5 ms·
Of course not. Turing's test isn't supposed to be quantified universally over examiners. The criterion is "does there exist a human examiner who can tell the di
by noqc 3y ago
Of course not. Turing's test isn't supposed to be quantified universally over examiners. The criterion is "does there exist a human examiner who can tell the difference between talking to a computer and the human".
I would argue that Turing's test has actually become less dead. It can now additionally be used to establish someone's qualifications to identify reasoned explanation. The computer isn't capable of engaging in it (yet), so being unable to tell if you're talking to a computer generally means that you are incapable of distinguishing bullshit from reasoning.
- ethanwillis 3y agoI'm just reminded of an old Invader Zim episode where Zim harvests and consumes(integrates) lots of human organs to "be more human." He ends up having like 5 livers, 20 spleens, etc. Just absolutely ridiculous. The school nurse when examining him says "perfectly healthy with such plentiful organs." That's where LLMs are at. "Perfectly human with such plentiful words."
- kevinventullo 3y agoIt can now additionally be used to establish someone's qualifications to identify reasoned explanation. The test subject in this case is now the person instead of the algorithm. Quite the paradigm shift I would say!
- noqc 3y agoIn addition, to paraphrase my friend, the turing test was dead when computers could reliably tell the difference between humans and computers. It was utterly useless as a measure before, but now it is genuinely alive.
- SAI_Peregrinus 3y agoYep, the original Turing test is adversarial: There's a computer subject trying to act human, a human subject trying to act human, and an examiner trying to tell which of the subjects is the human. If there exists an examiner which can distinguish the subjects with probability non-negligibly different from 0.5, then the computer fails the test.