Newspaper Article on Frdiay

[john reade]

Actually the article in the Science Magazine does say Tinsley withdrew because of illness. See below for the quote.

In 1992, Tinsley defeated Schaeffer's program in a London tournament, but in a rematch in Boston two years later, Tinsley and the computer played to a draw for six matches before he fell ill and was forced to withdraw. He was later diagnosed with pancreatic cancer and died the next year, Schaeffer said.

[EdTrice]

I had numerous conversations with reporters also. It seems they care more about meeting their deadline than getting very detail-oriented.

One reporter said something like this to me: "So there are about 144 different ways to put the checkers on the board without a capture, right? 12 for one side, 12 for the other."

I was like: "Well, actually, more like 90 quintillion."

reporter: "90 what?"

me: "Quintillion. Quin -- till -- yun. It's a big number."

reporter: "Never heard of it. You made that up?"

me: "No. Take a billion, multiply by 1,000 and you get a trillion. Take a trillion, multiply by 1,000 and you get a quadrillion. Take a quadrillion, multiply by 1,000 and you get a quintillion."

reporter: "There's no way 12 checkers against 12 checkers has that many arragements."

me: "Ok, let me ask you something. How many ways can you place one type of material on the board. Hint: there are 32 squares."

reporter: "32, obviously."

me: "Buzz, wrong answer. Correct answer = 120."

reporter: "You just said there were 32 squares."

me: "I know. You can place one red king on any of those 32 squares. Running total of legal one-piece arrangments = 32, so far. You can also place a white king on any of the 32 squares on a blank board, so that makes 64 different positions comprised of one piece. Clear the board, and you can put a red checker on 28 squares, since the remaining 4 would be promotion squares. Likewise, there are 28 squares on which a white checker could reside. 32 + 32 + 28 + 28 = 120. See?"

reporter: {mumbled something}

me: "You have 120 positions in the game of checkers that represent wins once the last jump was made on the board. But there are 6,972 positions with 2 pieces on the board that may or may not lead to these 120 final positions. And there are 261,224 different 3-piece positions that may or may not simplify into those 2-piece positions. The numbers get very large very quickly."

I let him know that once you work your way up to just 5 pieces against 5 pieces, the total number represents more than 1,000 times the present population of the earth.

After that he said:

"Now I have to re-write the entire article."

[EdTrice]

http://www.nature.com/news/2007/070716/ ... 16-13.html

I said it was no big deal to the checkers players who enjoy the game. After all, people continued to run for sport after the invention of the automobile.

[Palomino]

... people continued to run for sport after the invention of the automobile.

Excellent analogy Ed.

It seems as if reporters never get the whole story correct, every time I've read an article in the newpaper after a reporter has interviewed me and then written an article about the work I do, I say to myself, "I didn't say that!" LOL, but I found the nature.com article to be pretty good.

Thanks for sharing.

Regards, "PaL" Bucker

[whitefork]

me: "Quintillion. Quin -- till -- yun. It's a big number."

reporter: "Never heard of it. You made that up?"

... [snip] ...

After that he said:

"Now I have to re-write the entire article."

That's beautiful. Thank you.

[Martin Fierz]

The simple fact is, they will not allow a person to have his entire library to be able to look up his moves in his books. He can't even use his own manuscript as an aid in a tournament game. The computers cannot play at the master level without using their opening and engame "books". I don't believe the computer program exists that can beat any human master player on equal terms, i.e. without books or manuscript, just over the board analysis using the same time limitations for both players.

This wil probably not always be the case, computer speed an memory capabilities are expanding at an enormous rate,

Hi Bob,

this is an interesting statement. I don't really think I believe you! You can make the experiment by turning Cake's book off, and renaming the database directory in the CheckerBoard directory from db to anything else, eg. nodb - then Cake will run without endgame database, and without opening book. I would gladly take a bet that it still plays at master level under such conditions. Also, keep in mind that endgame heuristics in the programs are not well developed because we have the databases - so we never had to develop particular heuristics for positions with very few pieces. I would think that this area of play could be improved upon if necessary (but who wants to do that if he can use the databases instead??).

Of course, there are some openings which are very hard for the computer to play properly under time restraints and without books - but probably for the human too...

cheers
Martin

[Alex_Moiseyev]

Bob, if program turn off opening book and ending database - does it mean that human also can't use his PP and ending knoweldges ?

Personally, such sort of knoweeldges plays the major role in my play and is about 60%-70% of all my abilities. I also use some transpositions and my experience from previous games which I played in the past ... and program does very much the same the same.

You try to make program weaker - it is not fear ! The human has strengths and weaknesses, program also. The only thing is the same - game and rules, but human and program achieve the goal differently.

Speed, memorising and calculations are program strengths - why we should take them out ?

Alex

[EdTrice]

Computer programs are tactically superior, no doubt. But humans are strategically superior.

