Prime Intellect releases INTELLECT-2 (open-sourced first 32B parameter model trained via globally distributed reinforcement learning), which demonstrates that globally decentralized RL works, per them.

They also open sourced:
- PRIME-RL
- SHARDCAST
- TOPLOC Validators

would Urs Schreiber quantum monadology / topological quantum compilation count? under some diagrammatic rewriting models, the diagram itself is the simulation of a kind...

or perhaps some open-ended world model learner that uses ZX Diagrams for its world?

arxiv.org/abs/2201.132...

** It's happening!! **

We managed to convince Google to open source PebbleOS. Took a while, but they just did it today! github.com/google/pebble

With that, we're bringing Pebble back! I blogged about it - ericmigi.com/blog/why-wer...

Overall, it's pretty hard to argue that any surprise function S(p) shouldn't satisfy the following properties:

- S(1) = 0
- S(0) = ∞
- S decreases monotonically
- S is continuous
- S(pq) = S(p) + S(q)

Entropy is one of those formulas that many of us learn, swallow whole, and even use regularly without really understanding.

(E.g., where does that “log” come from? Are there other possible formulas?)

Yet there's an intuitive & almost inevitable way to arrive at this expression.

Today's advent of papers explores a fascinating programming construct from Alex Warth and Alan Kay called Worlds.

Worlds let you control mutation by branching your program state. Later if you decide you want those changes, you can commit them back.

jimmyhmiller.github.io/advent-of-pa...

I wrote a summary of the main ingredients of the neat proof by Hugo Lavenant that diffusion models do not generally define optimal transport. github.com/mathematical...

Morley's theorem (1899) states that the intersections of the adjacent angle trisectors of any triangle defines an equilateral triangle. https://www.cut-the-knot.org/triangle/Morley/

Reconnect

I wanted to make my first post about a project close to my heart. Linear algebra is an underappreciated foundation for machine learning. Our new framework CoLA (Compositional Linear Algebra) exploits algebraic structure arising from modelling assumptions for significant computational savings! 1/4

This is SO AWESOME.
Embeddings, embeddings, Embeddings!!
Thanks so much for doing this.
[It's on my queue to embed MajorTOM with Clay v1.5 so we can facilitate comparisons]

s/awk/tuah/

guess what

Then I find superposition of transverse modes which reflects as little as possible. And finally make a wave package of the modes with different energies. Voila, behold the result!

Applied Categorical Duck Cyberneticist combines expertise in social cognition, behavioral ecology, and computational modeling, applying this knowledge in a practical way to improve our understanding of duck behavior and, by extension, our understanding of social systems more broadly.

🦆⊛👁️‍🗨️