the lever
at just that right moment, or something like like candy Land,
where you're just rolling dice and moving pieces. Yeah. Yeah,
candy Land is like a progressive slot machine. Yeah. I
(07:02):
wish I could make a candy line joke, but it's
been a long time since I've played that game. Alright. So,
so games that do involve strategy, you should try it sometimes.
It's pretty sweet, nice, nice, all right. So anyway, games
that are have straightforward strategies include stuff like chess. That's
pretty obvious, right, Checkers or drafts if you're English und
(07:24):
draft Yeah, drafts, I've never heard this. Draft Checkers, that
is the pro that's the proper name for Checkers. Actually, yeah,
the the actual older name for Checkers as drafts. Do
they have cracker barrels in the UK? I hope not.
I doubt it. Yeah, I don't. I've never seen one.
(07:45):
Uh if if you work at a cracker barrel in
the UK, give us a shout. So other games that
also involved strategy connect for Tic Tac toe go is
another one. There are lots and lots of these, and
some of the games have enough strict parameters or rules
(08:06):
in play that make them solvable. Right, And so this
is a concept in in game strategy that there are
some games that are solved in other games that are
not solved. And basically, a solved game is a game
where if you know the starting conditions and players play
at optimal performance, meaning they all make the perfect or
(08:29):
best decision every time. Yeah, you can predict exactly what
the game is going to be. Right, this would be
This would be a ken to having two people sit
down at a chessboard and before anyone's even touched a piece,
one person says mate, and twelve moves and the other
guys like, very well done, and they shake hands they leave.
But of course chess is not solved, No, it is not.
(08:50):
Chess is much too complicated to have solved. You can't
know what all of the possible moves in chess are
and not like our current computing powers, right, Uh, yeah,
I mean there's so many possible series of moves in chess.
I kind of doubt that could ever be calculated. You
are not the only person to doubt that. As for
solved games, there are three broad categories of solved games.
(09:15):
There's ultra weekly solved games, weekly being w E a K.
Not not like every week a game this really really
comes out every week. This would be this is really
really not strong. Uh. You can only predict the outcome
accurately from the initial position, meaning before any player has
(09:36):
made a move. And uh, at that point, it would
just say like, all right, well, assuming that everyone plays optimally.
For example, with Connect four, the first player is always
gonna wine if you always you know, if no matter
if both players are playing perfectly, player one always wins
Connect four. Uh. Weekly solved games include not just the
(09:57):
prediction of the outcome, but also a strategy for its
eiaving it from the initial positions. So essentially saying, here
is the the process through which you will always hit
this particular outcome. Uh. Strongly solved games include strategy to
achieve the best possible outcome from any point of the game,
even parts of the game where someone may have made
(10:18):
a mistake earlier on. UH. So this would be where
if you would look at a game that's already in
progress and say, all right, from this point forward, we're
going to have perfect play on both sides and you
could still predict who was going to win. So you
gave the example of tic tac toe in the movie
War Games. But this is a solved game if players,
If both players are playing perfectly, and there are perfect
(10:41):
ways to play, there the optimal moves you can make
in any given game, then it's always going to end
in a draw. That's exactly right, which is why I
say that's why the computer it comes to the conclusion
that there is no point in playing because there's no
way to win. So winning a tic tac toe do
pens on your opponent making a mistake right exactly. You
(11:04):
also mentioned Connect four. Yes, that one was solved in twice.
It was by two different independent researchers, and it was
weekly solved to show that the first player can always
force a win given perfect play, and in it was
actually strongly solved to the point where at any given
moment within a game, if you took over from that
(11:25):
point forward, you could you could predict who was going
to win based upon that previous positioning. It wouldn't always
be player one in that case, because they may have
made mistakes. So next we have nine Men's Morris. I
don't know what that is. If you played Assassin's Creed
Black Flag, you would because it's a game that exists
within that game and you get achievements for playing it
(11:47):
and winning. So I played a lot of nine Men's Morris.
