Despite the considerable efforts to build a machine with artificial intelligence there has been little real success. Fifty years ago Turing elaborated a test that could be performed on a machine (robot/computer/software) that would determine if they possessed human-like intelligence. Ray Kurzweil recently predicted that a computer will have the raw computing capacity of the human brain by the year 2020 and that of the entirety of collective humanity by the year 2050. There are many detractors that consider that to be overly optimistic and yet others who doubt any machine will ever achieve human-like intelligence.
In 1770 a clever showman toured Europe with a machine he had built called The Turk. This exotic looking Ottoman automaton took on all challengers and was able to beat nearly anyone at chess. It was revealed many decades later that inside the machine was hidden a compact human chess master who was covertly controlling the machine and beating his opponents with apparent ease. Instead of Intel Inside, the logo on the side of the machine should have read Human Inside.
Chess playing computers have come a long way since then. The machine Deep Blue is now the best chess player in the world having beaten Gary Kasparov a few years ago. It can compute staggeringly large numbers of possible outcomes to decide on its next move, eventually wearing its human opponent down with its sheer number-crunching capabilities. Machines are really great at doing such calculation intensive tasks. Banks, insurance companies and other large organizations were the first to use artificial intelligence to do such work in a fraction of the time required to do the job by the Bob Cratchits of the world.
Recent advances in computing power have closely followed Moore's Law and attained efficient number-crunching capabilities approaching petaFLOP speeds. So one would guess that there were nothing left that a machine wouldn't be able to compute better than a human. What about looking at a picture and determining whether anyone in that picture is eating pizza? This is an example of a task that artificial intelligence has a real hard time doing. Any child could pull it off and so could several clever gorillas and orangutans.
So what are you to do if your company needs to sort through thousands of pictures to find pictures of people eating pizza? Artificial intelligence doesn't seem to be up to the challenge so what is required is artificial artificial intelligence or human intelligence. Just like that touring renaissance chess machine the computer requires the assistance of Human Inside architecture. So Amazon is making such a service available and they have cleverly called it the Mechanical Turk. If you've got a few spare cycles you can sign up and take on such simple work and be paid for it at the rate of a few cents per task.
Human intelligence is the new artificial intelligence.
3 comments:
ottoman automaton! nice.
does kurzweil really claim that processing power = intelligence?
Kurzweil is one of those curious people who eagerly and optimistically anticipate the future of artificial intelligence. He does seem to imply that given enough computing power, one could arrive at a human-like intelligence. I question that approach since human-like intelligence seems to be of another variety but perhaps massive computing power can result in something highly indistinguishable from our intelligence. I think Kurzweil simply doesn't think there would be any important difference and moreover the blending of cyborg technologies with our human substrate would blur the lines between them anyway.
For an interesting view that is contrary to Kurzweil see the Unabomber Manifesto.
heh. or searle (another view contra kurzweil).
Post a Comment