There is a reddit thread that claims AG is playing on Tygem using an account named Deepmind.

It links to this page. I can neither read the page, nor access Tygem to verify. Needless to say, I am skeptical, and suspect Deepmind is played by Aja Hwang. But I'd love to be wrong!

It would be fantastic if someone posted some recent Deepmind games here, so we can decide for ourselves!

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Do you mean this page (from the Reddit post)?

http://www.tygem.com/news/news/viewpage ... &findword=

If so, there isn't much content, but I can summarize.

Paragraph 1: There's various talk that AlphaGo uses the 'deep mind' account on Tygem as a stage for it's real-game training, and s/he (the account) has again revealed himself!

Paragraph 2: On the 25th at 9am, or 1am in London taking into account the 8 hour time difference, Tygem's 'deepmind' beat (by resignation, as white) Japanese 9d, 'universe55' in just 82 moves.

Paragraph 3: In today's match, 'deepmind' played a double 3-3 opening as white. Last March in the Lee Sedol vs. AlphaGo Google DeepMind Challenge, AlphaGo played a double 4-4 opening in all of the 3 games it played as white, which is a bit of a contrast.

---

That's about it. The rest of the article consists of captioned images (a few comments on the game, and then the player profile).

_________________
be immersed

I, too, am skeptical that it's 100% full strength AlphaGo. Before the Lee Sedol match, Tygem posted about this account and it gave pros more confidence Lee Sedol would win the match. The real AlphaGo turned out to be much stronger.

_________________
be immersed

On the Reddit thread in question, Uberdude quoted Aja Huang who said,

"We haven't decided to play AlphaGo online yet, but when the decision is made we will use AlphaGo(P) on tygem and AlphaGoBot on KGS."
from http://permalink.gmane.org/gmane.games.devel.go/30213

Personally, I suspect that there might still be some connection between 'deepmind' and AlphaGo (or its development team). It's a pretty strong player, and some other Tygem articles have sometimes noted its unique play. There's also the fact that the account was created around the time development for AlphaGo started.

Maybe Aja or someone else that's on the development team uses Tygem as a testbed to train some subset or component of AlphaGo.

_________________
be immersed

There's also the article we discussed here on this forum, where Demis seemed to hint at the possibility that they'd used (some version of?) AlphaGo on Tygem.

_________________
be immersed

the Official Statement on this was that this was Aja Huang's account. IIRC, he claimed it on Facebook.

I seem to recall this, too. It's just suprising to me that Aja Huang could be 9d on Tygem without the, uh... assistance of something else

_________________
be immersed

Fan Hui is in DeepMind's team. Also the true strength of Aja Huang is like a black box, he recently inflated to 8D on KGS.

What is his account on KGS?

_________________
be immersed

By the way, per Brady's original request, here are a few games. They are from December of 2015.







'deepmind' is white in all cases.

_________________
be immersed

I just happened to find those on a blog post. When I get home, I can post some more recent games of 'deepmind' (unless someone else here would like to do it first ).

_________________
be immersed

There was a TYGEM article confirming by Demis Hassabis that deeonubd is id that used to training.

http://www.tygem.com/news/news/viewpage ... &findword=

Thanks, trout.

Anyway, as promised, here are some more recent games, all from the 25th - including the "double 3-3 opening" game that the Tygem article talks about.












Note that 'deepmind' hasn't won all of the recent games. Here's a look at the recent game history:



So there you have it. Some of 'deepmind's most recent wins.

Is 'deepmind' truly AlphaGo? Judge for yourself.

_________________
be immersed

Thanks, Brian!

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Another aspect to consider in the "is 'deepmind' AlphaGo" discussion are the time settings played in games on Tygem. If it's believed that 'deepmind' is weaker than AlphaGo, it could be a factor.

On a different thread 'seigenblues' suggested that AlphaGo plays better with more thinking time. If that is true, then we might expect 'deepmind' to have weaker play on Tygem.

_________________
be immersed

Unless he's gotten stronger recently, it's fairly safe to say he is European 5d level. I (Euro 4d) have played him twice and lost twice, the first was a fairly sensible game, slightly losing until I misread some important fight, the 2nd was just after I lost the first British Championship game so I was in a frustrated mood and played crazily and we both had chances to win, he missed a few tesujis I found but then I got angry and went bonkers kamikaze style. You can see him winning and losing to 5ds but mostly losing to 6ds here:
http://www.europeangodatabase.eu/EGD/Pl ... y=17374016

How strong are you on Tygem, Uberdude?

_________________
be immersed

Does anyone know how to download someone else's games off Tygem? While I can look at such games, the save button doesn't seem to work in the English client. (I don't care about format, gib to sgf is easy.)