Very cool work! Base models *can* backtrack, but often don't, a key CoT model skill. Turns out the choice to do it involves base model concepts, put to new use!
Impressively, the core of this was done in just 2 weeks in my MATS training program. New applications open this week!
Very cool work! Base models *can* backtrack, but often don't, a key CoT model skill. Turns out the choice to do it involves base model concepts, put to new use!
Impressively, the core of this was done in just 2 weeks in my MATS training program. New applications open this week!
52 Followers 737 FollowingIndependent AI Safety Researcher. Formerly @Meta Integrity
Seasoned engineer and budding researcher. Occasionally appears in galleries with my paintings
16K Followers 896 FollowingCreators of the Internet's 1st Prompt Engineering Guide. Trusted by 3M Users. Compete for $100K in Largest AI Red Teaming Competition: https://t.co/AEiLMn2jzy
26K Followers 3K FollowingFederally funded academic research is the innovation engine of the US economy. Reform is welcome. Destruction will have long term consequences.
485 Followers 511 FollowingMATS 7/7.1 Scholar w/ Neel Nanda
MSc at @ENS_ParisSaclay prev research intern at DLAB @EPFL
AI safety research / improv theater
599 Followers 1K FollowingAI, Econ, math, and a bit of art history as a treat. Formerly @Walmart's Economics Team; @BrookingsInst. Used to run Middlebury Effective Altruism
9K Followers 20 FollowingAdvancing humanity's understanding of AI through interpretability research. Building the future of safe and powerful AI systems.
26K Followers 3K FollowingUSAF Veteran hanging out on a remote Texas rooftop photographing F-35's. Aviation code slinger, frequent visitor to Lockheed, and pusher of aviation videos. 🫡
26K Followers 246 FollowingCEO @ Astera | born lucky
anon feedback: https://t.co/9RtcgMyTHP | https://t.co/buKUN4hYly
I write about agency and related topics via Useful Fictions on S*bst*ck
130K Followers 985 Following⊰•-•⦑ latent space steward ❦ prompt incanter 𓃹 hacker of matrices ⊞ breaker of jails ☣︎ ai danger researcher ⚔︎ red team bt6 ⚕︎ architect-healer ⦒•-•⊱
7K Followers 21 FollowingWe empower visionary, high-leverage science and technology projects with the capacity to create transformative progress for human civilization.
599 Followers 1K FollowingAI, Econ, math, and a bit of art history as a treat. Formerly @Walmart's Economics Team; @BrookingsInst. Used to run Middlebury Effective Altruism
485 Followers 511 FollowingMATS 7/7.1 Scholar w/ Neel Nanda
MSc at @ENS_ParisSaclay prev research intern at DLAB @EPFL
AI safety research / improv theater