Tags

, , , , ,

An article in The New York Times by Steven Strogatz titled One Giant Step for a Chess-Playing Machine brought up the computer currently known as AlphaZero. Among other things, it can play chess really well, but the important thing was that no one taught it how to play chess. AlphaZero learned how to play using machine-learning algorithms where it was simply exposed to large numbers of the problems it was supposed to learn how to cope with. It learned chess, shogi, and Go that way. It was set to competing with Stockfish, which was the best of the current generation of chess programs, but Stockfish plays with the simple brute force of comparing millions of potential moves. AlphaZero instead plays with insight and strategy, and even when given ten times the amount of thinking time Stockfish still didn’t win a single game of one hundred played. AlphaZero wasn’t programmed or coached by any human players, with a thousand years of chess cultural history; no, it figured our elegant moves by its self and even discovered ones that were unknown to the current chess masters.

What’s next? The Times article states “By discovering the principles of chess on its own, AlphaZero developed a style of play that ‘reflects the truth’ about the game rather than ‘the priorities and prejudices of programmers,’ Garry Kasparov wrote… The question now is whether machine learning can help humans discover similar truths about the things we really care about.”

What’s next? A year ago I was constantly swearing at my smartphone because I would ask Google what seemed to me at least like clearly pronounced and clearly stated questions, and it would send me down the rabbit hole into a vast wonderland of wrong answers. Today, while chatting with my dudes, a couple of times I asked not particularly well-defined questions, such as about an artist whose last name I couldn’t remember, “Russian icon-artist Olga living in Oregon“. BINGO in half a second, I clicked images because that is what I wanted to show to my artist friends, and once again BINGO! If that’s the kind of progress we can expect in one year what will happen in five years?

What’s next? Last week, Scott Pelley on 60 Minutes TV interviewed Kai-Fu Lee, the head of China’s largest computer chip manufacturer and asked that question. The answer, “In about ten years 40% of all jobs will be replaced by computers.” If you listened carefully, you could hear a billion jaws drop in amazement and horror, “Am I going to be replaced?”

What’s next? is one of my favorite questions, and my answer to the Artificial Intelligence (AI) problem is that we are a transitional species. We are moving from an organic form of intelligence to an inorganic, basically silicon form of self-conscious existence. In our current form we have a life expectancy of a hundred years but in our AI form, we have a consciousness that can last for billions of years. Our choice is if we would rather die forever in a couple of years, or live on in a vast web of consciousness for billions of years and then die.

Who can know what would be more pleasurable? But one thing is certain: if we die we will never have the opportunity to find out.