[fiftycents]

Well, I wrote a novel last year and now I'm fighting a prequel/sequel manuscript. So I asked GPT to help me with the plot. I explained the whole plot to him first and then what I have now... He gave me a lot of "common places", clichés, uninteresting plots. A week ago I asked him how to finish a melody I was writing ---yes, I'm master of none--- but that time it was just nonsense answers. Weird because then I asked him about a math problem. He explained it very well, and when I asked for a formula...I got the right one.

Edit: trying to fix bad English

[jraph]

>"lugares comunes", cliches, uninteresting story-lines

I guess that's an interesting consequence of such model languages working by completing with the most probable words. lugares comunes are... common and supposedly what is seen most in the training data.

[mejutoco]

Did you try asking GPT for unexpected arcs or plots? maybe that will fix the common scenarios.

[mcv]

I've also seen it produce complete gibberish when faced with a math problem. With a very clear but wrong explanation of why is gibberish was the correct answer.