AI. Will 1.4 stop the following?

[Kallocain]

1. is solved in 1.4

 

[Chaingun]

That's a misconception, chess AI ineed has limited amounts of moves, but it still require more difficult algorithms (ratings) then used in most modern games, moreover it's not required for HoI AI to behave like genius, just not to look retarded.

It's just Al doesn't have enough attention (unlike say in Stardock games) and that's sad.

Programming a decent chess AI that beats anyone but grand masters takes weeks, programming an RTS AI that can beat a player with one month of experience can take years.

Of course, chess is very well researched in comparison. Yet, it is a WAY simpler game given the limited state:

typedef struct
{
	piece_t pieces[2][16];
	color_t color;
	int draw_clock;
	int en_passant_dest;
	int turn;
	bool castle_w_short;
	bool castle_w_long;
	bool castle_b_short;
	bool castle_b_long;
} pos_t;

 

[Chaingun]

Not true, the concept of heuristic pruning is used in AI to limit the search space - this is effectively discarding moves without evaluation or, as you put it, not examining all branches. Google Alpha-beta pruning if you're interested in more detail.

True, alpha beta can prune at most half of the moves on average per branch, which means the AI can search twice as deep in the search tree than it could otherwise. It is optimal in the sense it doesn't effect the search result.

Still, a human prunes a lot more (but thus risks making tactical mistakes, which means an AI only searching a few levels deep beats casual players who prune too much).