David Pfau

I’m beyond thrilled to share that our work on using deep learning to compute excited states of molecules is out today in @ScienceMagazine! This is the first time that deep learning has accurately solved some of the hardest problems in quantum physics. science.org/doi/abs/10.112…

I have been going down a rabbit hole on the history of pranks and mischief in San Francisco. Blog post below:

It's time to bring 3D meshes into modern machine learning properly! 🛸 Our work solves the non-differentiability of the Exp map on meshes, enabling gradients to flow directly through geodesics. It’s differentiable, GPU-fast, and fully parallelised. circle-group.github.io/research/DSG

The entire NSF research budget is ~$9B/year. This is literally funding every awarded PI at every field and every institution. But we've decided that all of basic science is a rounding error in comparison to venture bets. Please consider funding basic science more.

@5amoljen You and Igor Babuschkin.

Crazy to think I got into both DeepMind and CERN…

@AndyMasley Thank you, I feel like I'm taking crazy pills.

First time I’ve disagreed with a community note! A little confused by the reactions to this. 1) There is a real hot mic moment that imo is nothing, but it is real 2) Everyone’s replying to Julia implying she thinks the AI Dario was on the hot mic, but that’s obviously not what she’s saying 3) I agree that the article is terrible overall

The USA is 250 years old this July and instead of launching a Lincoln/FDR scale rejuvenation of the country for another 250 years of glory, has instead chosen to vindicate every theorist of cyclical history by self-detonating the empire right on the semiquincentennial mark.

Sam Roweis was my undergrad advisor, and is very regrettably deceased. He had a lovely PDF for Gaussian identities that I've used for years: people.orie.cornell.edu/andrew/orie674… I've turned it into a markdown file for posterity: github.com/jonbarron/gaus… RIP Sam, thanks for everything.

Like many AI researchers, I cut my teeth as a fundamental physicist. If the UK wants to power growth via AI, the government should not be cutting fundamental science funding. The current plan is an absolute own goal. bbc.co.uk/news/articles/…

@Brad08414464 Doing good research requires 100% of your focus and is therefore next to impossible to do well without institutional support or being independently wealthy. Can't work 9-5 and just do science in the evening.

why isn’t there an underground research community for people to publish research outside of the system and outside of conventional channels? the world doesn’t have very good infrastructure to enable this

Is it possible to add a community note to a community note?

@EnablerGPT I'm taking crazy pills. I must be.

@pfau Approved community notes, now. Yeesh. It’s a stupid, boring hot mic moment and it was weird to hype it, but it definitely was a hot mic moment!

You would think that all of the people getting upset over this tweet would be the exact kind of people to recognize that this is a scissor statement.

@MasonKamb @victorvalbert It's better than bad, it's good.

@pfau @victorvalbert Log Intelligence was a rejected talk in the session on biological computing

Exciting, scary, and thoughtful session on the future of AI at #apsmarch and possible breakdown of peer review. Matthew D. Schwartz: "AI will keep getting better and better at physics. We will not."

I am now getting shown proposed community notes suggesting that the article does not contain a hot mic moment when it clearly does. Have you all lost your minds?

So all the people getting angry over the second twist (which isn't even a hot mic moment) are just revealing a basic lack of reading comprehension. It's really an incredibly potent discourse generator.

@MattBurtell @Thomas_Woodside The hot mic moment came before the fake Dario interview, so if you made it that far you already saw it.

@Thomas_Woodside Forgive me for not finishing the piece

The "truly bonkers hot mic moment" is that the interview is fake, and actually, you were reading slop. Journalists aren't exactly known for respecting their readership but cmon lol

Feels like I couldn't have picked a better time to tweet this than the day before Charles Bennett wins the Turing Award.

@Sam___Bone Twitter is the great equalizer. You're welcome.

@pfau I just read your website and it's crazy that a google deep mind employee taught me something today. Thanks!

The degree to which AI research at the big labs has almost entirely been reduced to hill climbing is actually an aberration and not reflective of the rest of science at all. Ironically this means AI research is probably the easiest branch of research to automate.