Originally posted by Sam The Sham
Uhhhmmm, I don't think they could be called "crap" at anything. GM's have a hard time beating them on standard PC's nowadays. They'll still have a positive score if they bring their best game to the table, but they'll lose and draw a lot of games along the way.
openings for one. engines are still quite clueless without the
human-made opening books. they're not that good at endgame (or any other) strategy either, unless they're allowed to rely on
pre-computed tablebases. they can't crack that stuff in realtime.
in fact the only thing they do well, is calculation. human players still tend to try beating engines at that, which is completely undoable.
I don't think there has been strong human players understanding search algorithms who have played against engines yet. they're still relying on 'coffee house computing' misconceptions, dated weaknesses that maybe were true 20-30 years ago etc, kasparov being a prime example. -we haven't seen the best human vs engine play yet. if someone like kramnik understood deeply how search algorithms worked, and were willing to adjust himself accordingly (that would take years), we'd still see the top GMs blowing the engines right out of the water.
but there's no real reason for them to do that, no money and more importantly no point, so it's not likely to happen. so we're doomed to hear GMs talking about engine evaluation as if they
were humans, "what's rybka thinking with that move?" "rybka thinks this" rybka thinks that", which is completely misleading. engine evaluation is still just a subjective opinion, which can be wrong just like human evaluation. it just makes different kind of errors, because it
doesn't think. it's a glorified pocket calculator, incapable of
understanding the data it's fed with.
at least until the chess engine paradigma shifts from the brute force approach. but with the exception of a few tries at neural nets it doesn't seem to be happening yet. which ironically enough, is due to the lack of people understanding both computing and chess deeply. the math exists, even some hardware exists, but everybody's still trying to make the same old serial processing pocket calculators, only a little faster than the previous guy. there's been zero innovation in chess engines for decades, they're still doing the same thing as in the 80s, with minor performance tweaks. a student can code up a program which beats the other guys, because it's still the same exact crap it's been for decades. case in point: zappa engine.