So we’re pretty jazzed that marimo got GPU integration working, as of today (info here: https://t.co/1WXkUJPaMR). You hand someone a URL — they're driving a 3-D finite difference time domain simulator from a browser.
Making physics legible to AI systems today starts with digital twins – simulations of the system that are referenced back to empirical data, e.g., experiments. Our core tech today dramatically accelerates the engineer-in-the-loop build time for a digital twin. (1/4)
But simulation tools are often a bit old (GPU optional) They also are typically pretty hard to get going. With AI, we’re able to mitigate some of these disadvantages quickly, which is great. But it can be hard for the user to realize that without substantial back-and-forth. (3/4)
Thrilled to see @NIST’s AI consortium is opening up to new members again! Mandate also expanded to enable the foundations of integrating AI with our economy: “proven, scalable, and interoperable techniques and metrics to promote the development and use of AI.” Link below
Great fun today covering how verification unlocks agentic AI for science and engineering at @RIKEN_JP today. Many thanks to the RQC team and Prof Nakamura for hosting!
Today @Axiomatic_AI’s Winston Yin showcased our upcoming tool that makes LEAN’s mathlib, the core source of truth for lean proofs today, accessible to agents — semantic theorem searches, similarity clustering, and extraction to support tactics and proving capabilities.
Members of our LEAN team are showcasing how a minimal agent can make good proofs, and discuss the sorryDB, a living benchmark of unsolved problems at ICLR's verifai workshop today. Come say hi to Borja, Austin, Krystian and @LeopoldoSarra at the poster session!
Gravity is probably quantized into gravitons.
If not, however, there are experimental consequences. In particular, some level of irreversibility/noise. We finally classified ~all such models and calculated the noise.
https://t.co/qN1qQ0AELS