How Deep Blue Beat Us At Chess – Computers Beating Humans #1

How Deep Blue Beat Us At Chess – Computers Beating Humans #1

That’s it! I’m taking my ball and I’m going home. For all of history, humans have been #1 at
board games, but now there’s a new kid in town. Oh and that kid? He runs on code. Welcome to CompChomp, the only show on the
internets where player 2 has an eidetic memory. So, uh, flipping the table? That is not gonna save you. There was a time, not long ago, so not long
ago that your grandma Mimi probably remembers it, when computers did not play a single game. These were serious machines and they served
serious purposes. (Purposi?) (Porpoises?) They needed to do things that were important,
like calculating missile trajectories and cracking wartime codes. But even in the chaos of WWII, engineers found
stolen moments here and there to teach computers to play games. One of the earliest examples was of the computer
called the Nimatron and it played a math-based game called Nim. The engineers at Westinghouse Electric created
it as part of their showcase for the World’s Fair. 50,000 people played against the computer
during the World’s Fair, and almost none of them won. The engineers actually built in a delay between
when the computer calculated what its next move would be and when it actually made that
move just so people didn’t feel bad every time they lost. So WWII starts winding down, but at the same
time the Cold War is ramping up, and computer scientists on both sides of the Iron Curtain
start dreaming of a day that a computer-based intelligence is going to help them defeat
their enemies. Or, at least, quickly translate all of their
enemy documents. In the beginning of Artificial Intelligence,
the ability for a computer to play a game like chess was seen as proof that computers
were smart. And in the beginning computers sucked at it! Experts say that a really good human player
can see about 10 moves ahead (5 turns for each player) using the eyes of their mind…
their intuition. But computers, they can’t make decisions based
on guts. Cuz umm, they haven’t got any guts! So instead, they have to make their decision
by looking at every possible move they could make and every possible outcome from every
possible move and then deciding which next move was the best one. MATH ALERT!!! You know this is gonna be some good math cause
I’ve got on my math glasses. There are roughly 30 legal moves per chess
position. For one set of turns, that would be 30 * 30,
or 900. For two sets of turns you’d multiply that
by itself: 900 times 900=810,000. Does that seem right? It is. By the time you get to just 5 sets of turns
you are talking over nine quadrillion possibilities. That is a whole lot of possibilities. Your shiny shiny brand new quad core pentium
10000 whatever that you’ve got sitting on your desk right now could not calculate all
of that in the 3 minutes allowed by tournament play, so what do you think that 1950’s PC
behemoth is supposed to do? Forgetaboutit! The humans continued to dominate all the way
until the 1970s. That is the first moment when computer chess
players were able to start beating high ranking human players. One of the ways they improved the computer
programs was by slightly tweaking the algorithm. It’s actually kind of cool. So, before they were trying to look at every
single possibility of every single move, and they just didn’t have the power for that. Instead, they said, you know what?There are
pieces in chess that are more important than other pieces, like the queen. And they started looking at moves that would
put those pieces in a better position and those were the only moves that they would
then like calculate out and choose just the one best result for that. And that was a tiny bit of the improvement. Most of it actually came from raw computing
power. Crunching those numbers! No computer was able to beat a chess grand
master until 1996 when IBM’s Deep Blue faced off against Garry Kasparov and won! One game. Kasparov actually took the tournament. Sorry, computers. A year later a highly upgraded version of
Deep Blue was able to beat Garry Kasparov in an entire tournament, making it the very
first time in all of history that a computer defeated a chess grand master under tournament
conditions. Why?!? That souped-up version of Deep Blue was able
to see 200 million positions per second and calculate 6 to 8 moves ahead. Once again, it was tiny algorithm tweaks and
a whole lot of raw computing power that gave the victory to the computers. Deep Blue retired after that. And by retired I mean we ripped it in two
and sent it to two different museums. Sweet revenge! But it was too late. There were new chess computers popping up
every single year like daisies. By 2006, humans couldn’t even beat the top
computers anymore. We were down, but we were not out. Humanity had a secret weapon, and that weapon
was GO. CHOMP!

15 comments / Add your comment below

  1. Conspiracy alert:
    Garry kasparov accused IBM of cheating. He quoted "It was a wonderful and extremely HUMAN move.".
    He claimed that a human was involve and wanted the source code for the deep blue II.However IBM refused to publish the code.

    Bad move IBM, BAD move.

  2. You should talk about how alpha beta pruning reduces the number of moves a minimax search has to look at.

  3. A computer has now beaten some of the best GO players in the world. I don't think watching a computer playing chess is very interesting. What is more interesting would be to get a computer that can play like a regular person so that matches are fair. Too often either the computer will be set to be too weak to where it is boring to play against or it is too strong to where it is discouraging.

  4. If I recall correctly there is an interesting article around the net stating that the program malfunctioned at some point, making Kasparov flip.
    I found it:

  5. Kaspy played in a style which he doesnt usually play to trick the machine. Now we have to wonder if he treated deep blue like a top 20 GM, could humans have stayed on top longer

Leave a Reply

Your email address will not be published. Required fields are marked *