recentpopularlog in

deepmind

« earlier   
Twitter
needs mere 4 hours of self-training to become a chess overlord:
BigData  AI  DeepMind  from twitter_favs
10 days ago by rukku
Google's AlphaZero destroys Stockfish in 100-game match • Chess.com
Mike Klein:
<p>Chess changed forever today. And maybe the rest of the world did, too.

A little more than a year after AlphaGo sensationally won against the top Go player, the artificial-intelligence program AlphaZero has obliterated the highest-rated chess engine. 

Stockfish, which for most top players is their go-to preparation tool, and which won the 2016 TCEC Championship and the 2017 Chess.com Computer Chess Championship, didn't stand a chance. AlphaZero won the closed-door, 100-game match with 28 wins, 72 draws, and zero losses.

Oh, and it took AlphaZero only four hours to "learn" chess. Sorry humans, you had a good run.

That's right - the programmers of AlphaZero, housed within the DeepMind division of Google, had it use a type of "machine learning," specifically reinforcement learning. Put more plainly, AlphaZero was not "taught" the game in the traditional sense. That means no opening book, no endgame tables, and apparently no complicated algorithms dissecting minute differences between center pawns and side pawns…

…GM Peter Heine Nielsen, the longtime second of World Champion GM Magnus Carlsen, is now on board with the FIDE president in one way: aliens. As he told Chess.com, "After <a href="https://arxiv.org/pdf/1712.01815.pdf">reading the paper</a> but especially seeing the games I thought, well, I always wondered how it would be if a superior species landed on earth and showed us how they play chess. I feel now I know."</p>

The article includes one of the games. It feels quite different from how a human plays. AlphaGo seems to play as though it has all the time in the world; that it's not particularly worried by threats, but equally wants to make exchanges on its own terms. Stockfish never seems to force it. AlphaZero even shows which openings are best. Queen's Gambit and English Opening, apparently. (I prefer Bird's Opening. Get things started.)

As Eric David <a href="https://siliconangle.com/blog/2017/12/06/deepminds-alphago-mastered-chess-spare-time/">notes at Silicon Angle</a>:
<p>What makes DeepMind’s latest accomplishment is noteworthy is the fact that it conquered three games with very different rule sets using a single AI. AlphaGo Zero, the latest version of AlphaGo, began “tabula rasa” without any prior knowledge or understanding of Go, shogi or chess, but the AI managed to achieve “superhuman performance” in all three games with stunning speed. IBM spent more than 10 years perfecting Deep Blue before it successfully mastered chess. AlphaGo Zero did it in just 24 hours.</p>
chess  deepmind  ai  learning  machinelearning 
11 days ago by charlesarthur
High-fidelity speech synthesis with WaveNet | DeepMind
"The training method has parallels to the set-up for generative adversarial networks (GANs), with the student playing the role of generator and the teacher as the discriminator. However, unlike GANs, the student’s aim is not to “fool” the teacher but to cooperate and try to match the teacher’s performance.

Although the training technique works well, we also need to add a few extra loss functions to guide the student towards the desired behaviour. Specifically, we add a perceptual loss to avoid bad pronunciations, a contrastive loss to further reduce the noise, and a power loss to help match the energy of the human speech. Without the latter, for example, the trained model whispers rather than speaking out loud."
wavenet  speech-synthesis  deepmind 
22 days ago by arsyed

Copy this bookmark:





to read