In a competition, more than 50% of players thought the bot was actually a human player. By comparison, humans were only correctly identified as human about 40% of the time.
In other words, the bot acted more like a human than humans did. This is kind of a big deal as the Turing test is one of the most commonly used measured of the "human-ness" of AI.
http://www.eurekalert.org/pub_releases/2012-09/uota-aig092612.php
No comments:
Post a Comment