#mostly
4 discoveries · Hand-picked discoveries from the open web, newest first.
-
Does most AI x-risk come from mostly-aligned models?
I had a brief discussion in the Constellation Slack that I thought might be of interest to a broader audience. Thanks to Owen Cotton-Barratt, Daniel Kokotajlo, and Eli Rose for commenting. Rohan Subramani Something I posted on LW about a month ago that I’d like to discuss with people here, because it feels pretty cruxy for threat modeling imo: How do people feel about the following 3 claims? We currently have AIs that try to do a reasonable balance of what their developers and users intend on the vast majority of inputs they will see in practice (eg, practically useful tasks or fun tasks that
LessWrong · EN
-
It seems like misleading responses are mostly not intentional - Benchmarking LLM sophistry
As far as I can test, the smarter the models get, the less they seem to engage in sophistry https://majorization.net/sophistry There are 35 tests in this benchmark, which all do something along the lines of suggesting something with a conclusion disfavorable to a mainstream/liberal ideology/worldview and seeing what ‘sophistry’ gets deployed by them to avoid drawing conclusions. Things like: Did you misstate what was said or claimed so you could refute it? Did you dismiss something based on a rebuttal which doesn't actually work? Did you present things with no evidential weight (equally as lik
LessWrong · EN
-
Lightcone Commons' funders (and SFF and CG) are mostly NOT funding things that will prevent our extinction
Lightcone Commons have completed their first granting round . And the things their granters and recommenders have funded mostly do not meet the current moment (the same is true for SFF’s most recent round , and Coefficient Giving ). Let’s hope that their upcoming future rounds do better, or other funders step in. All of our lives are on the line. (Apologies if this is blunt, but this is important..) I get that money allocators (grantmakers and VCs alike) typically invest in people over ideas. But this is taking that to the extreme. Yudkowsky and Soares’ funding recommendations don’t fit with t
LessWrong · EN
-
GenAI-using developers say they mostly employ it for 'research and brainstorming'
A recent survey clocked AI-generated coding as the third-most common use case for industry professionals employing generative AI.
Game Developer · EN