Nine Men's Morris involves it is a not no, it's
not specifically a pirate game. There's a board game in
which you have spaces where you can place a piece,
and so the first part of the game involves placing
your pieces on the board strategically. The second part of
(12:09):
the game involves moving those pieces along specific pathways that
are available to you and you're trying to line up
three pieces in a line, and if you can line
up three pieces in a line, then you are allowed
to take one of your opponent's pieces off the board. Well,
whereas your opponent is trying to do the same thing
while also preventing you from lining up three three pieces
(12:30):
in a line. So one of the things I've noticed
so far about all of these examples is that they
don't include any element of luck. Yeah, these are all
again very much strategic games. These are games where it
requires the player to make a decision, and then the
player's decision is uh, you know, whatever happens from that.
The consequences of that are all based upon the strategy
(12:52):
of the other player as well. So I'm sure we'll
talk more in later in the episode about games that
are a combination of skill in law. Yes, but but
checkers or or droughts, drafts, Well, we're we're American. We
can say drafts, some of the British douce drafts. But
so you're saying checkers like they play at the cracker barrel?
(13:13):
Was that's a solved game? Solved in two thousand seven.
It had been close to being solved for much longer
than that, but officially solved in two thousand seven, which
was proving that from the initial standpoint standpoint, both players
can play to a draw with optimal play. So so
even if you go second, yeah, you can at least
(13:36):
play to a draw. You may not win, but you
can at least play to a draw. So if you
are player too and you lose, it's your fault. Uh.
It took. It took eighteen years to solve checkers. At
one point they had two hundred computers working on this. Yeah,
that's because there are five hundred billion billion possible arrangements
(13:56):
that could appear on an eight pie eight checkerboard. So
it's real impractical to analyze every single one of those
that this The solving of it was the odyssey of
one Jonathan Schaefer, who's a comp scientist who began with
just sixteen megs of memory on a computer in and
and built out this this solvability issue. Uh it took.
(14:17):
It took really the scaling up of computer power to
make it possible. This is you know this, this is
not simple stuff. Obviously, it's interesting because a lot of
these games are very easy for us to grasp as players.
We understand the rules and the basics and the and
general strategies pretty intuitively. But when it comes to proving
(14:38):
the you know that you have solved the game, that
you know definitively how the outcome will will be based
upon perfect play, that's a lot trickier. Some games are
only partially solved, So chess is partially solved, but only
from an end game standpoint. Yeah. Yeah, So if you say, um,
you know, the only pieces left on the board, or
(15:01):
these five or something, and they're in these positions, there
is a perfect way to play. And as you get
more pieces on the board, it gets way more complicated
because you have more options, more variables, and so I
think when you get up to about seven pieces, it's
really that's really like the limit of how far you
can solve. And not all of those uh permutations are solved,
(15:24):
yet some of them are, because again it depends on
the combination of pieces. Pieces move in different ways depending
upon there that what piece you're talking about, whether it's
upon or a bishop or a rook or whatever. Uh.
And then there's some other ones that have been solved
for smaller versions of the game, like Go Go as
a game, which is really interesting, and that you can
play it on different size boards that determine how many
(15:46):
pieces are in place. So a five by five board
of go has been solved. There's a perfect way to
play it, Yes, but most people play it on something
like a nineteen by nineteen board, which is nowhere close
to being solved. So again, is partially because of just
the amazing complexity of the game at that scale. Sure, Now,
(16:08):
whether or not a game has been solved doesn't necessarily
mean that humans can beat a computer at playing it consistently. Yeah,
as it turns out, computers don't have to have a
solved that they don't have to know quote unquote the
solution in order to still beat the pants off a
human opponent, even a really good human opponent. Yeah. So
(16:29):
chess is not solved, but we've gotten to the point
where computers will always be the best human chess player,
even way back. Yeah, this was the famous case. It's
sort of like the people people likened this to the
story of John Henry and the and the uh steam
engine that was laying down tracks, like the man versus
(16:51):
machine story. But this was the man versus machine story
for the twentieth century. So in u there was actually
a chess rematch between Gary Kasparov and IBM's Deep Blue computer.
