James Kirkpatrick

Such a great honor, thanks a lot @NeurIPSConf and congrats to my esteemed co-authors @ilyasut & @quocleix! The 2014 talk also stood the test of time IMO. Here is a slide from it (powerful models of today == large transformers). Believe it or not this talk was controversial at the time (I got a lot of negative feedback at NeurIPS).

Happy to share my TED talk recorded live at TEDAI in San Francisco last month, giving an insight into a new paradigm I see emerging: creating AI analogues of our real messy world, to accelerate new scientific breakthroughs. Watch it here: bit.ly/3Z6YyiU Some of the topics explored: - The sheer scale of new knowledge about our protein universe that we now have access due to AlphaFold, with models replacing the need for real world experimental work - The potential to bring together the exploding world of AI to the increasingly costly world of drug design - Why I believe that new breakthroughs in machine learning have allowed us to produce an AI Analogue of biomolecular systems in the human body - How could we can use AI like AlphaFold 3 to modulate the function of proteins involved in cancer - How we are moving to a world where AI agents can begin to design new drugs purely in-silico, on a computer

At Isomorphic labs we are looking for ML researchers to build on top of the next generation of AlphaFold models, developing the tools that will allow us to crack systems biology and develop better drugs cheaper and faster than ever before. lnkd.in/eP7Rytgb

We have over 30 countries present at this years Indaba but clearly Nigeria is the most represented (speculating). Also present is #BurkinaFaso . Tag us in with your selfie and country you are representing. __ @ankahira

Andrew Malkinson case shows importance of securely stored scientific evidence | Forensic science | The Guardian theguardian.com/science/2023/s…

@ShuiwangJi @DaniloJRezende Wow! What a comprehensive survey

Excited to share our latest survey paper: "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems"! 🚀 ArXiv: arxiv.org/abs/2307.08423.

Pretty wild that simple text compression algorithms demonstrate few-shot learning. aclanthology.org/2023.findings-…

How can we improve the crystallisation and homogeneity of wide-bandgap perovskite materials? Our study provides a mechanism which explains the role of methylammonium chloride, ultimately improving 1.8 eV perovskite solar cells for all-perovskite tandems doi.org/10.1002/adma.2…

@RuehleVictor @ChrJunghans hanging out with the original @votca_software crew!

wish they taught us this in high school instead of useless crap like taxes or nutrition. never gonna use that in real life.

Never underestimate a droid: robots gather ⁦@AIforGood⁩ summit in Geneva. theguardian.com/technology/202…

Industrial Revolution iron method ‘was taken from Jamaica by Briton’ - eye-opening historical investigation by ⁦@jennybulstrode⁩ theguardian.com/science/2023/j…

Midjourney July 2022: July 2023:

Spent 6h driving to events yesterday, utterly stressful & hours of dead time (Pile of work left to do when I got in) Today - Bournemouth by train & on to Oxford. Could work, dose & get everything done whilst traveling. No parking stress or cost. Driving is not freedom. Trains are

Yes, a 𝐬𝐞𝐜𝐨𝐧𝐝 team has now simulated the recent IBM experiment within just two weeks since it came out! Why is there no media hype about these amazing classical breakthroughs that can beat even state-of-the-art quantum computers? arxiv.org/abs/2306.16372

@_iainmcculloch looks like a puntable route :D

Oxford roads are a mess. It can take almost three quarters of an hour to travel less than two miles. The disruption and pollution as a result is scandalous. There must be a better way.

Congratulations to NANOGrav for detecting evidence that gravitational waves fill the cosmos. These ripples in the fabric of space happen when massive objects like black holes circle one another before colliding. The result brings us closer to understanding how galaxies evolve. go.nasa.gov/3PwynPD

🧵(1/6) We propose EquiformerV2 with state-of-the-art results on all OC20 tasks and AdsorbML. Joint work with @bwood_m , @abhshkdz from @OpenCatalyst and @tesssmidt . Besides, thank @_akhaliq for tweeting before. Paper: arxiv.org/abs/2306.12059 Code: github.com/atomicarchitec…

Amazing work by @aicooper and his team published on @digital_rsc ! Thank you for sending our work to us. We also collaborate with Andy in a joint project so stay tuned for our joint @acceleration_c work.