I finally figured out a frame work for a (maybe the) hardware that is specifially to play Go. In the past I have been strategically acquiring certain skills so that I can build a an embeded system or a computer interfaced system in my basement. (Well I live in a cold climate, so the garage workshop is not a good idea. I think it's possible that a garage workshop holds advantage over a basement workshop. This may be the reason why silicon valley happened in California, not in, for example, Minnesota or Alaska.) These skills include digital circuit design, PCB layout, PCB manufacturing, PCB assembly (mounting a microprocessor on a PCB is not trivial). This Go computer can be built using COTS (Commercially Available Off the Shelf parts - a defense industry term). This avoids the using of ASICs which carrys a price tag of half million dollar minimum. It can be expected to search half to one million nodes per second on a whole board scale. Some people may wonder what can one million node per second do? It can solve the carpenter square by brute force in several minutes. How well can a one MID (million nodes per second) Go computer do against humen? With a well organized Go knowledge database I think it can play at the level of amateur 3 to 4 Dan.
Many people have observed that computer Go is not advancing as fast as in computer chess. One of the reasons is that it lacks the organized efforts in computer Go. Can anyone organize such an effort? If you do, I'm willing to quit my steady job in the electronic industry and work for you as an research associate for 3 to 5 year.
Daniel Liu
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go