Getting badly off topic here... Anybody want to start a new topic?
Computers can already solve othello. When I was playing othello semi-seriously I used to set the computer to 'hardest' which would mean it would solve the game with 24 moves to go and display the final score. After every move it would adjust the final score so I would know I made a mistake. Since it is trivial to start with an opening database of 24 moves, that means 48 of the 60 moves are solvable. Furthermore, this was ten years ago, since then computers have increased in speed by approximately 500x times I would estimate instead of solving 24 moves, they could instead solve 28. The opening book can be even larger since it could be precomputed, I would expect an ordinary 3GHz PC to be able to play perfect othello (i.e. no heuristics). Increasing to 10x10 puts othello well outside 'strongly solved', but the quality of the heuristic: "Minimise my opponent's moves" is so good in othello if you're looking 20 ply that it must still be extremely hard to beat. Incidentially, this was part of the reason I gave up othello -- why try so hard to master a game when you can never beat the machine.
Returning to Chess, there is no computer program which has solved chess. Furthermore, I do not expect chess to even be weakly solved (from the starting position) for at least another 20 years -- and that is assuming custom-designed hardware. However, computer chess now has the same feel to it as othello -- If I dedicate my life to it, will I ever be able to beat a computer program I write?
As for go, sadly I cannot write off computers so quickly -- though some of the programmers have (see Fortland's opinion for instance). I have seen a computer program that can play professional level early fuseki, but it is playing it at a smart 'monkey' level, in that it will absolutely fall apart in the late fuseki/early midgame.
Furthermore, perfect tsumego and perfect yose are entirely possible. Perfect tsumego is already integrated with at least one go playing program, though efficency reasons mean it _currently_ makes mistakes. Now, I could say the mistakes make it quite weak -- I can read how to save groups it classifies as dead on its hardest setting, but throwing more resources at the program will make it that much harder.
There has really been very little serious research into go playing outside CJK (I could probably name every western researcher off the top of my head :-(). But what will the future bring to go? I have my own ideas about how to write an extremely good go playing program, and I'm pretty confident I'll manage to build a professional level 9x9 playing program. As to whether it will scale to 19x19, well I'll give that a go if/once I get 9x9 to profesional level, but the technique I'm thinking of is not especially suited to scaling, so don't hold your breath. However, that's just me, who knows what some brilliant researcher will think of elsewhere?
I really appreciate that go playing programs suck -- it increases my enjoyment of the game. I also appreciate that when I sit down to a game on the net I know all my opponents are humans (or else they're about to get their backside kicked ;-).
I also note that computer assisted go is a field which is developing extremely fast. YuBin himself (my personal go hero) has done some good work in this area. For instance, imagine a go program with kogo's joseki dictionary integrated, of little use to a computer that doesn't understand the implications of a joseki and of no use to a professional, but to an amateur dan player it could help quite a lot. I've also downloaded a very early prototype of an (automatically derived) hamete refuting dictionary -- too crude to be useful _yet_ but a fascinating direction for research.
Then there are score estimation tools, if you set the computer to play out the game and it says you will lose by 10 points, then you know to play differently than if it says you will win by 10. As computer play gets stronger so you can trust the estimate more, this will become incredibly useful -- I already use it when reviewing games for beginners to show them that the game is so far lost or won that they should either resign or simply settle.
And what about fuseki? I think GoGoD and gobase are both far, far better than the encyclopedia of fuseki -- I know I've used them to find mistakes in my fuseki which would've been impossible using the encyclopedia. For instance, Isheida might say that some joseki should be played for top influence, but with gobase you can see exactly what the pros have chosen in similar situations.
Tsumego problem tools are getting better too -- Imagine a tool which custom-designs/chooses problems based on areas you are weak? Perhaps you're an ace at ishi no shita, but just can't spot a sacrifice of half your group. Or imagine if you're playing a game on IGS, and your client helpfully points out that some group can be reduced to one eye, forcing it to run? A disgusting cheat, but will that stop its availability?
As for yose, the same point applies. Imagine that yose tool I talked about picking a random pro game, skipping to the point it can play perfectly and failing you if you ever make a mistake -- I think it would revolutionise late yose virtually overnight! Or taking over for you on IGS to coast to a 0.5 victory...
*SIGH* Go has had so few cheats in the past since the computer tools have sucked. As they get useful it will be interesting to see if honour can keep these tools out of the internet clients and only in the study tools. I fear not.
PS: My phd is in language processing, not game playing so I'm not totally up with the field.
Little Golem