Cepheus - superpocitac ovladl poker heads up holdem

http://technet.idnes.cz/poker-superpocitac-cepheus-df7-/veda.aspx?c=A150115_165002_veda_pka#utm_source=sph.idnes&utm_medium=richtext&utm_content=clanek-box

Projekt kanadskeho tymu “rozlustil” nejlepsi pokerovou strategii.

http://poker.srv.ualberta.ca/

The program was trained for two months using more than four thousand CPUs each considering over six billion hands every second! This is more poker than has been played by the entire human race.

The concept underpinning the program that built Cepheus is, in a sense, relatively simple. A ‘dummy’ AI is generated and begins playing poker against a copy of itself, exploring the ways it might play each hand. It then reviews the hands, and every time that it discovers an action it regrets it reduces the frequency with which it takes that action, implicitly increasing the likelihood that it takes some other action. While this description forgoes the details, the end result is that the dummy AI continues to play and evolve until it eventually arrives at a strategy which it no longer wishes to change, a ‘best’ strategy where any change would increase regret. This strategy is Cepheus. Since poker is a symmetrical game, the end strategy which Cepheus plays is an unbeatable one. While chips can, of course, be won from Cepheus in the short term, there is no decision which can be made against it which will be a winner in the long term. If a perfect opponent, either human or computerized, were to play a semi-infinite number of hands against Cepheus the best possible result would be for them to break even. Any imperfect opponent, which unfortunately includes all human players, would make mistakes along the way and lose. That being said, what Cepheus cannot do is maximize its winnings against weak opponents, a skill which humans excel. Cepheus is simply an invincible, immovable bunker, a Maginot Line that actually works.

Podobne jako poker, zda se ze i Go je nyni vyreseno 12-vrstvou neuronovou siti naucenou pomoci monte carlo… Udajne 80 let hrani…

http://www.theguardian.com/technology/2016/mar/09/google-deepmind-alphago-ai-defeats-human-lee-sedol-first-game-go-contest

AlphaGo (by Google) porazil Lee Sedola.