And Kasparov and Deep Blue had met in ninety six
for a series of six games, which Kasparov won four
to two. But in the nineties seven match, Deep Blue
(17:14):
one it beat Kasparov. I think it's like three and
a half games to two and a half games, the
haves being I believe draws. So they you know, now
we saw a machine beat a chess master for the
first time. By the way, Kasparov actually demanded a rematch,
he had match. Yeah, he had claimed that there was
some hanky panky going on, and IBM declined his request
(17:38):
because they said they essentially had proven, they had proven
what they set out to prove. And since then these
chess programs have become more powerful and sophisticated. But what
they're actually doing isn't evaluating strategy so much as running
through every single possible option at that time. Give then
(18:00):
the board and the peace position. Yeah, like in encryption
or decryption, I should say, this is called brute force. Yeah,
you're essentially throwing everything you can at the system to
find out what works right. So that's that's really what
brute force is all about. So it might look at
the chess board and say, okay, if I move my
queen here, what would be my opponent's best next move
(18:22):
or what would be all of his or her possible
next ye yeah, yeah. So let's say let's say that
you say, all right, let's move my queen. Uh up,
you know, one square we're moving it, moving the queen
forward one square. What are all the possible responses to
that move? And then what are my responses to those?
And then after figuring all that out, all right, well,
(18:43):
what if we move the queen two squares? What are
all the possible responses and what are my And that's
for every single piece that's in a position that can
be moved. And remember chess has certain rules that also
complicate things like there's the rule about castling. Castling adds
another variable to that sort of stuff. There's also timing.
I mean, you can't have the computer takes seventeen hours
(19:07):
to decide what moved moved to make yea. So now
those roote force with computers that are sufficiently they're sufficiently fast,
they can make these decisions relatively quickly. If we humans
played like this, a chess game would last year entire life.
You would never finish the game because you would be
constantly evaluating all these potential moves before finally settling settling
(19:27):
on the one that's waited to be the strongest. But
that's essentially what chess games and other computer games are doing.
They are looking at all the different possibilities and picking
the one that's the most advantageous or at least advantageous
to a certain degree. Because if you set your difficulty,
because a lot of chess programs allow you to set
the difficulty of the thing, what they'll do is they'll say,
(19:49):
all right, well, we'll just look for x amount of
time and use the best one out of all of that,
rather than evaluate the entire board and all the pieces. Yeah,
and so in this case, the strength of the program
would probably be based on something like how many moves
in advance can it look right? And and how quickly
(20:13):
can it execute that, because if it can't execute it
within enough time, it may have to, you know, cut
back on that. Um So, one thing I thought was
really interesting is that there's a new approach to creating
a chess playing computer that does not rely on brute
force and It comes from Matthew Lay of the Imperial
(20:33):
College of London, and what he did was he developed
a learning algorithm and a neural network, which we've talked
about so much on this show, right, neural networks and
learning algorithms to teach a computer rather than to program
a computer. So he taught a computer chess. He uh.
He created a computer program called it Giraffe and started
(20:54):
to teach it how to play chess. Before he even
got started teaching at chess, he tested it against a
a standardized system that is used to evaluate how well
a computer program does in chess, as a top score
fifteen thousand, so before we had even schooled it, where
it knew the it knew the less. It knew the
(21:14):
rules of chess, but didn't hadn't really learned the value
of various gambits. It scored six thousand, which is not bad.
After he schooled it, and by that I mean he
fed on five million different chess positions into the database,
because you need huge data sets for learning algorithms to work.
(21:34):
Then he set Giraffe against itself. It played itself in
a series of games to start learning which positions were
the most valuable, which one's made you vulnerable, which one's
led to victory or defeat. He then tested it again
and this time it scored uh, like nine thousand, seven hundred,
so much higher. Yeah, yeah, And that still means that
(21:55):
it's not perfect, right, but it's it's at a level
that is consider stint with an international master of chess.
