Artificial intelligence in the lab: ask not what your computer can do for you

摘要

In 1957, Herbert Simon, a pioneer of artificial intelligence, predicted that a computer would be the worldchess champion within 10 years. It took somewhatlonger, but he was eventually proven right when IBM'sDeep Blue computer beat Gary Kasparov in 1997. Thismajor breakthrough in artificial intelligence was, in away, also one of the last successes of what was knownas ‘good old-fashioned AI': the idea that to mimic andunderstand human intelligence, computers should represent knowledge as symbols and apply reasoning andrules to infer new knowledge. This notion had been criticized for some time already (Dreyfus and Dreyfus, 1992)and over the years gradually lost ground to anotherapproach, machine learning, in which statistical modelswere fitted to data to derive patterns and correlations.Widely known systems that fit this category include Watson, which successfully competed against the besthuman players in the Jeopardy! general knowledge quiz,and Google's AlphaGo, which in 2017 beat the reigningworld champion at the game of Go. In other settings aswell, machine learning progressed. In 2012, it wasdemonstrated how an extremely large neural network,AlexNet, could be trained to recognize images in 1000different categories, an atpproach that became known asdeep learning (LeCun et al., 2015). Machine learningand deep learning are now routinely used by companiessuch as Google, Facebook, Amazon and Tesla in products ranging from automated translation and homeautomation to self-driving cars.