Talk:Deep Blue (chess computer)/Archive 2

Recently added paragraph on the outcome
I have concerns about the recently added paragraph on the outcome. The good is that this has a source and that adding more about the significance of Deep Blue would improve the article. The bad:
 * "Regardless of the claims and the reasons for disassembling Deep Blue, the implications for artificial intelligence were interesting."
 * What does disassembling Deep Blue have to do with the implications for AI? Claiming that that they are "interesting" is troublesome language for an encyclopedia as well (MOS:OPED).  Probably this sentence should just be omitted since it doesn't actually say anything encyclopedic.
 * "Computer scientists believed that playing chess was a good measurement for the effectiveness of artificial intelligence, and by beating a world champion chess player, IBM showed that they had made significant progress with the concept."
 * I'm not sure precisely what "the concept" means here or how it adds to the meaning of the sentence. Wouldn't it be better to simply say "IBM showed that they had made significant progress"?
 * "Deep Blue was able to analyze data and respond to that data with the best decision available in the computer’s software. While Deep Blue was not really making decisions, the victory over Kasparov was a huge milestone in artificial intelligence. It let the development team know that they were not wasting their time and that their ideas could be successful with more development."
 * So the first sentence says Deep Blue made the best decision available and then the very next sentence says Deep Blue was not really making decisions. What are we trying to say here?  That doesn't make any sense.  (Also "analyse data and  respond to that data with the best decision possible" is pretty terrible.  It would describe nearly any computer program.)  I'm also not sure what it means to say that "their ideas could be successful with more development".  Weren't their ideas already successful?  They set out to build a chess machine that could beat the world champion, and it did.  Why would that need more development to be successful?  It already succeeded.

I think this reference can be used to improve the article, but I don't like what we have now. Quale (talk) 03:44, 2 July 2018 (UTC)
 * Thought similar. Encyclopedic rewrite warranted. Need coffee first! --IHTS (talk) 03:32, 4 July 2018 (UTC)

Later programs not brute force search?
The article current claims that more recent (Chess, Go?) programs do not perform brute force search. I corrected this incorrect claim, but somebody reverted it, claiming that programs that use neural networks do not do brute force search. Monte Carlo tree search is as much "brute force" as alpha-beta minimaxing. Neural networks also do not do any kind of "thinking" in Go: they are the heuristic for choosing which parts of the search tree to search. The main body of the newest Go programs is still very much brute force search. If you could successfully play Go without MCTS and any other tree search method, then it would not be "brute force" any more. But neural networks do not help getting rid of that brute force part. Could somebody take care of fixing this issue? I do not plan to fight this one. — Preceding unsigned comment added by Sciken (talk • contribs) 15:49, 20 March 2020 (UTC)
 * I don't see a claim like that about newer programs on the current version of the page. It says "Modern chess programs like Houdini, Rybka, Deep Fritz or Deep Junior are more efficient" and then goes on to explain the claim of greater efficiency means the newer programs search fewer positions per second but manage to search deeper trees.  That seems reasonable to me.  Whether Monte Carlo tree search should be called "brute force" or not isn't really clear to me.  It's a heuristic, but the tree search used in Deep Blue and other programs use heuristics too.  Neither algorithm resembles the kind of highly selective search that human chess players use.  Fortunately I don't think this article needs to say anything about Monte Carlo tree search, so it's moot.  Quale (talk) 05:49, 21 March 2020 (UTC)