![[ru]](/mwf/v12/flags/ru.png "Russian Federation")
...maybe the world could have seen a Go program that plays as strong as humans!
The game of Chess has already been beaten by programmers (chess playing programmers, of course). How long do you think it will take Go to be beaten? My opinion: 10-15 years.
The game of Chess has already been beaten by programmers (chess playing programmers, of course). How long do you think it will take Go to be beaten? My opinion: 10-15 years.
![[lu]](/mwf/v12/flags/lu.png "Luxembourg")
...maybe the world could have seen a Go program that plays as strong as humans!
The game of Chess has already been beaten by programmers (chess playing programmers, of course). How long do you think it will take Go to be beaten? My opinion: 10-15 years.
On the 9x9 board, the strongest Go programs are already fairly good; Mogobot achieved a respectable 1-2 loss in a recent three-game-match on KGS with 7.5 komi and chinese counting against a 5d pro. I think in the case of 9x9-go, it is not unreasonable to expect being able to beat pros regularly in even games within ten years.
On the 19x19 board, things are of course more complicated. The strongest programs presently can hold ranks of 2k(CrazyStone) or 3k (Mogo) on kgs at reasonable time controls, so to reach the level of the strongest humans it will be necessary to gain probably around 13 stones in strength. Assuming that a doubling of computing speed gives you a bit less than one stone with presently known programming techniques, this will take around 30 years to do if Moore's law holds that long. Advancements on the software-side should however cut down on this significantly, so I would guess that 20 years will suffice and that 15 years is a possibility.
However, in 50 years I think Go programs will be able to win every even game of Go against any human opponent, whereas in chess I could imagine that humans will by then still be able to get away with an occasional draw.
Greetings,
Neuromancer
P.S.: In 500 years ;) I think robots will be superior to all humans with respect to every kind of intellectual task which is not easy enough for humans to be able to handle problems of that kind close to perfectly.
The game of Chess has already been beaten by programmers (chess playing programmers, of course). How long do you think it will take Go to be beaten? My opinion: 10-15 years.
On the 9x9 board, the strongest Go programs are already fairly good; Mogobot achieved a respectable 1-2 loss in a recent three-game-match on KGS with 7.5 komi and chinese counting against a 5d pro. I think in the case of 9x9-go, it is not unreasonable to expect being able to beat pros regularly in even games within ten years.
On the 19x19 board, things are of course more complicated. The strongest programs presently can hold ranks of 2k(CrazyStone) or 3k (Mogo) on kgs at reasonable time controls, so to reach the level of the strongest humans it will be necessary to gain probably around 13 stones in strength. Assuming that a doubling of computing speed gives you a bit less than one stone with presently known programming techniques, this will take around 30 years to do if Moore's law holds that long. Advancements on the software-side should however cut down on this significantly, so I would guess that 20 years will suffice and that 15 years is a possibility.
However, in 50 years I think Go programs will be able to win every even game of Go against any human opponent, whereas in chess I could imagine that humans will by then still be able to get away with an occasional draw.
Greetings,
Neuromancer
P.S.: In 500 years ;) I think robots will be superior to all humans with respect to every kind of intellectual task which is not easy enough for humans to be able to handle problems of that kind close to perfectly.
![[si]](/mwf/v12/flags/si.png "Slovenia")
Playing chess or go for coputers is not intelectual task. This will be intelectual whene you could input game rules and whit time computer will become stronger and stronger. Chess engines are at same strenght most of their soft life. Coputers are and will be used to solve specific tasks for humans and not for coputers.
Coputers cant absorbe and process knowlege at same level as human can do.
Coputers cant absorbe and process knowlege at same level as human can do.
I believe this will happen in fifty years, not five hundred.
![[us]](/mwf/v12/flags/us.png "United States")
And in year 2000, we would have flying cars in the streets? :-)
Long-term predictions are notoriously difficult to make. :-/
/* Steinar */
Long-term predictions are notoriously difficult to make. :-/
/* Steinar */
I think in fifty years, we will clearly have the hardware to do strong AI, but I think that the software problem is several orders of magnitude harder and will take a lot of time.
I think it is not even clear that today we are lacking the hardware for disembodied human-level AI, as in principle I could imagine something like Cyc to know basically everything I know about the world and run nicely on affordable hardware. However, I don't see anyone solving the problem of programming computers right any time soon (unfortunately, as I would very much like to see the emergence of superhuman intelligence).
I think it is not even clear that today we are lacking the hardware for disembodied human-level AI, as in principle I could imagine something like Cyc to know basically everything I know about the world and run nicely on affordable hardware. However, I don't see anyone solving the problem of programming computers right any time soon (unfortunately, as I would very much like to see the emergence of superhuman intelligence).
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill