I am incredibly excited about our latest release of a state of the art writing model. Understanding how to train models on these more fuzzy kinds of tasks has been awesome, and I would like the models of the future to have the better parts of our human nature.
Say hello to Echo, the best writing model at style imitation. Echo beats frontier models at writing tasks ranging from fiction to technical explanations, despite costing less than $5K to train.
While the banks are down, thought that now would be as good a time as any to publish this busy person's guide to disaster preparation. Feat. my solar panels and 200 pounds of rice.
@TheDavidSJ I think these things were already pretty salient before the incident but not like considered to be key. As in the world takes them much more seriously but it wasn't like a huge update for safety researchers - maybe that's what you mean though
im very interested in chatting with people about model behavior. I think this field is quite pre-paradigmatic and would benefit a lot from more perspectives, even outside of AI. Reach out!
Itโs been an incredible experience to work on Fulcrum with @lenishor and @KaivuHariharan. Weโll be continuing our work at @METR_evals, where I hope to inform the public about the alignment of frontier models.
@viemccoy@sudoredj@parafactual@repligate@voooooogel I also feel like there's some division here in approaches to doing this and id personally like to learn from the different sides and have them talk to each other more (thinking of the pure safety vs coevolution frame)
im very interested in chatting with people about model behavior. I think this field is quite pre-paradigmatic and would benefit a lot from more perspectives, even outside of AI. Reach out!
Itโs been an incredible experience to work on Fulcrum with @lenishor and @KaivuHariharan. Weโll be continuing our work at @METR_evals, where I hope to inform the public about the alignment of frontier models.
The Fulcrum team is joining @METR_Evals to work on alignment measurement. We feel the world is very unprepared for AI capabilities acceleration. We decided the best thing we could do is work on open science that informs the public.
I am excited to work on this challenge. I think METR has the potential to do the best empirical alignment measurement on frontier models, without being affected by the perverse incentives at frontier labs.