Writing a chess-playing AI like Alphago in Python

[caigy84 at gmail.com]

On Sunday, December 24, 2017 at 1:26:22 PM UTC+8, Steve D'Aprano wrote:
> On Sun, 24 Dec 2017 12:20 pm, Cai Gengyang wrote:
> 
> > How many lines of code in Python would it take to create a Go-playing AI
> > like AlphaGo ? Estimates ?
> 
> Somewhere between 1 and 1 billion.
> 
> How about you start by telling us:
> 
> - do you mean AlphaGo or AlphaGo Zero?
> 
> - how many lines of code AlphaGo [Zero] has;
> 
> - in what language or languages;
> 
> - is Python allowed to call out to libraries written in other 
>   languages, e.g. machine learning and neural net libraries,
>   or databases, or does it have to implement *everything*
>   from scratch?
> 
> 
> The Michi Go engine uses about 550 lines of Python:
> 
> https://github.com/pasky/michi
> 
> but I don't believe it does any machine learning.
> 
> 
> See also:
> 
> https://github.com/rossumai/nochi
> 
> https://medium.com/rossum/building-our-own-version-of-alphago-zero-b918642bd2b5
> 
> 
> 
> -- 
> Steve
> “Cheer up,” they said, “things could be worse.” So I cheered up, and sure
> enough, things got worse.

> - do you mean AlphaGo or AlphaGo Zero?              ----- AlphaGo
> 
> - how many lines of code AlphaGo [Zero] has;        ----- No Idea
> 
> - in what language or languages;                    ----- Python
> 
> - is Python allowed to call out to libraries written in other
>   languages, e.g. machine learning and neural net libraries,
>   or databases, or does it have to implement *everything*
>   from scratch?                                     ----- Allowed to call out