Discussion about this post

User's avatar
Michael K.'s avatar

You guys are amazing! Crush the hype! Expose the parrot! :)

Expand full comment
Miguel I. Solano's avatar

Re "Melanie Mitchell gives an example of an MBA test question where changing some details in a way that wouldn’t fool a person is enough to fool ChatGPT (running GPT-3.5). A more elaborate experiment along these lines would be valuable." — A quick experiment with GPT-4 suggests that, unlike ChatGPT, GPT-4 is far less sensible to changes in wording. In our (brief) experiments with the same prompt Mitchell used, GPT-4 output the correct answer every time.

Expand full comment
19 more comments...

No posts