Maybe 'creative' is the wrong word? Coming up with a novel idea in chess, where the rules are basic and it's possible to "win", is different from creating a beautiful poem, where there aren't the same strict rules/limits. AI can do the former, but can't do the latter?

AlphaZero is apparently completely self taught. It developed by playing itself millions of times, so maybe it did literally reach this ability on its own. Unlike Deep Blue and most other chess computers, which I think are more like calculators using brute force and human guidance.

Quite slow- what you'd call "classical" time controls. For some moves they will "think" for 20 seconds, others for 20 minutes.

@didau.bsky.social @trivium21c.bsky.social
On the topic, Matthew Sadler and Natasha Regan wrote 'Game Changer', about the impact of the AlphaZero chess engine. In the first paragraph they describe it as having a "creative style". I think it can probably be innovative, but can it really be creative?

Very interesting discussion- as ever. Interesting to hear the example of how nobody would want to watch two computers playing Fifa. In chess, there's a reasonable interest in the 'Top Chess Engine Championship', where supercomputers compete (all streamed on Twitch!) tcec-chess.com @didau.bsky.social