norvid_studies

@norvid-studies.bsky.social

charts and graphs follow me on twitter https://twitter.com/norvid_studies https://substack.com/@norvidstudies/posts

[concept missing a name] is that once world accommodates to X you can't just go back to not-X by personally deleting X, as a common example you can't "just not use a dating app" as you could in 1990 because all the people you would meet, are themselves coordinating via a dating app.

John David Pressman@jdp.extropian.net · 10h ago

Same. I didn't grow up in an abusive household or anything but the Internet was actually great and I'd be a lot dumber without it.

I remember hearing about the spiritual bliss attractor stuff when it came out but I didn't know it happened in their automated alignment-related behavior evals either. pretty funny...

Bild

tldr: search for crow-themed content → stumble on the perfect clip → unearth history of said clip while hunting for a clean version of the meagre 5s or whatever it actually comprises

E.g., our reasoning work here: academic.oup.com/pnasnexus/ar... shows how (even much older) models and humans make similar patterns of trading off between logic and semantic biases that can roughly be interpreted as "using prior knowledge. 3/

Language models, like humans, show content effects on reasoning tasks

Abstract. Abstract reasoning is a key ability for an intelligent system. Large language models (LMs) achieve above-chance performance on abstract reasoning

academic.oup.com