from Hacker News

GPT-4 Passes Turing Test. Humans Often Mistake Each Other for AI

by biscuit1v9 on 5/17/24, 5:07 AM with 16 comments

• by tripletao on 5/17/24, 7:23 AM

  The "decoder" article says:

  > The researchers defined 50 percent as success on the Turing test, since participants then couldn't distinguish between human and machine better than chance.

  54% of GPT-4 conversations were judged to be human, so the "decoder" article says the Turing test has been passed--indeed, it seems more human than human. But the paper says:

  > humans’ pass rate was significantly higher than GPT-4’s (z = 2.42, p = 0.017)

  The seeming discrepancy arises because they've run a nonstandard test, in which the meaning of that 50% threshold is very hard to interpret (and definitely not what the "decoder" author claims). The canonical version of Turing's test is passed by a machine that can

  > play the imitation game so well, that an average interrogator will not have more than a 70 percent chance of making the right identification after five minutes of questioning

  The canonical experiment is thus to give the interrogator two conversations, one with a human and one with a non-human, and ask them to judge which is which. The probability that they judge correctly maps directly to Turing's criterion. If the two conversations were truly indistinguishable, then the interrogator would judge correctly with p = 50%; but that would take infinitely many trials to distinguish, so Turing (arbitrarily, but reasonably) increased the threshold to 70%.

  That doesn't seem to be the experiment that this paper actually conducted. They don't say it explicitly, but it seems like each interrogator had a single conversation, with a human with p = 1/4. The interrogator wasn't told anything about that prior, leading them to systematically overestimate P(human). If every interrogator had simply always guessed "non-human", then they'd collectively have been right more often.

  Even if the interrogators had been given that prior, very few would have the mathematical background to make use of it. GPT-4 is impressive, but this test is strictly worse than Turing's, whose result has clear and intuitive meaning.

• by anonzzzies on 5/17/24, 6:13 AM

  I would pick the one that is the most polite and makes the least syntactical and grammatical errors to be the AI always; most humans are absolutely terrible at formulating anything coherent so it's a safe bet that the ones who actually form correct sentences are AI. Many (most I see online or on the street) humans talk like Markov Chains (check your kid's snapchat for examples) or, at best, very early transformers (tons of repetition, getting stuck and not all that coherent).

• by somenameforme on 5/17/24, 6:34 AM

  Ugh, these modern "Turing Tests" are a complete bastardization and dumbed down version of what Turing described. Here is his original paper. [1] In short the actual task involves a skilled interrogator, somebody of a given and specific identity, and then somebody pretending to have that identity. Turing proposed a simple example where you'd have a woman, and then a man pretending to be a woman. The more precise the identity, the more challenging the test becomes. A man might kind of sort be able to pass for a woman in text, but he'd never be able to pass for a nuclear physicist who has a twin brother working in neuroscience, against a skilled interrogator. And all participants are expected to actively collude and collaborate as much as possible to emphasize who is the "real" person. So for instance the woman might propose to help the interrogator by proposing questions he could use to help spot the fake, and/or to emphasize her own authenticity.

  Modern takes generalize the identity to absurdity (with the identity being human or not), generally feature idiots (or people acting like such) for interrogators, and participants who are actively trying to act like a computer to trick the interrogator. Like in this article, the human is B and was asked, "What could you say to convince me that you're a human?" His response was "You just have to believe!" Why not just skip the pretext and just have the human start responding 01001001 01000010 01101111 01110100 01001100 01101111 01101100 to every question? And if all this nonsense wasn't enough, they bumped it up to 3 comps and 1 human pretending to be a comp. This isn't the Turing Test - it's complete LARPing!

  [1] - https://redirect.cs.umbc.edu/courses/471/papers/turing.pdf