The project is interesting, the approach is funny *, but what passed there at the university is only one of the variations of the turing test that was passed.
Professor Warwick felt that the test passed was the best, well, ok. You can not agree with him.
Very surprised by his statement. He himself knows that in all modern tests there are no established themes.
And in fact, 33% means that there were only three judges, and only one was deceived.
In the case of the Löbner Prize, for example, 2013, there are four judges, and these are professionals on the subject.
In the case of the test, from what is evident from the press release, among the judges were “actor Robert Llewellyn” and “Lord Sharkey”, who have an indirect relationship to robotics, even though the first one played a robot, and the second was related to Turing. With this approach, a visual turing test for android robots could be performed with Schwarzenegger and Summer Glow as judges, and not with Professor Hiroshi Isiugro or David Henson.
Moreover, so far there are no published results, except for the press release (as opposed to the full background in the case of Löbner test), and everyone shouts about how the test is passed, as if they had not been there before.
(* The approach is funny, but, as already noted in the comments, it can be counted as a scam made specifically for turing testing.)
The comments are logical, but before making assumptions about the quality of the test, one could follow the link in the post and read about it. What could be found:
- 30 judges
- total 300 5-minute correspondence
- every 5 minutes the judge spoke simultaneously with the car and with the person
- the test was carried out at an independent examination of a friend of the Professor Professor John Barnden, University of Birmingham, formerly head of the British AI Society
Source: https://habr.com/ru/post/225599/
All Articles