1
$\begingroup$

I mean this in the sense that Go is unsolvable but AlphaGo seems able to make choices that are consistently more optimal than a human player's choices.

It is my understanding that Game Theory turned out to have limited applications in real world scenarios because of the profound complexity of such scenarios and degree of hidden information. Is it fair to say that there is now a method for dealing with this?

I fully understand that Go is a game of complete information, which has a very specific meaning, but it occurs to me that the inability to generate a complete game tree (computational intractability) could be seen as form of incomplete information, even if it is not traditionally thought of in those terms.


I should probably note that my perspective is one of a "serious" game designer, where complexity serves the same function as chance and hidden information, which is to say as a balancing factor that "levels the playing field".

$\endgroup$
1
$\begingroup$

I think that the technique AlphaGo used to solve the computational intractability problem of the search space are not new, it uses the Monte Carlo Tree Search. The real innovation in AlphaGo was to figure out how to compute the evaluation function of a move, that was the really tricky part. For this they used combinations Deep and Reinforcement learning techniques.

$\endgroup$
3
  • $\begingroup$ So basically a big step in a greater overall movement, with Go as a convenient proving ground. Am I wrong in thinking there are profound implications to where we're getting in this field? $\endgroup$
    – DukeZhou
    Dec 2 '16 at 18:45
  • 1
    $\begingroup$ You are not wrong but we need make sure that we don't get into the mindset of "When you have hammer everything looks like a nail". History have shown again and again that once we develop some new approach to a problem, people start to apply that approach to various problems and there is excitement but in the end the approach runs out of fuel and we need to look for something else $\endgroup$
    – Ankur
    Dec 3 '16 at 4:56
  • $\begingroup$ In some way the followup question was a reaction to a Business Insider article which touched on an initial application with the conclusion that the DeepMind acquisition looks like it will pay off, but didn't really take the next step to the idea that there are a myriad of technological systems where this could apply. $\endgroup$
    – DukeZhou
    Dec 3 '16 at 21:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.