Re: computer-go: Insight of a human

[Heikki Levanto <heikki@xxxxxx>]

Vincent Diepeveen <diep@xxxxxxxxxxxxxxxxx> wrote:

>>  - Describe the situation in simplified terms (live group here,
>>    strong-looking one there, this is dead, this looks weak)

> This is the basic question actually: how to describe them, as
> at the computer i only have 0 or 1 which either go to a simplified
> is true or not true, so even if i define a lot of simplified terms, then
> still there is the border problem; when does something belong completely
> to a certain term or not?

One could use some sort of descriptive language for this, or numbers (as has
already been pointed out). Either hard-coded numbers (since it has only one
eye, I call its strength 7), or something derived from the size of the local
reading trees, or something else again...

>>  - Isolating subgames. Computers do a little of it in the endgame,
>>    but could do much more, also much earlier in the game.

> This is the opposite of what i wanted to talk about. Obviously computers
> are superb in small spaces. The problem is having the oversight.

No, what I meant is that humans are good in seeing that this battle here has
nothing to do with that one. This is far from trivial for the computer
programs of today.


>>  - Persistent state. Humans remember much of their reasoning from one
>>    move to the next. Mostly status info (I know this group was alive
>>    3 moves ago, and nothing has been played in that corner since).
>>    Also plans and intentions (if I play this joseki variation
>>    and get out in sente, I can then extend in front of this wall).

> Let's not confuse having a PLAN with having information.
> All programs obviously have this information too, you can store
> anything in memory, logically you can still have filled hashtables.

What I meant here was more the ability to keep this overview even during a
long sequence of moves, and plan accordingly. Many programs I know of will
recalculate everything at every move, whereas it is very hard for humans to
get a "fresh look" at the situation.


>>  - Focus. Humans know what the game is "all about" at each stage

> This is a very good description. The right focus is indeed a very
> important word to describe the point at the board where one needs to
> make action. 

> oversight ==> better focus ==> better plan?

I think we may be talking of the same thing here, "focus" == "oversight" ??


>>> So not the search part but the evaluation part?
>> I am not even sure if this is a necessary or good distinction.
> I'm sure that within any program there is a clear distinction between
> search and evaluation.

Yes, but I am not at all sure that all programs do any search at all. of
course local tactical reading, but not necessarily any full-board strategic
reading. I know humans do very little, if any.

I can well imagine a program that will analyze the situation carefully
(including, of course, some tactical reading about life and connections),
describe it in some higher-level language ("weak group here", etc), make
plans on that level ("force the weak group to run this way, and secure
territory while attacking it"), and then find a move that implements the
plan (again using local tactical reading where necessary). If any (global)
searching is to be done, it will be done on a high level, among relevant
plans, (almost) disconnected from the board. I would like to experiment with
something like this, if I had more time.



- Heikki


-- 
Heikki Levanto     LSD Levanto Software Development   heikki@xxxxxxxxxxxxxxxxx
               "In Murphy we Turst"