[Computer-go] Our Silicon Overlord

Xavier Combelle xavier.combelle at gmail.com
Wed Jan 4 23:38:33 PST 2017



Le 05/01/2017 à 07:37, Robert Jasiek a écrit :
> On 04.01.2017 22:08, "Ingo Althöfer" wrote:
>> humanity's last hope
>
> The "last hope" are theoreticians creating arcane positions far
> outside the NN of AlphaGo so that its deep reading would be
> insufficient compensation! Another chance is long-term, subtle
> creation and use of aji.
>
The problem is that you have to find a way to constraint alphago to
reach the position you have prepared it will be very hard because it has
the choice of half of the moves which leads to the position.

From computer science point of view, theoricaly, the best move on an
arbitrary position being a PSPACE hard problem, any problem at least
easier than PSPACE could translate in a go problem. So there is a huge
amount of difficult problems (understand impossible to solve except on
toy size) which could be setup as target go positions but the real
problems is that you have to reach this position which are very unlikely
to happen in a real game.

An easy way to win against Alphago strength level of bot is to make two 
deterministic version of it, make it play one against the other and
replay the moves of the wining side.

--
Xavier




More information about the Computer-go mailing list