Turing Test and Artificial IntelligencePosted: November 22, 2011
The Turing Test is a key test of artificial intelligence: can robots fool humans into thinking they are intelligent? Despite optimistic projections for AI (e.g., Herbert Simon made some wild predictions),AI still underwhelms. Well, in some areas (in chess, for example, computers beat humans).
Perhaps the Turing test isn’t the right measure of ‘intelligence.’ But chatbots have yet to fool humans that they actually are human. The yearly Loebner prize puts this to the test, here’s the 2010 winner Suzette, or the 2011 winner Rosette. Chat with either of the bots, or any other for that matter, and you’ll quickly see the problems. Over at orgtheory.net we have ‘tested’ the winning chatbots several times – and inevitably they fail.
The clip below is a game show with some AI bots (ol’ Eliza versus Deep Blue versus an evolutionary algorithm – ok, it’s not the actual bots).
The best clip still is the chatbot v chatbot discussion at Cornell.
Here’s Ray Kurzweil a few weeks ago (at the 2011 Singularity Summit) talking about “from Eliza to Watson to passing the Turing Test.”