Erin Grant

Join us! We are opening many postdoc positions both in London and in Gothenburg!

London
🔗https://www.lesswrong.com/posts/GTt33CasvWjxxazJw/hiring-principia-research-fellows
📅 deadline: March 26th

Gothenburg
🔗 www.chalmers.se/en/about-cha...
📅 deadline: April 1st

Excited to launch Principia, a nonprofit research organisation at the intersection of deep learning theory and AI safety.

Our goal is to develop theory for modern machine learning systems that can help us understand complex network behaviors, including those critical for AI safety and alignment.

1

#CCN2026 Proceedings submissions are open and due in *two* weeks! Info about how to submit in the thread below👇

Come share your science and hang out in NYC in August. :)

@dataonbrainmind.bsky.social starting now in Room 10 with opening remarks from @crji.bsky.social and the first invited talk from @dyamins.bsky.social!

Thrilled to start 2026 as faculty in Psych & CS
@ualberta.bsky.social + Amii.ca Fellow! 🥳 Recruiting students to develop theories of cognition in natural & artificial systems 🤖💭🧠. Find me at #NeurIPS2025 workshops (speaking coginterp.github.io/neurips2025 & organising @dataonbrainmind.bsky.social)

Two posts from Bluesky. The first one shows a figure from a paper published in Nature Scientific Reports full of totally incoherent AI fabricated gibberish words. The other a comment on a recently published paper by eLife discussing the paper and its peer reviews which were published along with the paper.

Nature Sci Rep publishes incoherent AI slop. eLife publishes a paper which the reviewers didn't agree with, making all the comments and responses public with thoughtful commentary. One of these journals got delisted by Web of Science for quality concerns from not doing peer review. Guess which one?

Probabilistic ML in scientific pipelines

I'm on the academic job market!

I design and analyze probabilistic machine-learning methods---motivated by real-world scientific constraints, and developed in collaboration with scientists in biology, chemistry, and physics.

A few highlights of my research areas are:

Applying to do a postdoc or PhD in theoretical ML or neuroscience this year? Consider joining my group (starting next Fall) at UT Austin!
POD Postdoc: oden.utexas.edu/programs-and... CSEM PhD: oden.utexas.edu/academics/pr...

Postdoctoral Fellow - Large Language Models as Models for Human Development This position is part of the Post Doctoral Fellows Association and has an initial appointment of two years. This position has a comprehensive benefits package. Location - This role is in-person at Nor...

I'm hiring (another) post doc, this time in collaboration with Natalie Brito @nataliebrito.bsky.social at Columbia! We will be exploring some of the characteristics of human development using deep learning models. Email with questions!
iaejup.fa.ocs.oraclecloud.com/hcmUI/Candid...

Hoping you find out and share! 🤗

Congrats Richard!!

I am hiring a post doc at UAlberta, affiliated with Amii! We study language processing in the brain using LLMs and neuroimaging. Looking for someone with experience with ideally both neuroimaging and LLMs, or a willingness to learn. Email me with Qs
apps.ualberta.ca/careers/post...

Frontiers | Summary statistics of learning link changing neural representations to behavior How can we make sense of large-scale recordings of neural activity across learning? Theories of neural network learning with their origins in statistical phy...

Since I'm back on BlueSky - with @frostedblakess.bsky.social and @cpehlevan.bsky.social we wrote a brief perspective on how ideas about summary statistics from the statistical physics of learning could potentially help inform neural data analysis... (1/2)

📢 10 days left to submit to the Data on the Brain & Mind Workshop at #NeurIPS2025!

📝 Call for:
• Findings (4 or 8 pages)
• Tutorials

If you’re submitting to ICLR or NeurIPS, consider submitting here too—and highlight how to use a cog neuro dataset in our tutorial track!
🔗 data-brain-mind.github.io

I’m recruiting committee members for the Technical Program Committee at #CCN2026.

