![]() But as early as 1991, a bot called PC Therapist created by Joseph Weintraub took part in a Turing test and fooled 5 out of 10 judges into thinking it was human – a pass rate of 50 per cent. How exactly to translate Turing’s ideas, conceived long before chatbots existed, into a meaningful test of artificial intelligence has been a matter of debate for some time. Sounds sort-of impressive, though with just 30 judges, only a handful need to be fooled to produce the improvement. In the most recent tournament, at the Royal Society in London, which involved the same number of tests and judges, Eugene upped his score by 4 percentage points. In that tournament, Eugene fooled judges 29 per cent of the time – just 1 per cent shy of the 30 per cent needed to pass. ![]() With 150 separate conversations, or tests, carried out by 30 human judges, it was billed as the “most statistically significant” Turing test tournament yet to take place. We had to sit at computer terminals and chat with invisible chatbots and humans via an instant-messenger-like program. When I met Eugene, I was acting as a judge in a Turing test at Bletchley Park in Milton Keynes, UK, where Turing helped to crack the Nazi Enigma code, that was organised to mark the centenary of Turing’s birth. Turing said that a machine that fooled humans into thinking it was human 30 per cent of the time would have beaten the test. ![]() First conceived by the legendary Alan Turing in the early 1950s, the test challenges human judges to converse via a text interface with both hidden bots and humans – and say in each case whether they are chatting to a human or machine.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |