5 ms·

That is not correct though. With the question phrased this way the second and third logicians don’t know if the previous ones want a drink or not.

dmurray 4y ago
They should know. When the first logician answers "I don't know" it makes it clear that he does not want a drink.
- leereeves 4y ago
  That is true, but GPT-4 wrote: > The second logician hears the first one's answer and concludes that the first logician must want a drink, because if he didn't, he would have said "no" > The third logician hears both answers and realizes that since neither of the first two said "no," they both must want a drink. IOW, when asked the puzzle with "anyone", it repeated the explanation for "all", which seems to support the hypothesis that even GPT-4 is still just a "stochastic parrot" repeating what it sees and unable to reason about slight variations.
benj111 4y ago
Yes they do. The question is if any of them want a drink. Ie if at least one of them wants a drink. If the first person wants a drink then they know at least one person wants a drink and so could answer "yes". If they don't personally want a drink they don't know if either of the other 2 might want a drink so they would reply "I don't know"
- teawrecks 4y ago
  Correct. The original phrasing is a question of universality, the new phrasing is one of existence. Gpt4 continued to explain the joke in terms of the former when it should have been able to adjust for the latter. It has never heard the latter, but it sounded close enough. On the upside, this makes for a good way to test the logical abilities of a LLM. On the downside, whatever part was considered a "joke" before has been completely obliterated.
m463 4y ago
Humor is only coincidentally about being correct, since we all know horses have fore-hooves and two rear hooves, leading to the funny fact that horses have six hooves.

dmurray 4y ago

They should know. When the first logician answers "I don't know" it makes it clear that he does not want a drink.

leereeves 4y ago
That is true, but GPT-4 wrote: > The second logician hears the first one's answer and concludes that the first logician must want a drink, because if he didn't, he would have said "no" > The third logician hears both answers and realizes that since neither of the first two said "no," they both must want a drink. IOW, when asked the puzzle with "anyone", it repeated the explanation for "all", which seems to support the hypothesis that even GPT-4 is still just a "stochastic parrot" repeating what it sees and unable to reason about slight variations.

benj111 4y ago

Yes they do. The question is if any of them want a drink. Ie if at least one of them wants a drink. If the first person wants a drink then they know at least one person wants a drink and so could answer "yes". If they don't personally want a drink they don't know if either of the other 2 might want a drink so they would reply "I don't know"

teawrecks 4y ago
Correct. The original phrasing is a question of universality, the new phrasing is one of existence. Gpt4 continued to explain the joke in terms of the former when it should have been able to adjust for the latter. It has never heard the latter, but it sounded close enough. On the upside, this makes for a good way to test the logical abilities of a LLM. On the downside, whatever part was considered a "joke" before has been completely obliterated.

m463 4y ago

Humor is only coincidentally about being correct, since we all know horses have fore-hooves and two rear hooves, leading to the funny fact that horses have six hooves.