Pros making progress against AI

[pookpooi]

Not trying to diminish the success, but DeepZen has been playing in Tygem since yesterday and win 45 lose 1 (against TuTu by B+2.5) so it's not invincible yet.
Anyway, 100 consecutive wins make me think about changing the topic to 'AI making progress against Pro' instead
CGI and Leela also play in Tygem too, while they can maintain 9D status, their results are significantly worse than DeepZen. So only DeepZen and FineArt are two forefronts of 'active' Go AI.

Emphasis mine.

Sorry, folks, but let me reiterate my point that the most effective learning tasks are, in general, those with a success rate of around 50%. IMO the pros would do better to take handicaps from Zen. Or large komi.

FineArt already done this (accepting only 2-4 handicap games) but I'm not sure if it's for benefit of training Pro or training AI to give handicap stones to Pro...

[moha]

100 consecutive wins make me think about changing the topic to 'AI making progress against Pro' instead

It seems obvious that humanity won't catch up anymore, and the gap will widen, but this is natural (a specialised system should outperform a general one in it's own field).

And there are new things ahead (handi games or -komi, figuring out the maximum strength possible, learning from analysis with near-perfect programs - or maybe a new, massively distributed system, to overcome the gap in resources used by Google).

[luigi]

Doesn't look like the progress, if there ever was any, was sustained: Zen apparently won the last 100 games in a row (quite possibly a stronger version).

Given that DeepZen is most probably running on much lesser hardware than AlphaGo Master was when it won 60 games in a row against pros, and that 100 is greater than 60, does this mean that DeepZen is already better than AlphaGo Master?

[Uberdude]

Doesn't look like the progress, if there ever was any, was sustained: Zen apparently won the last 100 games in a row (quite possibly a stronger version).

Given that DeepZen is most probably running on much lesser hardware than AlphaGo Master was when it won 60 games in a row against pros, and that 100 is greater than 60, does this mean that DeepZen is already better than AlphaGo Master?

No, the (mostly young or old) Japanese pros it is playing are considerably weaker than the top pros Master faced. Occasionally it beats someone like Ichiriki Ryo who is about world top 50 level, but mostly it's much weaker pros. See some losses from Tygem where it is playing stronger presumably Korean/Chinese pros: viewtopic.php?p=224507#p224507

[luigi]

Doesn't look like the progress, if there ever was any, was sustained: Zen apparently won the last 100 games in a row (quite possibly a stronger version).

Given that DeepZen is most probably running on much lesser hardware than AlphaGo Master was when it won 60 games in a row against pros, and that 100 is greater than 60, does this mean that DeepZen is already better than AlphaGo Master?

No, the (mostly young or old) Japanese pros it is playing are considerably weaker than the top pros Master faced. Occasionally it beats someone like Ichiriki Ryo who is about world top 50 level, but mostly it's much weaker pros. See some losses from Tygem where it is playing stronger presumably Korean/Chinese pros: viewtopic.php?p=224507#p224507

Its rating is misleading then. On http://www.yss-aya.com/cgos/19x19/standings.html there are three Zen entries with ratings over 3800, and Ke Jie currently stands at 3666 on https://www.goratings.org/en/.

[pookpooi]

Its rating is misleading then. On http://www.yss-aya.com/cgos/19x19/standings.html there are three Zen entries with ratings over 3800, and Ke Jie currently stands at 3666 on https://www.goratings.org/en/.

CGOS rating anchored with Gnugo-3.7.10-a1 being fixed to 1800 Elo all the time. While Elo in AlphaGo paper anchored with goratings.org (GNU Go 3.8 in this paper only rate 431 Elo)

[luigi]

Its rating is misleading then. On http://www.yss-aya.com/cgos/19x19/standings.html there are three Zen entries with ratings over 3800, and Ke Jie currently stands at 3666 on https://www.goratings.org/en/.

CGOS rating anchored with Gnugo-3.7.10-a1 being fixed to 1800 Elo all the time. While Elo in AlphaGo paper anchored with goratings.org (GNU Go 3.8 in this paper only rate 431 Elo)

That explains a lot, thanks.

[pookpooi]

New AI account Fuheyuqi is very hot on Fox server. Actually it's not confirmed to be AI but it just win against 'at least 9D accounts' for 60 consecutive games. It win 99 out of 100 games it play with its new version. Here are two interesting game.

The first game is its only lose, against Ke Jie. This actually made news in many Chinese site. The time control (against all human opponent) is 1 minutes maintime + 3 periods of 30 seconds byoyomi. So it's quite fast and benefit from 'blitz' effect that favor AI.


The second game is its latest game (100th game) it won against FineArt UEC Cup version even giving 2 stones handicap! The time control is 1 hour maintime + 3 periods of 1 minute byoyomi.

[pookpooi]

FineArt(B) vs. Winner of Tencent Weiqi Tournament Tong Mengcheng 6p(W) Chinese rule 2 hours main time + 5 periods of 1 minute byoyomi

FineArt(B) vs. Runner-up of Tencent Weiqi Tournament Lian Xiao 9p(W) Chinese rule 1 hour main time + 5 periods of 1 minute byoyomi

[Uberdude]

Ah, so that's what TWTQQ meant in the game name on the Wbaduk broadcasts! Nice play from FineArt, it already seemed good for black (vs Tong MengCheng, who beat Lian Xiao in final by half a point) with capturing those 4 stones, but once black sabakid in the white moyo I thought it was over. Move 66 seemed very suspicious direction to me, making territory from the thickness side. Wouldn't something like k11 be better?