I wonder how much different that is from humans, even professional comedians. You don't usually get to see all the jokes that comedians try that don't work, unless you're an avid attendee at comedy clubs.
For comedians, perhaps. But there are plenty of humans who aren’t very good at comedy.
In these LLM comparisons, people tend to compare them to the best humans, not average or below average. It’s a kind of motivated reasoning, with the goal of deprecating the abilities of LLMs.
Well, you did ask for a comparison to professionals.
But I think humor is pretty easy to rate for a human. Did it make you laugh? Yes? Then it might be a good one. This simple human filter radically improves your efficiency as a joke validator, leaving any LLM far behind.
For pictures I'm not sure if humans are much better. For music, yes, after a few listens (recognizing a hit usually takes time). For games, definitely yes.
[–] tikotus 2 days ago
[–] antonvs 1 day ago
In these LLM comparisons, people tend to compare them to the best humans, not average or below average. It’s a kind of motivated reasoning, with the goal of deprecating the abilities of LLMs.
[–] tikotus 1 day ago
But I think humor is pretty easy to rate for a human. Did it make you laugh? Yes? Then it might be a good one. This simple human filter radically improves your efficiency as a joke validator, leaving any LLM far behind.
For pictures I'm not sure if humans are much better. For music, yes, after a few listens (recognizing a hit usually takes time). For games, definitely yes.