This is a great retrospective on the Jarzynski equality. But I've gotta dispute the final sentence: "Rarely has a breakthrough discovery been so clearly evident at the time of its publication". It took almost a year to get the paper past the referees! https://t.co/qFbAw1yNuA

Neural Thermodynamics: Entropic Forces in Deep and Universal Representation Learning

The San Francisco maximalism posting will continue until morale improves.

“The future is already here — it's just not very evenly distributed.” — William Gibson

arXiv moving towards being a postprint rather than a preprint server. Understandable, but not good.

[*] But @GillVerd, for the love of all that is beautiful in the world, can you make the hole in the foam the same shape as the device?

A seven sided 3d printed enclosure would not have been my choice, but we're all talking about it aren't we? As a colleague remarked "It is provocative. And I am provoked." Mission accomplished. [*]

Time article discussing our work We recently had the opportunity to talk to TIME magazine about the burgeoning landscape of virtual cell research across the frontier AI and biology teams, which I see as a position piece about the future potential of this vision. Also fun to see

Indeed! Research funding in this country has been far, far too conservative in this country for the last few decades. There's been almost zero support for anything outside of a few fashionable fields. But now the massive demand for AI compute is spilling over and leading to a

New artificial neurons replicate real brain chemistry for smarter AI hardware design | Aamir Khollam, Interesting Engineering The design fits each neuron into a single transistor, a leap that could make future AI hardware smaller, faster, and more efficient. Researchers at the

Vaire is making solid progress in developing adiabatic computing. (Less flash but more substance than some other startups.) Very promising approach to energy efficient computing, but not without its challenges of course.

There is absolutely no fundamental reason we build AI the way we do today. There certainly is a radically different approach that is orders of magnitude more energy efficient. I’m going to find it before I die https://t.co/ML4fBRTsvv

Extropic wrote a paper. And at first glance it looks meaty. 👀

this is one of the legit nascent drug discovery companies...I don't think this is sufficient to revolutionize drug discovery but it is necessary to do so

I wish more people understood this. The dominant computer architecture (a large memory pool far away from compute units) is great for general-purpose computing but highly inefficient for learning. GPUs are a step in the right direction but don't go far enough. Chips with

The future of intelligence is non-von Neumann. Compute where the data lives.

Yeah. I assume: 1. local memory ~ cheap; 2. local compute ~ somewhat expensive; 3. transferring data ~ really expensive, with cost proportional to bits*millimetres. Hence graph computing architectures with co-located storage and compute, and message passing for synchronisation.

Thrilled to share our latest preprint with @D_Sekizawa and @ito_sosuke! https://t.co/wHJGzVU52l We merge Koopman operator analysis with stochastic thermodynamics to uncover the oscillatory origins of thermodynamic dissipation in nonlinear systems. This framework pinpoints

Achira is going to revolutionize computational drug discovery. Come be part of a brilliant team.

Multiple ML/AI roles available at https://t.co/jgDolM2K8Z focused on building atomistic foundation simulation models. If you’re in the market and interested in AI-infra\research engineering, or GenAI-research roles reinventing our understanding of matter and reasoning about

I tried this test on my son who knows little about medicine, but does have an encyclopedic knowledge of Pokemon. Response: "Come on man. You gotta try harder than that"