So this is a computer program that learned chess, and
it looks at the entire board and the piece positions
on the board and then evaluates that based upon their
attacks and defense capabilities. So it's kind of the way
(22:15):
we play chess. Yeah, exactly. I think statistically speaking, it
only picks the quote unquote best move about half of
the time. Yeah, a little less than half, like forty
six percent of the time it picks the the quote
unquote best move, but the best move is it tends
to be in the top three choices around seventy percent
of the time. And that's pretty good when you consider
it's not doing brute force, it's not planning all of
(22:37):
these out. It's it's essentially taking a look and then intuitively,
which is a weird thing to say, but that's as
close as I can get feeling out which one is
the best approach, So about half the time it gets
the absolute best choice, and seventy percent of the time
the best choice is within its top three options. But
(22:57):
this doesn't mean that you cannot beat the pants right
off of a computer in some games. Yeah, you can
totally do that in some games. There are some games
that computers still cannot beat the best human players, And
then there are other games where computers are just horrible,
like they can't even beat average players. Yeah, if you're
really stronge, well, yeah, when you when you play a
(23:21):
game where the rules are made up on the fly,
it's very hard for computers. X k c D is
responsible for that particular example. Uh So go A is
a good example, like I was saying before, where you
have a board of like nineteen by nineteen positions that
you can start to play in. A really strong human
player can beat a computer in that case, because computers
(23:43):
just can't deal with all the variables. And uh, it's
still a challenge. I mean, it's still really hard, but
it's possible. Other games don't rely heavily on strategy, so
then the computers are at a loss because the one
thing they do really well may not play as an
important role in those games, and especially if you have
(24:03):
incorporated randomness into your gameplay in some respect, like whether
it's dice or shuffling up cards, that kind of stuff
where your position, however you wanted to fine, that is
not fully determined just by your strategy, but also by luck. Yeah. Yeah,
And lots of games involve a little bit of luck,
(24:24):
like a poker or scrabble. Yeah, poker is a is
a very interesting example to me because you could argue
about that there are different levels of skill involved in poker.
I mean, there are a lot of it is just luck,
it's the cards you're drawing, but skill, the the effect
of skill on poker emerges over the course of playing
(24:46):
many hands. There there are three things I would argue
that go into determining whether or not you're going to
be successful at any given poker game where you're not
not a hand but a game ease of hands. One
is luck the cards that you draw. That's going to
be part of it too. And keep in mind that
(25:08):
you know, bluffing is still a strategy in poker. If
you're effective at bluffing, you may be able to win
even with lousy cards. Two is an understanding of the
statistical probability of what other people are holding in their hands.
So knowing what you should bet and how you should
play based on the cards you have and the probability
of receiving other cards. So if it's texas hold them
(25:28):
and you're looking at the flop, which are the first
three cards that are laid down face up, and you have,
you're looking at the cards in your hand. You can
then start to think, well, those cards are now eliminated
from my opponent's hands. There's no way they can hold
those cards because I know what they are, I see them.
What are the combinations that could beat the cards I
have combined with the flop. That's the sort of stuff
(25:50):
you have to start thinking about, what are the odds
that someone at this table can beat the hand that
I have. Then there's the third part that computers really
can't handle, just getting a read on the playing style
of your opponents, knowing which ones are aggressive, which ones
are timid, which ones might be playing on tilt. While
it's especially difficult because humans are good at playing against expectations,
(26:14):
or good humans are, I mean, so a computer can
look at the what you've done in previous hands and
say well, okay, this player tends to be raising the
bed every time, and so I think that this player
is probably bluffing sometimes. Like you could design a program
that wouldn't make that analysis, but it's hard to design
a program that understands that that other player is trying
(26:38):
to get you to think that that's how he or
she plays, so he or she can then surprise you
when you're vulnerable the good old check raise or something
along those things lines. I remember seeing poker games where
really really strong poker players, could you know what they
would start talking during the tournament, and like as they're
(26:58):
looking at their opponent with and they're deciding what to do,
and they even figure out what cards the person's holding,
like at least one of them, uh, one of them
in particular, I remember him saying, like, you gotta I
guess you've got to have a queen. And I look
at the cards that have been shown, including the ones
that you know held by the other players, that I
(27:20):
know what those cards are, but he doesn't. He hasn't
seen them because he's not watching the camera where it
reveals where all the cards are. And it was amazing
to me because the guy held a queen and it
was phenomenal that he was able to use deductive reasoning
based upon the guy's behavior and the cards that were
already out and the ones that he held in his hand,
and was able to be that accurate. Now, computers could cheat,
(27:43):
they could know that if it was built into the program.
But I gotta tell you, I've played a lot of
games that are that put you against a computer opponent
in poker. They're lousy, they're not consistent, they don't they
don't behave with any logic. I mean, if you're playing
Texas hold Them and you're playing one of those poker games,
half the time they're going for they're going to see
(28:04):
the entire hand, like the flop and the turn the river.
