The Turing Test is considered the reference test for AI systems. One person (the jury) communicates with two other subjects - one being another person, the other being a computer running an AI program trying to simulate human behavior.
If the jury cannot tell which is which, it is said the computer has passed the test.
Up till now, no computer was ever able to full the jury, but we got close this Sunday!
The test determines the computer needs to fool at least 30% of the jury, and in this case the winning machine was "only" able to fool 25% of them - just 5% more to go.
Skynet is getting closer... :)
[Update]
If you want, you can play with elbot yourself, (the program that nearly won this competition.)
No comments:
Post a Comment