@Polymarket market proposal: theory of everything by end of 2027.
A theory that unifies general relativity and quantum field theory.
I think scaling laws and RL will get us there.
@Polymarket Resolution: peer-reviewed paper published in recognized scientific journal that
- provides unified theoretical framework for QFT & general relativity
- reproduces Standard Model of particle physics
- Makes >=1 novel, quantitative prediction not explained by Standard Model alone.
What if bigger models are actually more interpretable?
Mixture of Experts (MoE) have become central to scaling LLMs and are used in nearly every frontier model.
Yet we still lack a mechanistic understanding of how MoEs represent features differently than dense models.
How do experts affect superposition (how networks represent more features than there are neurons) and how do experts specialize?
In our recent NeuIPS workshop paper, we found that MoEs are more interpretable than dense models.
work done w/ @foundinrome@JeremiNuer at Arcadia Research Team
@KyleVedder@TPenguin4748 Because discrete tokenization is not a good scheme or because diffusion action head is v important? And if the latter, what makes it so--being able to have stochastic policy?
"how neural networks think at scale"
neural networks are a black box
mech interp allows you to peek inside the black box
in a new blog, we explain how models represent meaningful concepts
how are these concepts arranged in the network
and what are the building blocks for understanding the model internals
@thesamparr The Mysterious Case of Rudolf Diesel -- pretty unknown guy, yet he revolutionized fuel.
Biography + has some mystery / murder in it.
https://t.co/HQhkQTNhLq
Hope its not too late... we had to fly in from college to make it happen.
Might be the slowest My First Muscle entry. But one of the funnier ones. Consider us for nike swag 🙏
@ShaanVP@thesamparr
https://t.co/LTPsw29g4E
@finntbaker@finntbaker listen to the podcast from Freakanomics: Is google worse? Addresses this directly + is interesting
link: https://t.co/TBDiuU2Gow
@_mark_murray Hey Mark! Do you think bills like AB274 (pushing recycling content standard onto PET thermoforms) being expanded to to other #3-7 plastics could help establish demand for these plastics that have no market?
Would love to discuss
@PhilTing Hey Phil! Working on a project for plastic recycling, and plastics #3-7. Do you think a plastic content requirement bill for say, #5 PP plastic would increase demand and incentivize sorting facilities to start sorting them? Would love to discuss
Now that applications are close to being over, I've been looking into plastic recycling, and how AI can be leveraged to solve it... excited about this avenue and going to explore it for the next couple months.