I'm aware that you're a different commenter but you are addressing yourself to a comment that was in reply to them and therefore not necessarily appropriate to measure such a comment against entirely new criteria that you want to bring into the conversation.
Despite your protestations to the contrary, these descriptions seem perfectly fine in that they're accurate and meaningful. And it if you want to start getting fast and loose with all kinds of new extra criteria and requirements for what it's supposed to do, they all seem squarely within the reach of the capabilities on offer, with some prompt tweaks.
>these descriptions seem perfectly fine in that they're accurate and meaningful
The description of Wicked doesn't mention either The Wizard of Oz or the Broadway musical. So yes, the descriptions don't contain obscene mistakes like calling Wicked a courtroom drama. If that is enough for you to call these "accurate" while ignoring the vagueness or the 1 in 10 failure rate on the Anora description, fine by me. But you must have some weird definition of the word "meaningful" to apply that to descriptions like the one of Wicked. That simply isn't a helpful way to describe that movie.
The comment thread you're at the end of started with this:
> I would expect nothing but hallucinations and nonsense coming out of any LLM regarding recently-released movies (aka. the ones you often find on flights).
The comment that replied to it (the one that you're arguing against) provides evidence that proves it wrong. You are correcting someone who isn't incorrect, and I think the person you're responding to is very justified in saying you're moving the goalposts here.
A reply downthread is not an endorsement of everything said upthread. I'm happy to discuss the points I made, but I’m not going to be made to defend something I didn’t say.
Well if you're not endorsing what was said upthread, then your comment is a complete non-sequitir. The parent comment said "LLMs can't give movie recommendations for recent movies because they'll hallucinate or spout nonsense", the next comment responds with a list of accurate movie recommendations, and then you come in and say this:
> Those descriptions are less detailed than the information you will see on basically any streaming interface and yet it still manages to not being very good.
The points you made were not relevant to the discussion at hand. It's like if people were having a debate about where to find the best tacos in town and you stepped in to say "tacos aren't as good as hamburgers, you know" and then got upset that nobody wanted to debate that point with you. It's not everybody else's fault if you don't understand how conversations work!
I don’t know why you are letting that one reply define the bounds of this conversation. My comment was directly relevant to the first comment in this thread and the comment I was replying to.
Despite your protestations to the contrary, these descriptions seem perfectly fine in that they're accurate and meaningful. And it if you want to start getting fast and loose with all kinds of new extra criteria and requirements for what it's supposed to do, they all seem squarely within the reach of the capabilities on offer, with some prompt tweaks.