“Game over? Computer beats human champ in ancient Chinese game”
In a milestone for artificial intelligence, a computer has beaten a human champion at a strategy game that requires “intuition” rather than brute processing power to prevail, its makers said Wednesday.
{ // A snip from the article…
Dubbed AlphaGo, the system honed its own skills through a process of trial and error, playing millions of games against itself until it was battle-ready, and surprised even its creators with its prowess.
“AlphaGo won five-nil, and it was stronger than perhaps we were expecting,” said Demis Hassabis, the chief executive of Google DeepMind, a British artificial intelligence (AI) company.
Even if AlphaGo is able to defeat the world champion of Go in a set of games, it still has not “solved” or mastered the game. Or will we treat it that way? Human beings are guided by emotion and feelings, as well as by reason, logic, pattern recognition, history, and calculation. Computers can mimic the latter to make the appearance of mastery of a skill. Without the former, AI will not be real intelligence because the lack of appreciation of nuance and the sense of wonder that occurs when the mind is engaged and grasps the essence of a problem. We are overawed by the tasks that AI can do. The programs have access to databases, which enhance the memory, and power, which is not limited. In all of the matches, the human being is working under a considerable handicap. Defeating a fatigued human being by outsitting him is a dim accomplishment.
One of the annoying habits of the computer people involved in trying to beat human beings in Go, chess, and other games is the glee they exhibit when they try to kill the games. Their interest seems to be not just to solve a problem and add to scientific knowledge, but to destroy our enjoyment in our diversions. Soon they will claim to solve poetry, art, and music so that we don’t have to waste time doing them since a computer can do them of a fashion, too. It is particularly annoying when they refer to human beings as “wetware”, which is considered an inferior element compared to software and hardware. As one computer person told me at a chess tournament, “Human thought is only complex biochemistry, and as deterministic as following an algorithm to its logical conclusion.” A particularly cold, sterile view of man. Then I saw him get crushed by a little kid with bright eyes and the whole world in front of him. Maybe there is still hope for us.
Someone raised the following question the other day, “When those in AI destroy the tens of millions of jobs that people do every day by 2030, what will they do when they are the only ones who will be taxed to provide income for the rest of humanity?” Google recently purchased DeepMind, the company that produced AlphaGo.
I think you are being overly inclusive in your use of the term “computer people”. Yes, some may think that way, but people, even computer people, are different and look at things in different ways.
I think tmagchesspgh is being over fatalistic. This all part of the advance of civilization. Steam shovels replaced manual earth movers with steam shovel drivers. Machines can lift many times more weight that a person can, and yet, weight lifting is still an Olympic sport. A bigger fear that I have is everyone getting to be more sedentary. This is especially true for many chess players. So between rounds, go for a walk, take a swim, hit the gym, play some golf or tennis. A healthy body plus a healthy mind equals success at the chess board!
This is an amazing accomplishment, but I still don’t think the go project has made that much progress in “thinking like a human.” Does the computer have intuitive pattern recognition superior to the best humans? Not really.
Hubert Dreyfus’s critique of AI is still relevant. You can program your engine to recognize this context as drawn (we humans all see the fortress draw immediately, and all decent software programs recognize this position as drawn, too):
But will you bother to program your engine to recognize the following context? (I saw a similar example on Rybka Forum a few years ago and have been quoting it since then. The curious may feed this position to Stockfish 7 and ask it for its evaluation.)
And even if you did teach your engine to cope with this second circumstance and thousands of analogous positions, humans will cope with these unusual contexts much better than computers. If our brains had the processing power and the memory, we’d still be better players than the computers. (But we don’t, and we aren’t.)
Having said that (sticking to simpler games at which I’m slightly less incompetent), playing backgammon with the neural net program Jellyfish taught me that I should be bolder in creating blots on the inner table. And I’m sure that the great go masters will learn much from working with engines, just as our grandmasters have.
More to the point, this is a step towards true AI. These programs trying to figure out how go should be played are a bit like Steinitz in the 1870s. After he’d become the world’s strongest player, he tried to figure out the rules governing chess.
I found this pigeon-sourcing project much more stunning. Could three pigeons in consultation soon be playing chess better than I?
Google just announced a Go match with the present world champion for $1 million. It will take place on March 9-15. The match will be shown on youtube. Some critics have said that the DeepMind, which has produced AlphaGo is better at marketing than AI. It is surprising that the match was set up so quickly and for so little in prize money.
If the program loses, the computer people will complain, as usual, that it was a hardware problem. They will come back to play again and again until they win. Then they will duck any future match since they will have “solved” go and are on to other profitable ventures. If the world champion loses, it will be a blow to the game which is very popular in Japan and China. They will probably take it a lot worse than the chess world did when Kasparov lost to Deep Blue.
AlphaGo leads 2-0 in the match with Lee Sedol. 33 year old Sedol is the present Go champion and ranked #4 in the world. Before the match, he said he would win the match 5-0 or 4-1. The winner of the match receives one million dollars. (Cue Dr. Evil) After Go, the AI people are setting their next sites on poker. They understand the math; the dirty, sleazy psychology of the game is what they are working on now.
The match is over. AlphaGo defeated Lee Sedol 4-1. A tearful Sedol said, “I failed.” Before the match he was very confident, claiming that he would defeat the program 5-0 or 4-1. The program took a 3-0 lead before Sedol finally won the fourth game in a four hour duel. The fifth game started with an early “mistake” by the program. Sedol could not hold his advantage and the program made a relentless comeback to win the final game. A $1 million dollar prize went to the winners.
AlphaGo is produced by DeepMind, a company that was recently purchased by Google. DeepMind has been researching neural networks to apply them to decision strategies. The complex process is expected to be applied to medicine and other fields. More likely, it will be used for military projects. The DeepMind people say that they will spend several months studying the games of the go match to see where they went right and wrong before publishing their findings. This match is likely the last the AlphaGo people will play. Now they will turn to practical things.
The Go community was astonished by the amount of interest and coverage of the match. The final result was stunning. It is not clear what will happen to Go as a game which is widely practiced in Asia among amateurs and professional players.