They want to see the whole thing. And they might
have like a two seven, and you're thinking, no human
apart from a couple of crazy people, would ever stay
in this hand that long. They would have folded immediately.
But computer, if you yeah, if you haven't programmed in
(28:26):
the concept of of personal stakes in a game, then
a computer is going to behave radically. It also doesn't
help that in most computer games, you're obviously playing with
fake money, so there's no there's no consequence to losing,
and probably it's not like in Reboot where the computer
losing is going to destroy part of its little interior
computer town. Probably not. Another game that includes both random
(28:49):
chance and strategy would be something like Yachtzi. Pretty simple game, right,
But when you make a role, then you have to
decide what category does that role go into on your sheet?
And every time you put down a category that one is,
it's then inaccessible to you for the rest of that
round of the game. Right, So if I roll three fives,
I could say, all right, well I want this to
(29:10):
go on my fives category, or I might say, no,
I want this to go in my three of a
kind category. Um, so there's a bit of strategy as
well as the random chance. You don't know what your
next role is going to be. Another one I would
think of is Risk, Like there's some amount of strategy,
but there's also a large degree of luck based on
dice rolls. Yeah, here's the way I know how how
(29:30):
the game is gonna go. For Risk, if you ask
me to play, I've already lost. Uh yeah, I would say.
A way of predicting the end of Risk is did
you start playing it will end with people quitting and
rage slipping the table and hating you forever. I just go,
I just go straight to the hate because I I've
only ever played it once and it was one of
those things where because of the way the game went,
(29:53):
I was eliminated before I even had a chance to
do anything. Oh yeah, yeah, things like like Monopoly and
Settle of Catan are other examples of games that are
partial strategy and partial random chance uh largely table flipping.
There are, of course, other games that involve absolutely no strategy,
like Shoots and Ladders or Snakes and Ladders or the
(30:14):
aforementioned candy Land. Yeah. These are games that are more
like distractions than anything else. It's it's you know, you
can kind of get the feeling of hey, I won
or darn I lost, but then eventually you get to
the point where you're like, no, I roll dice and
that's what determined everything wicked. We move one space at
a time, a citizens quote, isn't it all right? And
(30:37):
then there are games that rely heavily on intuition and interpretation.
These are games that are not designed with like cards
or dice. Necessarily, they can incorporate those, but I'm thinking
of stuff like pictionary. Yeah, see, that's great. You can
have definite skill at this game. You can get good
at it. But I'm sure computers are just awful. Well yeah,
(30:57):
because I mean you can you can create abstract representations
of things in these games which humans can understand because
we can have these weird associative things in our brains that, oh,
because you're acting that way, I happen to know that
the thing you're acting out is pulp fiction, you know,
which could could be whether it's, you know, mimicking something
(31:18):
that's going on in the movie and that's how I
know it, or it's because you've acted out the individual words.
However that might be Wait a second, I take that back,
because actually I think a computer could probably destroy people
at pictionary by just Google image searching the term and
then drawing pictures of the results that come up. A
computer is probably better at drawing than I am. I'm
(31:40):
pretty crap at drawing. It would be good at drawing,
It would not necessarily be good at guessing, because it
would then have to look at the abstract nature of
things that we draw and then figure out what concrete
concept is linked to that. Although I'm wondering right now
if something like like shrade, well not shrades, probably because
that would involve full my robot, really lame party. But
(32:02):
it's something like like Google Deep Dream could be successful
at pictionary by kind of reverse engineering, I don't know.
And well, we're seeing some more sophisticated image recognition software
out there where it looks like we could get to
a point where computers could at least make a guess
based upon uh, certain drawings. And there's certain certain shapes
(32:25):
that seem to be more or less universal when you
want to draw something, so like a cat, you know,
you see, besides those who are actually really good at drawing,
for those of us like me who struggle with keeping
within the lines in a coloring book, it's going to
be a circle and a couple of triangles for years.
You know, you know that, oh that's a cat. Um.
There's certain things I think that are almost universal in
(32:47):
that respect, and computers could easily start to recognize those
as well. So the big difference obviously is the role
that rules play and that strategy works within those rules.
If you have very well defined rules and you have
a lot of restrictions on what can actually happen within
the game, and there's not a lot of variability there.
