Sign In

Communications of the ACM

ACM TechNews

Smart Guide to 2012: How Best to Test Machine Iq

Alan Turing

Did Alan Turing's 1936 paper 'On Computable Numbers' influence the early history of computer building?

Artificial intelligence (AI) programs did not fare well in convincing judges at the 2011 Loebner prize competition that they were people. The contest is based on the most famous benchmark of machine intelligence, the test conceived by Alan Turing as he reflected on the possibilities of a thinking machine.

The new year marks the centenary of Turing's birth, and there likely will be more competitions in his honor. However, some experts say the Turning test may not be the best gauge for today's AI programs. The test may be too difficult because it requires a program to capture the nuances of human speech. Linguistic smarts also is a narrow focus now that AI bots are influencing the stock market, landing planes, and may soon start driving cars.

One alternative is to use a suite of mini Turing tests, each designed to evaluate machine intelligence in a specific area, such as a new visual test to gauge the ability to understand the spatial relationships between objects in an image against that of a human. Some observers want to use a universal, mathematical definition of intelligence to prevent human bias in testing.

From New Scientist
View Full Article

Abstracts Copyright © 2011 Information Inc. External Link, Bethesda, Maryland, USA 


No entries found