This statement combines a lot of what I think is good and less good about the thread of LLM criticism it’s from. I don’t think the stochastic parrot interpretation is more than one interesting model, I believe the relevant problems with longermism are almost completely disjoint from the ones mentioned, and I am unconvinced that marking media by consciousness of author is either philosophically clear or practically useful. However, I agree firmly with the tl;dr. https://www.dair-institute.org/blog/letter-statement-March2023

I continue to feel pretty alone in the interpretation that these models are neither trivial nor conscious; that they are neither nukes nor penicillin; that practically all the urgent questions about them are really the same urgent questions we’ve had about political economy, offshoring, and corporate governance for decades; and so on. I finally get to be a filthy centrist on an important issue and it turns out it’s boring.