Eddie, it would be doubtful only if we ask a question, bit if we don't ask a question - no doubts.EddieMoore wrote:and it's doubtful he would have lost the full match had he been healthy.
Alex_Moiseyev wrote:Thats all only yellow media and Chinook team is not guilty of this at all. In my life I saw very small number of reporters who wanted to find a true and make an article full and accurate with fatcs and interpretation.
Remember, if you are given an interview to someone - you both are on the different parts of barricades ! Most of them always care only the interest, attention and number of readers. For this purpose they can write anything which will help them to meet and accomplish this goal !
Alex
Schaeffer's goal was to have Chinook win a human world checkers championship. Though it essentially won the title in 1994, the victory came as a result of a default. Marion Tinsley, a mathematics professor widely regarded as the greatest checkers player in history, withdrew from the competition because of illness and died eight months later of pancreatic cancer.
That left Schaeffer with only one way to prove his computer was better than Tinsley. He refocused his efforts on solving the game to demonstrate once and for all that a properly programmed computer could not be beat by a human
http://www.cs.ualberta.ca/~chinook/publ ... ckers.htmlMostFamousDane wrote:Does anybody have any techinal details on how the chinook team has constructed this "proof" ?
Seems like they were using advanced search techniques like transposition tables in their search. Many of the advanced search techniques are not guaranteed to give the correct result but are used because the result is usually better than a shorter search. I wonder if they thought about this - it doesn't seem to be adressed in the 2005 article. Using even something relatively simple like e.g. zobrist keys would invalidate the proof.Alex_Moiseyev wrote:http://www.cs.ualberta.ca/~chinook/publ ... ckers.htmlMostFamousDane wrote:Does anybody have any techinal details on how the chinook team has constructed this "proof" ?
Sune,MostFamousDane wrote:would invalidate the proof.
- The details are nessary if one is to judge how good their argumentation that checkers is drawn is. Since their proof appearently relies heavily on search it immediately raise a red flag in my mind. but also a pure alpha/beta search of ALL nessary lines terminated only at db lookups would indeed be a very convicing proof - but the search tree is just big for this to seem likely. Words like iterative deeping and transposition tables which appear in the 2005 article could indicate the use of imperfect search techniques like zobrist keys and futility pruning have been used. Alex_Moiseyev wrote:Sune,♦MostFamousDane wrote:would invalidate the proof.
what really bother me here - alot of uncertainces and "warning red signs" in media, which puts in question a clearance of proof. For instance:
1) In one article I read that they used a help of human player (!!!) and other programs for validation of certain lines.
2) In other article I read that they cut a weak moves, which also would eliminate proof immediately.
3) Any technical system always guarantee only 99.99999999999% with many nine's, bur never 100%. Running a search for 3 years on 20-50 machines would raise (theoretically) a chance of simple hardware glitch which would not terminate the search but change the results.
If we simply ask a question of 100% proof and accuracy of hardware - it would be really hard to prove this scientifically.
Alex
