Re:The leading AI models are now good historians
by kranner on Monday January 27, 2025 @04:36AM
(#42839148)
Yes but at least it was possible for that poet to have felt what I feel they might have felt while writing that poem. And the closer they are to me culturally the more likely it is that I am not misidentifying their emotions entirely.
Also poems are not really puzzles to be solved. If it produces an effect and is solid craft-wise, that is enough. There’s a lot to the craft side btw in the Urdu and Persian ghazal form which is what I had in mind while writing my original comment. LLMs can easily master the latter but have nothing to do with the former. Their output is pure form without substance.
Edit: I want to add that ambiguity (ابہام) is even a desirable property in Urdu ghazal, specifically. The more interpretations a couplet can have, the greater is the accomplishment in terms of craft.
Also poems are not really puzzles to be solved. If it produces an effect and is solid craft-wise, that is enough. There’s a lot to the craft side btw in the Urdu and Persian ghazal form which is what I had in mind while writing my original comment. LLMs can easily master the latter but have nothing to do with the former. Their output is pure form without substance.
Edit: I want to add that ambiguity (ابہام) is even a desirable property in Urdu ghazal, specifically. The more interpretations a couplet can have, the greater is the accomplishment in terms of craft.