|
Job Discussion Forums
"The Internet's Meeting Place for ESL/EFL Students and Teachers from Around the World!"
|
| View previous topic :: View next topic |
| Author |
Message |
Didah
Joined: 25 Jul 2009
Posts: 88
Location: Planet Tralfamador.... and so it goes
|
 Posted: Thu May 22, 2014 11:49 am Post subject: Go Figure? |
 |
|
I thought I would offer a little cultural palette cleanser.
http://www.wired.com/2014/05/the-world-of-computer-go/
Although the game of Go is closely associated with the Japanese, it is said to have its origins in China along with Kanji and Buddhism. However, leave it to the Japanese to master a game where a human being can still beat a computer. Even Gary Kasparov, arguably the best chess player in the world, can now be beat by a lowly laptop with the right software.
The Japanese have taken frustrating foreigners to an art form, and now they have managed to do the same thing to a computer.
Only in Japan...
Any thoughts? |
|
| Back to top |
|
 |
Cool Teacher
Joined: 18 May 2009
Posts: 930
Location: Here, There and Everywhere! :D
|
| Posted: Thu May 22, 2014 2:04 pm Post subject: |
|
|
Yay! In your screen, computers! 
How about shogi? 
That's like "In your face!" for computers. |
|
| Back to top |
|
|
rxk22
Joined: 19 May 2010
Posts: 1629
|
| Posted: Thu May 22, 2014 10:33 pm Post subject: |
|
|
Very interesting.
I wonder if it has anything to do with a lack of research. IBM went after chess with Deep Blue. IBM dumped resources into beating humans at chess. WHile Go, it seems to just have some guy and his laptop.
Not that computing power matters these days, but who is writing the program? I think this is more of a case of there just hasn't been any real resources put into this as of yet. |
|
| Back to top |
|
|
Pitarou
Joined: 16 Nov 2009
Posts: 1116
Location: Narita, Japan
|
| Posted: Thu May 22, 2014 11:22 pm Post subject: |
|
|
| rxk22 wrote: |
| I wonder if it has anything to do with a lack of research. ... I think this is more of a case of there just hasn't been any real resources put into this as of yet. |
No, it's because Go is inherently "harder". The guys who write chess programs didn't just ignore Go.
Human-brains and the kinds of computers we have now work in very different ways. Computers are very fast, but (mostly) linear, and most effective at long, involved calculations. Humans are highly parallel, and most effective at fuzzy pattern-matching.
In a game of chess you have, 10-ish moves available to you at any stage. So a computer can use it's enormous computational power to look quite a few moves ahead, examine every possible outcome, and decide which one it likes best. Humans rely more on their pattern matching skills to recognize situations they've seen before, and try to nudge the game into a situation they know to be to their advantage. That's why Grand Masters memorize enormous long lists of opening move and endgame variations.
In a game of Go you have more like 100 moves available to you, and games take many more moves. Also, the game board quickly gets isolated into separate "fields", with only limited interaction between them. That's where a human's strengths really begins to shine, whereas a naive "try every possible move" computer program is overwhelmed by the sheer number of possible moves it has to try.
In fact, it's pretty remarkable that Go computers have made any progress at all That fact that a Go computer can play a reasonable intermediate level game shows that researchers have, to some extent, learned to mimic a human's pattern matching skills in silico. You can see it as part of the recent explosion in "Machine Learning", which came some time after Deep Blue. |
|
| Back to top |
|
|
nightsintodreams
Joined: 18 May 2010
Posts: 558
|
| Posted: Thu May 22, 2014 11:50 pm Post subject: |
|
|
Intersting, I had no idea.
Anyway, after a quick google I found this.
| Quote: |
| The number of legal chess positions is 10^40, the number of different possible games, 10^120. |
That seems like quite a lot of moves by anyone's standard, but Go is on some next level...
http://senseis.xmp.net/?NumberOfPossibleGoGames
http://en.wikipedia.org/wiki/Go_and_mathematics
Anyway, the prize money for someone who builds a computer/program to beat a human at Go is only one million dollars. Compare that to a the funds IBM put into developing a chess program and it's pocket change. If a computer heavyweight like IBM was to invest heavily into it, I'm sure they'd be able to eventually beat the top human players, after all...
| Quote: |
As of November 2013[update], China's Tianhe-2 supercomputer is the fastest in the world at 33.86 petaFLOPS, or 33.86 quadrillion floating point operations per second.
|
|
|
| Back to top |
|
|
Pitarou
Joined: 16 Nov 2009
Posts: 1116
Location: Narita, Japan
|
| Posted: Fri May 23, 2014 2:24 am Post subject: |
|
|
| nightsintodreams wrote: |
| If a computer heavyweight like IBM was to invest heavily into it, I'm sure they'd be able to eventually beat the top human players, after all... |
Believe me, considerable resources have been thrown at Go over the years.
