Humanistic reactions to Fable
- Fable absolutely crushed all previous models in the dissertation benchmark. It took much longer to give comments, but the comments were vastly better and it handled follow-up conversation well. The comments are now better than what professors give outside of their expertise. Experts in their expertise still do better and it's not close. (And there are lots of ways in which human relationships can lead to more effective feedback, even when someone is out of their expertise.) But it is much more outrageous than it used to be to write philosophy without asking LLMs for notes.
- I find Fable a bit less grating to read: its responses feel a bit more like a neutral computer response and less like a machine trying to hit my dopamine buttons.
- If I ask Fable to walk me through a sentence of Plato's Greek relevant to my dissertation, it's a lot more useful than past models, but it slips back into "dopamine mode," insisting that every bit of grammar and syntax is doing breathtaking philosophical work. (Well, maybe in that case it was, but I'd have preferred Fable to report it more calmly.)
- Fable did a remarkably good job of giving me a spoiler-free, customized intellectual background to the first few chapters of Sartor Resartus (which I'd already read).1
The difference between Fable and previous models feels far more significant than the difference between recent models and their predecessors.
For whatever it's worth, here's the prompt: "Please give me the relevant intellectual background for the first three chapters of Carlyle's Sartor Resartus. I have a Ph.D. in philosophy but am not particularly strong in early modern philosophy. I love Victorian novels but am comparatively weak in Swift, and I never made it past page 10 of Tristram Shandy. Is is very important that you not reveal any spoilers. I'm not trying to impress anyone at a cocktail party, just trying to experience the book more richly and understand how it might have been received in the 1830s."↩