Enrico Schulz

@Geha_Paul Thanks for sharing. Unfortunately, the link doesn't work...

This paper from Harvard and MIT quietly answers the most important AI question nobody benchmarks properly: Can LLMs actually discover science, or are they just good at talking about it? The paper is called “Evaluating Large Language Models in Scientific Discovery”, and instead of asking models trivia questions, it tests something much harder: Can models form hypotheses, design experiments, interpret results, and update beliefs like real scientists? Here’s what the authors did differently 👇 • They evaluate LLMs across the full discovery loop hypothesis → experiment → observation → revision • Tasks span biology, chemistry, and physics, not toy puzzles • Models must work with incomplete data, noisy results, and false leads • Success is measured by scientific progress, not fluency or confidence What they found is sobering. LLMs are decent at suggesting hypotheses, but brittle at everything that follows. ✓ They overfit to surface patterns ✓ They struggle to abandon bad hypotheses even when evidence contradicts them ✓ They confuse correlation for causation ✓ They hallucinate explanations when experiments fail ✓ They optimize for plausibility, not truth Most striking result: `High benchmark scores do not correlate with scientific discovery ability.` Some top models that dominate standard reasoning tests completely fail when forced to run iterative experiments and update theories. Why this matters: Real science is not one-shot reasoning. It’s feedback, failure, revision, and restraint. LLMs today: • Talk like scientists • Write like scientists • But don’t think like scientists yet The paper’s core takeaway: Scientific intelligence is not language intelligence. It requires memory, hypothesis tracking, causal reasoning, and the ability to say “I was wrong.” Until models can reliably do that, claims about “AI scientists” are mostly premature. This paper doesn’t hype AI. It defines the gap we still need to close. And that’s exactly why it’s important.

Individualized #iPSC-based readouts explain variability of clinical phenotypes in #schizophrenia. Our study aims to support mechniscially-informed personlized medicine. ja.ma/4s0Ljg7 @mpi_psychiatry @MCSP_LMU

New paper w/ @_LeeJaeJoong_ in Nat Neurosci!🧠 We developed personalized fMRI predictive models tracking ongoing spontaneous pain in chronic pain patients, trained on 6+ mo. of densely sampled data. A true team effort. Deeply grateful to our participants! doi.org/10.1038/s41593…

Early decline of intra-hippocampal connectivity tracks emerging clinical trajectories in the psychosis risk state. Our longitudinal fMRI study in 434 individuals is now available as a preprint 👇 medrxiv.org/content/10.648… Thanks to @AndrewZalesky @yetianmed @SidChop and others!

The first time you start a research lab, there’s no manual. So I made one. A practical toolkit for postdocs and early-career faculty launching their first lab: hiring, startup planning, collaborations, and building a sustainable research program. zenodo.org/records/188835…

@cl9681 I won't make it to Vancouver but I'd be happy to discuss individual level data analysis. I started doing it a few years ago and I have the impression we are only a few...

I’ll be speaking at CNS 2026 in Symposium Session 6: “Not Your Average Brain: Individual-Level fMRI as a Paradigm Shift for Cognitive Neuroscience” on Monday, March 9, 10am–12pm. Excited for a session centered on how individual-level fMRI can reshape cognitive and clinical neuroscience. cogneurosociety.org/symposia/?sym=… This symposium has an all-star group of presenters covering cutting-edge work in precision and individual-level neuroimaging. It should be a really fun and substantive discussion. @RodBraga @ArielleKeller @jingnandu049 Andre Zamani

Have you ever been asked from a journal to remove all your preprint citations from any type of article (review, mini-review, normal article) and remove all the content referring to them? I found that so nonsense 😐 #asking4friend

Newsletter 50 is LIVE! 🧠✨ Inside this milestone issue: ✅ 13 new research projects on the Neuroscience of Pain ✅ Launch of the EP BrainHealth partnership ✅ Details on 2 new 2026 funding calls Read the full update here: 🔗 neuron-eranet.eu/wp-content/upl… #Neuroscience #BrainHealth

@EraNeuron Thank you very much for funding our project. We look forward to conducting the study and sharing exciting results.

@EraNeuron @LMU_Muenchen @LMU_Radiology

Please share: Post-doctoral position in the Human Nature Lab in the area of human social chemosignaling. What role do the odors we emit and our capacity for smell play in shaping social behavior? academicjobsonline.org/ajo/jobs/30989

I tried it weeks ago, and I think it's a game-changer! I hope this really shakes up the editorial system

It’s crazy how much Bezos and @washingtonpost are eager to just kiss Trumps A.

What can *genetic insensitivity to opioids* teach us about endogenous opioid function in humans? Fully funded position in Oslo (PhD student or postdoc) Please RT Interested in pain, mu/kappa opioids, behavioural genetics, RCTs, or related? Apply here: 2411.webcruiter.no/Main2/Recruit/…

@OdedRechavi Interesting. No need to blow up anything for a big story. Is it peer reviewed?

We published a small "micropublication" - it's a cool format for publishing isolated results! (not every study has to be a "full story"): "Reduced insulin/IGF-1 signaling and loss of pals-22 promote repetitive DNA silencing in C. elegans" micropublication.org/journals/biolo…

@derspiegel Was für ein Unsinn. Der "Neurowissenschaftler" hat seit Jahren nicht mehr geforscht und hat mit seiner Forschung auch keinen tiefen Abdruck hinterlassen. Bücher über seine subjektive Meinung zu schreiben ersetzt keine Wissenschaft.

Der Neurowissenschaftler Joseph Jebelli erforscht das Ruhenetzwerk im Gehirn – und erklärt, wie wir auf brillante Ideen kommen. Ein Gespräch über Leere, Tagträume und warum die meisten von uns es ihrem Hirn grundlos schwer machen.#ref=rss" target="_blank" rel="nofollow noopener">spiegel.de/karriere/hirnf…