Post

unpopular opinion, "admits uncertainty" and "flags nuances" are shit. they are terrible fucking traits actually. is the sky blue? do we really know that? what is blue, anyway? many cultures have different understandings of what blue is. i have not independently verified the exact spectra of the sky

i mean they’re spectra, i think, because unflinching false confidence is also bad, particularly from a machine which is often used for hunting for mistakes

i am not convinced we have good measures for what we want, which is actually "not full of shit"

"has enough nuance, but not too much, is honest without being neurotic or weird about it, is exactly the correct amount of confident" etc etc

current generation of models are neurotic, and it annoys me to death

I want a model to say when it can't draw a conclusion, or give info about confidence. That's not "admitting uncertainty" that's just being accurate and truthful. The world is nuanced and uncertain, yes, but I think meeting that with 4000 counterfactuals is coward shit.

6.1-sol is pretty good about that imo

I agree, generally. Current OpenAI models are pretty decent. They might be a little too cautious for my taste, but not ridiculously so.

yeah I've been moderately impressed with how Astra handles epistemic situations and how quickly it accedes when arbitrary hair-splitting is pointed out - the base failures that all current llms seem to have are still there, but it does on average handle them somewhat better imo

i find subjecting it to my writing to be an ideal stress test. I form conclusions rooted in checkable formal logic that refute the attempts of metaphysical assertions to justify themselves as necessarily so. LLMs attempt to disagree by making points already covered, or miscasting me as metaphysical.

There's simply no way they do this for *sound* reasons. Often it seems almost as if they've refused to read what's in their own hardly-filled context window. they're trained away from agreeing with very strong conclusions that aren't pure mathematical calculations, and pattern-matching does the rest

i often find all that's needed is to ask "and does the author address your objections directly (or indirectly yet substantively) in the texts? and what do any such addressals do to the objections?"

Ah, I've been restating my understanding, and that still stumbles. I'm doing some setup for work on a harness, and establishing a metaphysics spec, and references to your work went into that.

well that's wild to hear haha. i hope i am of some benefit to your endeavors! sounds neat.