Demis Hassabis describes how AlphaZero, starting from scratch, became 'the greatest Chess playing entity that's ever existed' in only 9 hours

AlphaZeroはチェスに対する知識がゼロから9時間でどのチェスAIよりも優れたシステムになった。

AlphaZero becoming the best chess player in the world, with *no prior knowledge, or training on human play* and just by purely playing against itself *in 9 hours* from breakfast until dinner -- is a wonderful demonstration of how exciting things will get soon.

We've come a long way since. AlphaZero, a 2017 breakthrough in 'self-supervised learning', was trained solely by playing games against itself! Not a single human game or opening was fed to it.

arxiv.org/pdf/1712.01815

AlphaZero beats humans because searching over a number of branching possibilities is difficult for humans, and especially easy for machines when combined with modern ML techniques.

HOPE is easier found when you hear Demis Hassabis talk about incredible drug breakthroughs that’ll happen (quickly)! He’s over delivered w/AlphaGO, AlphaZero& AlphaFold; therefore, I believe his prediction.
Full interview that was posted earlier youtu.be/0_M_syPuFos?fe…

Mehregan امید فقط معنی hope نمیده

AlphaZeroで大局将棋のAIを作ってみたいなと昔思ったことがあるけど、ぶっ壊れ駒が多すぎて実は数十手で必勝手順あったりしそう。

AI provides accuracy, but humans fill in the blanks with intuition. From AlphaZero to the future of Shogi AI, we can expect a fusion of human creativity with AI precision. As Mr. Habu puts it, 'The AI is now mixed in the world of Shogi.'

#Shogi #AI
youtube.com/watch?v=_iMaog…

Sound! Euphonium Season 3 Episode 5. Career Path. 響け！ユーフォニアム. youtu.be/K50JY9AJn6Y?si… via YouTube

Demis Hassabis: AlphaZero, starting from scratch, became 'the greatest Chess playing entity that's ever existed' in only 9 hours

youtu.be/0_M_syPuFos?si…

One Piece ワンピース Episode 1102. Escape Plan Begins! youtu.be/kBO-a9jnjzI?si… via YouTube

“Kramnik has more opening novelties than all of us combined…” Gelfand on 2013 Candidates.

Kramnik defeated Kasparov with Berlin, the best opening according to Alphazero.

Kramnik now: let me teach you about statistics 🙈

Spice and Wolf: Merchant Meets the Wise Wolf Episode 5. Rescue and on th... youtu.be/EXBmbsV7BT8?si… via YouTube

Chris Shashank Joshi The simplest explanation is that it was not designed or trained for playing games. AlphaZero does that, quite well, and you can write the solution to tic-tac-toe in a page of code. LLMs have limited reasoning depth, but do reason, as in the example I gave you.

The best part of video games is killing god

clips.twitch.tv/PuzzledTriumph…

Gracias Demis & team !!!
#GoogleDeepMind es impresionante 🤯🔝. No dejes de verlo para entender mejor la IA ... #alphazero #alphago #alphafold #reinforcedlearning ... #ciencia
(Enseñáselo a tus hijos, sobrinos, gente joven con gusto por la ciencia, aprender, curiosidad 💡👏🏼)