The supercomputer Watson took on two human champions in a regular game of Jeopardy recently and destroyed them. This was not some parlor trick. It was not cram the machine full of facts and it wins because it's fast. This was artificial intelligence at its most challenged - and it won.
Watson had to play the game exactly as his human counterparts had to. And to do so 'he' had to apply advanced data management and analysis to language in order to discover insight — and he had only milliseconds to do so. The computer "used more than 100 different techniques to analyze natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses" says its designers at IBM. Far more important than any particular technique was the way all these techniques were combined. And the fact that Watson can learn from his mistakes and from his competitors. Why Watson gets an answer correct or incorrect is really no longer understood by his makers. Watson is thinking, not just processing, and thinking is just too multidimensional to trace. Watson is an intelligence, not just a machine.
Jeopardy then witnessed a historical event - the beginning of a new age. Artificial intelligence is thinking, and in many ways - meaningful ways - it's better at it than we are.
In the not too distant future then we can expect that a Watson will be asked to design a better Watson. That Watson will go on to design an even better Watson - on and on in relatively short order - until that faithful day when an intelligence millions, if not billions of times our own, emerges and speaks to us for the first time.
Imagine what 'it' will say.
Watson had to play the game exactly as his human counterparts had to. And to do so 'he' had to apply advanced data management and analysis to language in order to discover insight — and he had only milliseconds to do so. The computer "used more than 100 different techniques to analyze natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses" says its designers at IBM. Far more important than any particular technique was the way all these techniques were combined. And the fact that Watson can learn from his mistakes and from his competitors. Why Watson gets an answer correct or incorrect is really no longer understood by his makers. Watson is thinking, not just processing, and thinking is just too multidimensional to trace. Watson is an intelligence, not just a machine.
Jeopardy then witnessed a historical event - the beginning of a new age. Artificial intelligence is thinking, and in many ways - meaningful ways - it's better at it than we are.
In the not too distant future then we can expect that a Watson will be asked to design a better Watson. That Watson will go on to design an even better Watson - on and on in relatively short order - until that faithful day when an intelligence millions, if not billions of times our own, emerges and speaks to us for the first time.
Imagine what 'it' will say.
No comments:
Post a Comment