How much AI research has been done since 1975? A lot, right?
Wrong. There has actually been little to no true AI research in the past thirty seven years. We have made great strides in granting machines mastery of discreet skills, such as voice recognition / synthesis and navigating real world environments, but this is not Artificial Intelligence. A chimpanzee can be trained to juggle as well or better as a human can, yet they are clearly less intelligence.
Learning is not intellect. If you place a human--or many species of animal for that matter--in an unfamiliar situation with no prior instruction, it will endevour to find a solution to the problem. Its first attempts will most likely fail, but it will usually figure out something that works eventually, even if that action is less than optimal.
Intelligence is a measurement of how well a being functions when presented will a problem at least partially outside of its pre-existing knowledge and expertise.
There isn't a computer on this planet that can even attempt to solve a problem it was not programmed for. The closest anyone has come has been to create software that is meant to find solutions to issues related to a specific topic, but ultimately if you take a machine outside of its area of expertise it will not just perform less well, it will be flat out paralysed.
It is understandable that computer scientist and engineers have focused on basic skill acquisition. There is little point in creating an AI if it lacks the means to interact with its surroundings effectively. But, while still somewhat crude, robotic senses and motor skills have advanced to the point where construction of an artificial mind can actually be usefull. A change in focus is now called for.
No comments:
Post a Comment