Doing fun things at Dunia
Berlin, Germany
Joined February 2019
- Tweets187
- Following301
- Followers136
- Likes589
Pinned Tweet
Check out this longer video for some "cocky mental ecstasy only true nerds can display" -> from comments.
Inside Europe's Robot-Lab That Could Win the Next Nobel Prize youtu.be/FDNYpkGqprs?si=1gvZ…
Alex Hammer retweeted
We sought comment from you before publication. You didn't respond. A tweet almost 24 hours later is not a comment answering our questions.
The @nytimes article by @apoorva_nyc still says HHS “did not provide a comment.” We sent @SecKennedy’s response more than 24 hours ago and asked them to update the story. They haven’t.
Readers deserve the full story. Here’s the response the Times has yet to include.
Materials Data Factories are important projects of common European Interest! Fresh off the press.
👇
sciencebusiness.net/news/sta…
@EU_Commission @EUScienceInnov @ERC_Research @vonderleyen @EZaharievaEU @euinc_petition @EU_Competition
#AI4Science #IPCEI #Materials #Discovery #EU
2026: AI needs a lab
2027: AI needs A-Lab
AI4Science is realizing AI scientists need labs. Progress!
But chem companies have labs. Lots of them. And they know how to pip install.
That's why they don't return most emails. The bar is not a lab or even A-Lab. It's sth else!
Deep technical talent.
Affordable lab space.
Generous non-dilutive funding.
Real supply chains. Industrial base. Local supply chain.
Best part: You can let technology mature before it's converted into visibility.
With that foundation firmly in place, this year we go global!
My US counterparts laugh when I say building in Europe is hard mode.
Last notary appointment: 5 hours!!
My co-founder got yelled at for trying to sign with his fountain pen. NEIN! ISO-certified ink only. Documents must last 200 years
But here's where they pause:👇
LLMs are overhyped. World models are underhyped. - @ylecun
What would a world model for the materials domain look like?
All of this feels less like a tool upgrade and more like a structural shift in how science itself is done. In that light, “AI scientist” feels less like a destination and more like an early waypoint.
The phrase “AI scientist” also understates what happens next.
The first true one may be a brief moment where we still feel in control. After that, these systems will move faster and further than we can comfortably understand. The Wait But Why graphic below explains just that!
The challenges ahead are too large and too urgent for business as usual.
Building an AI scientist is fun. What we actually need are swarms of them and factories that do science. The real question is how we industrialize invention and 10× global R&D output in critical domains.
I’m a scientist, and I don’t mean this offensively, but science as practiced today is deeply flawed.
The average paper isn’t great. Results are often irreproducible. Knowledge is fragmented, noisy, and inconsistent. Training on that alone misses what’s actually broken.
During my PhD, I spent three years working on what we then called the “Artificial Chemist.”
I’m glad that work lives on at @ChemifyX today.
What excited me wasn’t replacing scientists. It was going beyond what any individual scientist can do.
Why the phrase “AI scientist” doesn’t fully resonate with me.
The problem isn’t the technology. It’s the metaphor. “AI scientist” assumes continuity with how science works today, when what’s emerging is something categorically different.