Computer defeats women's shogi champion

October 12, 2010 12:21 PM

Four different shogi-playing software programs combined forces to "aggressively pursue" and defeat female champion Ichiyo Shimizu in 86 moves. (previously)
posted by Joe Beese (25 comments total)

"The shogi association says it will conduct an in-depth analysis of the match before it decides whether to allow the software to challenge a higher-ranking male professional player."

(Emphasis added.)
posted by LSK at 12:26 PM on October 12, 2010

That hasn't happened yet, but after all, western chess is a relatively simple game, with only about 10^123 possible games existing that can be played out. Shogi is a bit more complex, though, offering about 10^224 possible games.

10^224 is 'a bit' more than 10^123 in the same way that the observable universe is 'a bit' larger than an electron.
posted by empath at 12:34 PM on October 12, 2010 [20 favorites]

So it takes four of them to gang up and take down a girl? That's really gentlemanly of them.
posted by Old'n'Busted at 12:38 PM on October 12, 2010

THE MACHINES HAVE NO GENDER
posted by I Foody at 12:41 PM on October 12, 2010

THE MACHINES HAVE NO GENDER

Doesn't matter, they're still going to screw us in the end.
posted by nomadicink at 12:51 PM on October 12, 2010

computers are better then girls, i knew it! i lived my life for this moment!
posted by Mach5 at 12:53 PM on October 12, 2010 [6 favorites]

As much as I'd like the invincibility of the human intellect, I think we should just be happy at how much computing power is necessary to win these games. It's no shame to say, "I lost to a machine that would normally used to calculate weather patterns".
posted by yeloson at 12:55 PM on October 12, 2010

10^224 is 'a bit' more than 10^123 in the same way that the observable universe is 'a bit' larger than an electron.

What's incredible is even that is still really far off. The observable universe is 'only' about 3 * 10^41 electron-widths in diameter. Even if you compared it to the Planck length, it's 5.4 * 10^61 Planck lengths wide. To bring it up to the same ratio as Shogi games:chess games you'd have to compare an electron to something 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 universes wide.

It's roughly like if every atom in our galaxy were an entire universe and you laid those end-to-end, then compared that to an electron. Or something. Combinatorics has a way of producing numbers that are completely outside of the natural world.
posted by jedicus at 1:06 PM on October 12, 2010 [8 favorites]

Fortunately, most of those games suck.
posted by Malor at 1:18 PM on October 12, 2010

That hasn't happened yet, but after all, western chess is a relatively simple game, with only about 10^123 possible games existing that can be played out. Shogi is a bit more complex, though, offering about 10^224 possible games.

From what I know of Shogi, this definition of complexity, sheer combinatorial abundance, is false. There are many more possible moves at each point of the game, but a larger proportion of them are meaningless or bad choices. That's why Shogi is challenging masters without IBM having to bring big iron to bear to do it.
posted by JHarris at 1:19 PM on October 12, 2010 [1 favorite]

Er, why computer Shogi is challenging masters. Is there a neurological condition that causes one to frequently drop important words?
posted by JHarris at 1:21 PM on October 12, 2010

"The shogi association says it will conduct an in-depth analysis of the match before it decides whether to allow the software to challenge a higher-ranking male professional player."

(Emphasis added.)

I don't know what the emphasis is supposed to imply. I don't think there are any higher ranking female players - Ichiyo is the highest ranked woman.
posted by mrgoldenbrown at 1:32 PM on October 12, 2010

Is there a neurological condition that causes one to frequently drop important words?

Yes, it's called.
posted by "Elbows" O'Donoghue at 1:44 PM on October 12, 2010 [16 favorites]

Dai_shogi looks like even more fun!
posted by Joe Chip at 1:45 PM on October 12, 2010

Combinatorics has a way of producing numbers that are completely outside of the natural world.

Ramsey theory really is the best for that. Graham's Number is so big regular exponential notation can't describe it. If you use 3 as the base, the number of levels needed to write out Graham's Number in exponential tower notation dwarfs the possible number of particles in the observable universe.

Graham's Number is a weak upper bound on a problem in Ramsey Theory. (The actual upper bound is much smaller but everything above still applies.) The current best lower bound? 11.
posted by kmz at 1:49 PM on October 12, 2010 [2 favorites]

Combinatorics has a way of producing numbers that are completely outside of the natural world.

A way that I would describe briefly as "multiplication".
posted by Wolfdog at 1:59 PM on October 12, 2010 [3 favorites]

As regards the difference between 10^123 and 10^224:

There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers.
        -Richard Feynman, physicist, Nobel laureate (1918-1988)

But maybe we should instead call them computational? Anyhow, this discussion is a reasonable place to link to a thing I wrote about visualizing large numbers, which concludes that the only way to be accurate is with probabilities, but the only way to impress people is by underestimating and just using "earths weight in carbon molecules" and the like. Ramsey theory has famously terrifying big numbers, but it feels fake somehow - these large numbers end up getting used not because of necessity, but because our proof techniques are weak. Whereas the number of chess games or card shuffles or whatever is actually something that is inherent to those systems, and those systems are of mild interest to non-specialists.
posted by pmb at 2:40 PM on October 12, 2010

I love Shogi, but this is hardly big news. Call me when a machine beats a human playing Go. It will never happen.
posted by KokuRyu at 3:35 PM on October 12, 2010

Is Shogi competition normally segregated by gender? Are they all just mentioning gender as a handy subset of which she is the top human player?
posted by scottreynen at 5:51 PM on October 12, 2010

Call me when a machine beats a human playing Go. It will never happen.
It could beat me today. I assume you mean a good player.
posted by MtDewd at 6:11 PM on October 12, 2010 [1 favorite]

No doubt. My phone beats me all the goddamn time.
posted by box at 6:32 PM on October 12, 2010

I love Shogi, but this is hardly big news. Call me when a machine beats a human playing Go. It will never happen.

I'd bet you all the money in the world that it will. And I'd guess in the next 3-4 years.
posted by empath at 8:40 PM on October 12, 2010

10^224?? Bah. Child's play. The # of Go games is 10^768, says no less authority than Wikipedia.

If you take 10 billion years of yoctoseconds, stack a mole of those, then bundle one of those for every star in the visible universe, then cluster one of those for every atom in your body, then heap a google of those ... then WolframAlpha says you've up to about 10^216. It takes a heapin helpin of big numbers to Katamari to 10^768!
posted by Twang at 2:35 AM on October 13, 2010

I am sure that a machine will beat a 9-dan professional Go player eventually, but I would be astonished if it happens within 10 years. They've made great strides, but they're still pretty far away.
posted by dfan at 7:10 AM on October 13, 2010

How about a Magic:The Gathering AI that can design decks and play given an arbitrary card environment? Seems like the ultimate AI challenge, to me. Go has a relatively simple set of rules, and the rules don't change in the middle of a game.
posted by empath at 9:18 AM on October 13, 2010