In the field of artificial intelligence (AI for short), scientists are constantly working to give a computer the ability to reason. So far, there hasn't been much success. A new system from Google is now tackling a neat idea: trying to figure out where any particular picture was taken. They call their new project PlaNet, and it's a neural network AI (basically, it's a primitive brain, which can make all sorts of connections and basically program itself). Technology aside, the premise is asking how we humans recognize places and doing the same from a machine perspective. You can try a game about this idea at www.geoguessr.com.
I'm not particularly worried that computers will overtake humans in the world of thinking. Yes, computers are amazing at calculations, and geo-locating is just that, but reason is beyond programming. Logic, cold and hard facts, massive databases, blitzing to the answer, these are things computers do well. I read that just recently a proficient GO program was beating most players. That's excellent, yet the program is only examining all possible moves rather than intuiting the best flow for the game (as a human does). It's called brute force computing (throw enough power and resources at the problem and you will eventually find the best answer).
Until we can give computers any kind of intuitive ability to leap beyond facts and to an answer, computers will never reason. But, there is danger in trying, because we have a word for people who can reason but have no morality--sociopaths. Computers have no morals, holding only what has been specifically programmed into them. That's why I'm not worried about robots taking over. At some basic level no human wants to make such a beast. Instead, we make intelligent tools for specific tasks and watch in awe as they perform with perfect ability.
- M
No comments:
Post a Comment