Jonas Köhler
@jonkhler
Followers
2K
Following
10K
Media
111
Statuses
3K
Scaling Machine Learning + Molecular Simulation @cusp_ai ex-{@MSFTResearch,@GoogleDeepMind,@MPI_IS} born @ 353 ppm he/him block toxicity
Berlin
Joined April 2019
Soon @ @icmlconf !. 🌪️Normalizing flows for rigid bodies and the rotation group SO(3)! 🌪️. We found a way to design smooth normalizing flows for rigid body systems, like ice 👇. Joint work with amazing @inve_michele @pimdehaan and @FrankNoeBerlin . 1/.
5
34
204
I'm getting bored by this "AI" research on LLMs recently. Great tools for niche tasks / helpful here and there - no questions asked! I am getting more and more interested in what happens to the actual brains interacting with it and what it does to people and their context.
0
0
0
For context for those who don't follow this thread: regular LLM use with user adaptive memory behaviour can create your own echo chamber in which you become a madman. And that sh*t is used unregulated by young folks who still find their place in the world? Complete nonsense.
1
0
0
We(=irresponsible AI companies, but also society not acting) are breeding a collective psychosis mankind has never seen. There are strong (at least legal) guardrails around drugs, porn, violent media in place. But here it's just wild west. Insane.
1
0
0
Great to see this out!.
Our development of machine-learned transferable coarse-grained models in now on Nat Chem! I am so proud of my group for this work! Particularly first authors Nick Charron, Klara Bonneau, @sayeg84, Andrea Guljas.
0
0
2
RT @jo_brandstetter: General relativity 🤝 neural fields. This simulation of a black hole is coming from our neural networks 🚀 . We introduc….
0
70
0
RT @rcbregman: Huge new study finds: people think AI will worsen almost everything they care about. By a 3-to-1 margin, they want more reg….
0
802
0
RT @martinmbauer: In Germany, international students can study physics tuition-free because public universities are funded by taxpayers. Si….
0
216
0
RT @martinmbauer: Instead of discovering new physics, AI is struggling to rediscover old physics.
0
266
0
Cool work.
Can an AI model predict perfectly and still have a terrible world model?. What would that even mean?. Our new ICML paper formalizes these questions. One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
0
1
8
RT @docmilanfar: If you expect people to pay attention and be productive, every meeting room, conference room, classroom, office, and home….
0
14
0
RT @KyleCranmer: 🎉 Great news: Our Machine Learning and Physical Sciences workshop at @NeurIPSConf will be back again this year! 🎉.Keep an….
0
33
0
Not everything is run-once batch compute. Same meme applied to real-time audio/control? Individual MD/MCMC steps? Your N-wise evaluation in some O(f(N)) >> O(N) algorithm 🫥.
0
0
4
