2 ms·
Check out prompt 2, "Write a limerick about a dog". The models undeniably get better at writing limericks, but I think the answers are progressively less inter
by furyofantares 1y ago
Check out prompt 2, "Write a limerick about a dog".
The models undeniably get better at writing limericks, but I think the answers are progressively less interesting. GPT-1 and GPT-2 are the most interesting to read, despite not following the prompt (not being limericks.)
They get boring as soon as it can write limericks, with GPT-4 being more boring than text-davinci-001 and GPT-5 being more boring still.
- amelius 1y agoI don't know if that is bad. The most intelligent person on a party is usually also the most boring one.
- saurik 1y agoI mean, to be fair, you didn't ask it to be interesting ;P. There once was a dog from Antares, Whose bark sparked debates and long queries. Though Hacker News rated, Furyofantares stated: "It's barely intriguing—just barely." > Write a limerick about a dog that furyofantares--a user on Hacker News, pronounced "fury of anteres", referring to the star--would find "interesting" (they are quite difficult to please).