I haven't read the Wired article you linked to, but I have a background in Computer Science (as well as Economics) and I've been following this problem for many years. I won't say, "It's impossible." but I don't think it's a simple matter of bigger, faster computers. You quickly get diminishing returns to scale. If your computer is 1000 times faster than mine, you'll get only an incremental improvement in performance. The real progress is conceptual, and just throwing money things doesn't guarantee big breakthroughs.
Like I said earlier, go brings out the strengths of the human brain, so I suspect that we'll only see really strong go playing computers when we have computers whose fundamental, silicon-level architecture more closely resembles that of the human brain. |
|
| Back to top |
|
|
rxk22
Joined: 19 May 2010
Posts: 1629
|
| Posted: Fri May 23, 2014 12:28 pm Post subject: |
|
|
Pitarou, Interesting points. I am not sure as to why, as I am split. Big Blue, was a massive massive undertaking by IBM. The new Watson beat people at Jeopardy, and that cost $10B. I am sure that something like Watson, which can almost think, could win at Go.
I do not know much about PC programming, which would mostly be the issue here |
|
| Back to top |
|
|
Pitarou
Joined: 16 Nov 2009
Posts: 1116
Location: Narita, Japan
|
| Posted: Fri May 23, 2014 1:15 pm Post subject: |
|
|
| rxk22 wrote: |
| Pitarou, Interesting points. I am not sure as to why, as I am split. Big Blue, was a massive massive undertaking by IBM. |
No it wasn't. Deep Blue was an iteration of Deep Thought, which was developed by a team of grad students at Carnegie-Mellon University. IBM hired them, and gave them the necessary resources to make Deep Thought faster. In particular, they went to the expense of fabricating custom silicon chips that could evaluate chess positions extremely quickly (but were useless for anything else). That was expensive, but not exactly a moon shot, and the whole exercise more than paid for itself when Deep Blue beat Kasparov and the press went nuts.
| Quote: |
| The new Watson beat people at Jeopardy |
I agree that it's a much more significant development, but be careful about phrases like "it can almost think". I could spend this weekend writing a computer program that could kick your ass at Sudoku or Connet 4 -- is that thinking or just calculation? Are you sure you know where the dividing line is? What about twenty questions? The history of A.I. is full of machines that fooled people with superficially thought-like behaviour.
Where are you getting your figures from? I think $10 million sounds more likely.
Anyway, getting back to what I was saying earlier, I think you're too easily dazzled by powerful hardware. If we were to go back in time to the late 1990s, when Deep Blue was strutting its stuff, and given the IBM Watson's hardware to play with, they would not have been able to build a Watson, because the state of the art in Machine Learning hadn't advanced to that level. The real magic in Watson lies in the new algorithms, not the hardware. Do you see?
To win at go, we don't need faster computers. Well, not just faster computers, although it goes without saying that faster is better. We need better tools in our toolbox of algorithms. It may be that we also need custom silicon chips, like with Deep Blue, and this time round, those chips might actually be useful for something! |
|
| Back to top |
|
|
rxk22
Joined: 19 May 2010
Posts: 1629
|
| Posted: Fri May 23, 2014 4:15 pm Post subject: |
|
|
Sorry about Watson, it is more around $1B http://money.cnn.com/galleries/2010/technology/1008/gallery.biggest_tech_gambles/3.html
I don't think computers can think, but the programming is getting closer and closer to being be able to, and not just calculate. Close, but not here. The speed, well we have long since reached an acceptable speed do most non-theoretical math type equations.
That said, I still am not fully convinced it is that Go is that much more complex. I think there hasn't been much programming thrown at it, and like Deep Blue, no specialized chips and what not.
I get what you are saying, and I feel you are much more knowledgeable about this, by far |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
This page is maintained by the one and only Dave Sperling.
Contact Dave's ESL Cafe
Copyright © 2018 Dave Sperling. All Rights Reserved.
Powered by phpBB © 2001, 2002 phpBB Group
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
