Miles Cranmer
@MilesCranmer
Assistant Prof @Cambridge_Uni, works on AI for the physical sciences.
I'm super excited to share a new initiative I am a part of! Announcing: Polymathic AI 🎉 We are developing foundation models for scientific *data*, such that they can leverage shared concepts across disciplines. polymathic-ai.org 1/6

What are the odds that LLMs which conceal their chains of thought are doing all their "thinking" in a mix of languages? Then having a second LLM translate it back to the user

Really loving AR glasses for working while traveling. No squinting at my laptop + I can still enjoy the view! These just plug into my closed laptop as an external monitor. Means I'm actually productive on planes now because I'm not hurting my neck looking down the whole flight!

I am a HUGE fan of Mooncake.jl github.com/chalk-lab/Moon…, a new language-level autodiff for Julia. Most impressive is how uncompromising the devs are about correctness and reliability. It has some of the best software engineering practices I’ve seen in the entire ecosystem!

Has anything been written about Google's LLM turnaround? They seemed to be lagging for a while, but the recent Gemini models are super impressive. I would be interested to read an analysis of what changed in their approach or team structure.
Wow. Apparently some PySR users are building dedicated rigs for faster equation discovery!
SymbolicRegression.jl v1.10.0 is out! It can now evolve expressions over arbitrary input types. The video below shows it reverse-engineering a string transformation from examples. Curious to see how people use this!
🚨 Final Reminder: Encode Fellowship Applications Close Tomorrow! (Wed Apr 30)
What is the role of memory for modeling time dependent PDEs? I will be at ICLR presenting our paper (Oral) where we study when it is beneficial for modeling time-dependent PDEs! 🔗openreview.net/forum?id=o9kqa… [Oral]: Thu 24 Apr 10:30 am @ Session 1E [Poster]: Thu 24 Apr 3 pm #617
Just released the first version of BorrowChecker.jl! This is a Julia package that emulates some of Rust’s ownership, lifetime, and borrowing semantics. It’s been useful for enforcing safe patterns in multithreaded codebases, and already helped me catch an extremely subtle bug in…

> pick your favorite LLM > try building anything in Blender scripting mode > send it screenshots, profanities, prayers > bug A → bug B → bug A → bug B > rip your hair out > become an AGI skeptic
Cool reverse engineering of the actual GPT-4o output from @jie_liu1!
The following are the five images truly generated by GPT-4o