Helmholtz AI Consultants Team Lead for Matter Research @HZDR_Dresden. Opinions are my own.
Dresden, Germany
Joined January 2018
- Tweets2.6K
- Following895
- Followers910
- Likes11.1K
Totally needed work👇! We are investigating predictive uncertainties with physics based prompts to LLMs. Hope to discuss this with others at #neurips2024
New Anthropic research: Adding Error Bars to Evals.
AI model evaluations don’t usually include statistics or uncertainty. We think they should.
Read the blog post here: anthropic.com/research/stati…
Awesome event. Great support it!
From 7-9 NOV in Berlin, join a
global network of leading experts in science, politics, business and media. Engage with groundbreaking research and hear from the 2024 Science Breakthroughs!
🌐 falling-walls.com/science-su…
#FallingWalls #ScienceSummit24
Couldn’t agree more. I hope we @helmholtz_ai can do our share to disseminate SBI to other lands of Physika!
A milestone! 10 years ago I had an epiphany about using machine learning to approximate likelihood ratios, enabling statistical inference when your model is a complex simulator. Since then ~1000 papers on SBI have been published, but this is the 1st from the @ATLASexperiment 🧵
This must be a fun event! Join, if you can!
We already have many fantastic registrations for the Snakemake hackathon 2025 at @CERN! But there are still some free seats left. Make sure to apply/register soon if you want to shape the future of Snakemake! indico.cern.ch/event/snakema…
Quite surprising! Big congrats to the awardees and quite a win for ML itself.
Learn more about the 2024 #NobelPrize in Physics
Press release: bit.ly/3TFxHIH
Popular information: bit.ly/3Bi9H8u
Advanced information: bit.ly/3N3pp9U
Peter Steinbach retweeted
Last call to submit to our #neurips compression workshop!
Less than 24 hours left to submit to the Machine Learning and Compression Workshop 🗜️🗜️ at #NeurIPS2024!
👉Workshop website: neuralcompression.github.io/…
👉Submission portal:
openreview.net/group?id=Neur…
@StephanMandt @karen_ullrich @emidup @ziv_ravid
I can only recommend the @nightsciencepod by @ItaiYanai and @MartinJLercher. I learned a lot so far.
Would be great to have more speakers from outside life science like @seanmcarroll or @MelMitchell1 and learn about their ways of doing science.
To be even more fair, I’d argue most researchers have a hard time defining what research is (and how it’s done) in the 1st place. Most is done intuitively, people hardly establish methods aside rules of logic.
But I agree, a lot of ML lit has flaws. #noprematuregeneralisation
Peter Steinbach retweeted
Very pleased to see this paper now published! Important work, and aligned with my observations. Communicating these concerns often leads to pushback—I know I have been hit by it hard—so kudos to the authors and the journal for publishing it 👏
Our new paper in @NatMachIntell tells a story about how, and why, ML methods for solving PDEs do not work as well as advertised.
We find that two reproducibility issues are widespread. As a result, we conclude that ML-for-PDE solving has reached overly optimistic conclusions.
Peter Steinbach retweeted
Right now, this is the most exciting race of all...
We are all benefitting from this competition...
Top of my reading list now! @KICampus @HIDAdigital
This Post is from an account that no longer exists. Learn more
Taking off in 25 minutes!
Join us for💾 "How to make data ready for AI" 🤖 at @SaxFDM's digital kitchen in one weeks time.
👉 saxfdm.de/veranstaltungen/da… 👈
Registration is required for the the video conference coordinates. Hope to see you there! @helmholtz_ai @HZDR_Dresden
Love the paper and the comment!
Another paper pointing out in details what we've known for a while: LLMs (used via prompting) cannot make sense of situations that substantially differ from the situations found in their training data. Which is to say, LLMs do not possess general intelligence to any meaningful degree.
What LLMs can be good for, is to serve as knowledge/routine stores for an actual AGI. They're a memory -- a representation of a data corpus -- and memory is a necessary component of intelligence. But keep in mind that intelligence is not just memory.
Join us for💾 "How to make data ready for AI" 🤖 at @SaxFDM's digital kitchen in one weeks time.
👉 saxfdm.de/veranstaltungen/da… 👈
Registration is required for the the video conference coordinates. Hope to see you there! @helmholtz_ai @HZDR_Dresden
I am returning again and again to compression these days (in python & julia). One important library in this remains Blosc2 and the growing ecosystem around it! I am happy to provide a testimonial for the team and the ecosystem:
ironarray.io/why-ironarray
@FrancescAlted @ironArray
For anyone doing #ComputerVision and #MachineLearning in the Dresden 🇩🇪 area, we are having a ComputerVision-Stammtisch tonight at 6pm at the #katanalabs office. I'll be talking about our approach to estimate #StarDist segmentation uncertainties. katana-labs.com/about/
My slides are available at doi.org/10.6084/m9.figshare.…
but I recommend you to dive into the paper, if you like.
So glad to see this out. Big CONGRATS to the team!
Our 🧬 language model GROVER has now been published in @NatMachIntell. GROVER is a foundation model on the human genome to understand how such models learn grammatical structure in 🧬 and fine-tune it for more specific genomics questions. doi.org/10.1038/s42256-024-0…
a 🧵 1/11
