Communications of the ACM
Why Can't My Computer nderstand Me?
A computer scientist says there are major flaws in what currently is considered artificial intelligence and the way in which it is measured.
Credit: ChasingAMiracle.com
University of Toronto computer scientist Hector Levesque recently presented a paper at the International Conference on Artificial Intelligence highlighting the flaws in artificial intelligence.
Although Alan Turing's renowned Turing test has been accepted for years as a measure of computer intelligence, Levesque says computers can easily bluff their way to success in the test. Rather than performing tricks to pass an arbitrary test, AI should focus on building true intelligence, Levesque says. He suggests an alternative test called the Winograd Schemas that he developed with his colleagues, which asks questions that are simple for an intelligent person but extremely challenging for a computer. For example, the test might ask, "Sam tried to paint a picture of shepherds with sheep, but they ended up looking more like golfers. What looked like golfers?" This type of question, impossible to answer using a search engine, requires knowledge of social interaction, common sense, and complex language understanding.
The field of AI is distracted by "serial silver bulletism," or always looking to the next big thing, instead of performing the necessary work of unraveling the complexity of ordinary human intelligence, Levesque says. "There is a lot to be gained by recognizing more fully what our own research does not address, and being willing to admit that other...approaches may be needed," he says.
From The New Yorker
View Full Article
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA
No entries found