[Computer-go] Japanese Go Congress: Talk on AlphaGo

"Ingo Althöfer" 3-Hirn-Verlag at gmx.de
Fri Jul 15 09:26:35 PDT 2016


Hi,

Prof. Marc Oliver Rieger from Trier University (Germany)
is a participant in the Japanese Go Congress (July 15-18, 2016)
in Takarazuka. In Germany's computer Go forum he gave some
first impressions from the talk given by a Google developer of
AlphaGo (including the question & answer session):

* Alpha go is still under Active development

* How many developers still work on that? 
Not as many as before

* Current version stronger than Lee Sedol version but definitely not four stones!

* Don't believe internet rumors! (he also mentioned the stupid Ko conspiracy story)

* Even without mc pretty strong. We haven't measured exactly

* They also used other data than kgs 
-> radically different results regarding style

* What data sets did you use?
You'd be surprised if you'd know what data sets we tried out! 
But I can't tell you, sorry.

* Very styles for different versions of alphago

* I review my games with alpha go 

* Personally, I would very much like to see alphago being sold commercially

* For me (5d Japanese) it would be strong enough without too much computational power

* Will alpha go play beautiful?
Didn't you think the shoulder hit was beautiful?   ;-)

The link to Marc's posting is here:
http://www.dgob.de/yabbse/index.php?topic=6158.msg203261#msg203261

An announcement of the congress, from June 26, is here:
http://www.asahi.com/ajw/articles/AJ201606260009.html

Ingo.



More information about the Computer-go mailing list