Geekpress linked to an interesting article on computer poker programs that are now becoming competitive with human poker champions. They're talking as if this is a level beyond Deep Blue for chess, and there's something to that. What's involved in beating a human poker champion is way beyond beating a human chess champion, at least when it comes to programming a computer. Still, there's something that seemed fairly problematic in how they're measuring the success of these computer poker programs, at least from how the article's presentation seemed to me to describe it. (I won't rule out the possibilities that unclarities in the article have disguised what they're doing.)
One crucial thing the article doesn't distinguish is very important for poker. It appears that some computer programs can play against some human players and play competitively against them. That's one measurement of how good a computer poker program is. But a more interesting measure would be to compare (a) how well the computer programs do against people with (b) how well the human players do against people. In other words, could the computer programs beat a novice like me as easily as the human champions do? Could they beat a champion as often as another champion could? Simply showing that they can beat champions almost as often as the champions beat them is something. But I'd be interested to see if they can handle human opponents as well as human players do. That seems to require a different play setup than just having all the humans play all the computer programs.