Those are games that are easier to solve and also
(33:09):
easier for computers to totally dominate. Um, the more you
get into variable rules or different styles of play, different
options that you could have during the course of the game. Intelligence,
that's a big one. Humans will do better at those two. Yeah,
And this is partially described the former part more so
(33:33):
by the game theory concept of convergence versus divergence, which
says that games that are more solvable tend to converge
their pieces and possible moves towards the end of the game.
In other words, there are fewer ways to win the
further along in the game that you get. Less solvable
games have more pieces in play and and more possible
(33:53):
moves towards the end, thus making it harder, even with
a very powerful computer, to to list all of the
potential piece positions and moves that can lead to a
winning board layout. So yeah, it's it's what do you
guys think? Do you guys think we're going to eventually
be in a future where no matter what the game is,
apart from you know, kill the computer, the computer is
(34:16):
always going to win. Uh yeah, I think for for
essentially all math based or strictly logic based games that
that involved what would you call them, you know, numerical
quantities or something as the ways of scoring and measuring advantage,
computers are going to dominate. Even even in the ones
where they're not winning now, they will soon be able
(34:38):
to defeat all human experts because it's really just just
a matter of of programming and processing power. Sure well,
I mean I if it comes down to ultimately there
is a best way to do this, computers have the
advantage and that they can evaluate all the other ways. Now,
the ones I don't know about are the ones like,
uh like the games that have social intelligence aspects which
(34:58):
I mentioned, like bluffing in poker, and obviously the randomness
of it also plays a factor in that there can
be quote unquote luck where the computer could end up
with the worst rolls, the worst cards, whatever it may be,
and that is enough to offset its strategic dominance of
the game. Although I do wonder that that even with
(35:19):
with poor luck, because because you can you can have
a poor hand and still when it poker. I wonder
if especially with with digital imaging and recognition. A computer
could be taught to to identify and exploit a person's
tells well, especially better than a human if you could
use a an algorithm similar to the one we talked
(35:39):
about with the sound, where it can actually detect your
heartbeat through your neck. It's looking it's looking at your
pulse through your neck and can tell what it increases,
and thus can map that to how excited you are
about the hand or how much dread you're feeling about
the hand you have. Yeah, you're talking about some some
I mean that's a that's a computer that can really
read in a pope that I'm only kind of joking. Yeah, yeah, no,
(36:03):
I mean that. And like my nude eye dilation temperature changes,
look at it in the in the infrared spectrum laser infrared,
uh voice changes. Sure, yeah, wow, we were I don't
think computers will ever be better than humans at the
most dangerous game hunting humans. I stand to disagree, but
(36:29):
that's a whole documentary series about that. Uh yeah, it's um.
You know, I think there are certain games that are
are inherently human that that I think computers will The
only time computers will really start winning is when humans
and computers. There's no longer a distinction between the two,
where the two have merged at some point and become
one unified thing. And who knows, we might be playing
(36:53):
totally different games at that point. But yeah, this was
a fun thing to think about, just like, what are
the games of the future going to be like, not
just you know, in the video game sphere, but just
games in general. What is play in the future? And um,
I'm sure we're going to see a lot more examples
of people still wanting to test their skills, both against
(37:14):
human and computer opponents. Um, it could be fun, it
could be really frustrating if you're like me and have
an incredible competitive streak paired with an almost comedic inability
to win a game. Um, and as long as I
can beat a computer at Betrayal at the House in
the Hill, I'm happy. So let's just keep that street going, folks.
(37:36):
That's all I'm saying. All right, guys, if you have
any suggestions for future episodes of forward Thinking, I recommend
you write them in and let us know the email
addresses FW thinking at how stuff Works dot com, or
you can drop us a line on Google Plus, Twitter
or Facebook at Google Plus and Twitter. We are FW
thinking over at Facebook. Just search FW thinking, it'll pull
(37:59):
up our profile. You can go in there and you
can leave us a message and we will talk to
you again really soon. For more on this topic and
the future of technology. This is forward sinking dot com,
(38:26):
brought to you by Toyota. Let's Go Places,
Fw:Thinking News
Hosts And Creators
Jonathan Strickland
Joe McCormick
Lauren Vogelbaum
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.