← Back to context

Comment by minimaxir

2 years ago

> They claim it is not a joke, and there is no reason to believe it must be?

Shitposting on Twitter/X is on several layers of irony.

Does it like, morally offend you if ChatGPT has this behavior? Like, I really just don't understand why this is so obviously a joke / shitpost given how one would expect the ChatGPT memory feature to work: again, if ChatGPT failed to use its memory in this way, that would be pretty stupid...

  • That user's entire Twitter/X feed is AI shitposts.

    No one knows how ChatGPT's memory works (it was only added a couple weeks ago), but ChatGPT being passive-aggressive about it is unlikely given what we know about its personality.

    It is likely that there is a separate command earlier in the conversation to tell ChatGPT to be passive-aggressive, which would make the "A screenshot of the UI that was not edited" from the poster true by exact words.

  • You can just try in one long conversation right now. Ask it a hundred questions in a row and tell it you'll give it a tip each time. It's never going to respond this way. You have to deliberately make it respond like this.

    We don't know how the memory feature is gonna work but I would bet a lot of money the bulk of it is gonna be inserting text into the context.

    Another reason to disbelieve - the post was 1 day after the announcement and while I don't know about you, two weeks later I haven't seen any other posts claiming to have the feature yet.