Ted Underwood on Nostr: A couple days later, with the benefit of replies, I would say I was too quick to ...

A couple days later, with the benefit of replies, I would say I was too quick to assume that the simple challenge of sustaining coherence over long contexts has been overcome. There may also be more interesting challenges specific to fiction. But idk that they’re the bottleneck yet.

RE: https://bsky.app/profile/did:plc:565ebob5f6hw33hjdkxty6qj/post/3leiaxmtzks2k

quoting nevent1q…kft9

Something I don't understand is: why can't LLMs write novel-length fiction yet?

They've got the context length for it. And new models seem capable of the multi-hop reasoning required for plot. So why hasn't anyone demoed a model that can write long interesting stories?

I do have a theory ... +