Grandmaster Savielly Tartakower had a great quote:

"Tactics is what you do when there is something to do. Strategy is what you do when there is nothing to do."

Take a look at this position:

white checkers: 32,31,30,28,25,24,23,19,18
red checkers: 1,2,3,4,5,6,7,12,17

It is white to move.

Even if you have a 10-piece database and the program could search 2,000,000 positions per second, you would need to search for about 4 hours to avoid the losing move.

...31-26?? loses to 5-9!!, which was Hellman's solution to win against this line that can arise from the Black Widow.

The only move to draw is 18-15.

Even after 31-26?? most programs can't find 5-9!! after a long search. Most want to play 12-16 19x12 3-8 12x3 then the waiting move to complete the in-and-out shot that leads to the 10-piece draw.

So don't discount humans completely. The master analysts are still top notch strategically speaking.

[EdTrice]

Computer programs are tactically superior, no doubt. But humans are strategically superior.

Grandmaster Savielly Tartakower had a great quote:

"Tactics is what you do when there is something to do. Strategy is what you do when there is nothing to do."

Take a look at this position:

white checkers: 32,31,30,28,25,24,23,19,18
red checkers: 1,2,3,4,5,6,7,12,17

It is white to move.

Even if you have a 10-piece database and the program could search 2,000,000 positions per second, you would need to search for about 4 hours to avoid the losing move.

...31-26?? loses to 5-9!!, which was Hellman's solution to win against this line that can arise from the Black Widow.

The only move to draw is 18-15.

Even after 31-26?? most programs can't find 5-9!! after a long search. Most want to play 12-16 19x12 3-8 12x3 then the waiting move to complete the in-and-out shot that leads to the 10-piece draw.

So don't discount humans completely. The master analysts are still top notch strategically speaking.

[Alex_Moiseyev]

Ed, white has to build a trading column 24-19-15 on time and tempo is critical. 31-26 is an important wasting time ! It is simple and obvious for human.

Regards,

Alex

[Martin Fierz]

hi ed,

thanks for an interesting position.

i'm not so sure about the 10pc db and 4 hours - i will see in a few months when i finally have the database and 10-piece-cake
but of course that's not really the point - and i'm not discounting humans at all - all i was saying is that the programs are very strong even without databases (opening or ending). perhaps they wouldn't beat alex - but who knows? anyway, to say that they would not beat a "human master" (whatever that means) sounds very unlikely to me. the big problem of all human-machine comparisons is that the humans always make mistakes - sometimes really dumb ones like kramnik who allowed a mate-in-one against the computer...

Most want to play 12-16 19x12 3-8 12x3 then the waiting move to complete the in-and-out shot that leads to the 10-piece draw.

and who can blame them? the 10-pc endgame (W:W32,28,24,23,18:BK30,5,4,2,1.) looks very good.... or is there any easy way to see that this is in fact just good-looking but not good?

cheers
martin (always looking for good ideas for Cake 2.0)

[Alex_Moiseyev]

the humans always make mistakes

Yes - this is our (humans) privelegy not available for programs !

[EdTrice]

Yes - this is our (humans) privelegy not available for programs !

Well, not entirely true either. I recently wrote a new move generator that uses 35-bits, not 32-bits, and it is twice as fast on a 64-bit machine than under 32-bit hardware.

Now you might say why do you need 35-bits on a board with only 32-squares? But, the thing is, using 35 bits and "skipping" bits 8, 17, and 26 (meaning they are "not on the board") can make the move generator faster by reducing the type of bit-AND operations that are needed after doing register shifts.

I mention all of this because my first attempt at this move generator did not work. So the program was making bad mistakes

[Alex_Moiseyev]

So the program was making bad mistakes

I assume programmer made mistakes, not program ? But in general I like your argument, and want to discuss this with my manager !

Alex

[Pedro Saavedra]

Ed, there is a problem with your math. Not only can a final position not include a checker in the promotion squares, it cannot include a checker in the first two rows (since the last move must be a jump if there is only one man on the board). Hence we have only 32+32+20+20=104 legal positions with only one checker or king on the board.

In pool or Russian checkers, your 120 would be valid.

ED Trice says:

> me: "I know. You can place one red king on any of those 32 squares.
> Running total of legal one-piece arrangments = 32, so far. You can also
> place a white king on any of the 32 squares on a blank board, so that
> makes 64 different positions comprised of one piece. Clear the board,
> and you can put a red checker on 28 squares, since the remaining 4
> would be promotion squares. Likewise, there are 28 squares on which a
> white checker could reside. 32 + 32 + 28 + 28 = 120. See?"
...
> "You have 120 positions in the game of checkers that represent wins
> once the last jump was made on the board. But there are 6,972
> positions with 2 pieces on the board that may or may not lead to these
> 120 final positions.