Please apply if you want to help make submission, review & selection of contributed work (Extended Abstracts & Proceedings) more useful for everyone! 🌐

Helps to have: programming/communications/editorial experience.

arguably the most important component of AI for neuroscience:

data, and its usability

The rumors are true! #CCN2026 will be held at NYU. @toddgureckis.bsky.social and I will be executive-chairing. Get in touch if you want to be involved!

many thanks to my collaborators, @saxelab.bsky.social and especially Lukas :)

I like the how Rosa Cao (sites.google.com/site/luosha) & @dyamins.bsky.social speculated about task constraints here (doi.org/10.1016/j.co...). I think the Platonic Representation hypothesis is a version of their argument, for multi-modal learning.

Definitely! Task constraints certainly play a role in determining representational structure, which might interact with what we consider here (efficiency of implementation). We don't explicitly study it. Someone should!

Main takeaway: Valid representational comparison relies on implicit assumptions (task-optimization *plus* efficient implementation). ⚠️ More work to do on making these assumptions explicit!

🧠 CCN poster (today): 2025.ccneuro.org/poster/?id=w...

📄 ICML paper (July): icml.cc/virtual/2025/poster/44890

Our theory predicts that representational alignment is consistent with *efficient* implementation of similar function. Comparing representations is ill-posed in general, but becomes well-posed under minimum-norm constraints, which we link to computational advantages (noise robustness).

Function-representation dissociation in ReLU networks. (A-B) MNIST representations before/after prediction-preserving reparametrisation. (C) RSM after function-preserving reparametrisation. (D-E) Performance under input/parameter noise for different solution types.

Function-representation dissociations and the representation-computation link persist in deep nonlinear networks! Using function-invariant reparametrisations (@bsimsek.bsky.social), we break representational identifiability but degrade generalization (a computational consequence).

Hidden-layer representations for a semantic hierarchy task. (A) Task structure. (B) Input/target encoding. (C-E) Hidden representations and representational similarity matrices for task-agnostic (C: LSS) vs. task-specific (D: MRNS, E: MWNS) solutions.

We demonstrate that representation analysis and comparison is ill-posed, giving both false negatives and false positives, unless we work with *task-specific representations*. These are interpretable *and* robust to noise (i.e., representational identifiability comes with computational advantages).

The solution manifold. (A) Solution manifold for a 3-parameter linear network, showing GLS and constrained LSS, MRNS, and MWNS solutions. (B-E) Input/output weight relationships and parametrisation structure for each solution type.

We parametrised this solution hierarchy to find differences in handling of task-irrelevant dimensions: Some solutions compress away (creating task-specific, interpretable representations), while others preserve arbitrary structure in null spaces (creating arbitrary, uninterpretable representations).

Task solution hierarchy defined by implicit regularisation objectives.

To analyse this dissociation in a tractable model of representation learning, we characterize *all* task solutions for two-layer linear networks. Within this solution manifold, we identify a solution hierarchy in terms of what implicit objectives are minimized (in addition to the task objective).

Example of a failure case. (A) A random walk on the solution manifold of a two-layer linear network reveals that weights can change continuously, inducing changes in the (B) network parametrisation and thus the (C) hidden-layer representations, while preserving the (D) network output.

Deep networks have parameter symmetries, so we can walk through solution space, changing all weights and representations, while keeping output fixed. In the worst case, function and representation are *dissociated*.

(Networks can have the same function with the same or different representation.)

Are similar representations in neural nets evidence of shared computation? In new theory work w/ Lukas Braun (lukasbraun.com) & @saxelab.bsky.social, we prove that representational comparisons are ill-posed in general, unless networks are efficient.

@icmlconf.bsky.social @cogcompneuro.bsky.social

Co-organized with @susanneharidi.bsky.social, @marcelbinz.bsky.social, Rodrigo Carrasco-Davis, @clementinedomine.bsky.social‬, @eringrant.me, @modirshanechi.bsky.social‬ 🌳

Want to contribute to this debate at #CCN2025? Please come to our session today, fill out the anonymous survey (forms.gle/yDBBcBZybGjogksC8), and comment on the GAC page (sites.google.com/ccneuro.org/gac2020/gacs-by-year/2025-gacs/2025-1)! Your perspectives will shape our eventual GAC paper. 👥