Can anyone steelman this "if it works, it's approximating a bayesian thing" idea from @ ESYudkowsky ? Like what, precisely, mathematically is he gesturing at? As far as I’m aware this is not true in general. Even the complete class theorem / PAC Bayes etc needs some regularity /

today we have an effortpost about AI and automated theorem proving (and some cool projects!)

When you store your knowledge and skills as parametric curves (as all deep learning models do), the only way you can generalize is via interpolation on the curve. The problem is that interpolated points *correlate* with the truth but have no *causal* link to the truth. Hence

Oh no I think the conformal prediction guy might be right

Out on ALEA! DOI: doi.org/10.30757/ALEA.…

Today Thinking Machines Lab is launching our research blog, Connectionism. Our first blog post is “Defeating Nondeterminism in LLM Inference” We believe that science is better when shared. Connectionism will cover topics as varied as our research is: from kernel numerics to

Today we're announcing Gauss, our first autoformalization agent that just completed Terry Tao & Alex Kontorovich's Strong Prime Number Theorem project in 3 weeks—an effort that took human experts 18+ months of partial progress.

What does it take to run Lean at scale? We're pulling back the curtain on our custom-built, semantically stateless REPL service, which lets our RL system scale to hundreds of thousands of CPUs independently of our GPU capacity. Dive into the details here: harmonic.fun/news#blog-post…

GAUSS: General Assessment of Underlying Structured Skills in Mathematics We’re excited to launch GAUSS, a next-generation math AI benchmark built to overcome the limitations of low skill resolution in today’s benchmarks. What it does GAUSS profiles LLMs across 12 cognitive