I had the pleasure of being interviewed on the @i-am-scientist.bsky.social podcast by @lisaschmors.bsky.social and @philipphubert.bsky.social . I talk about how I ended up writing my newsletter (path2phd.substack.com) and the people who supported me all along. Enjoy!

The path to PhD - Advice from a young scientist

I am Scientist · Episode

open.spotify.com

So—does this mean theory is doomed, and AI engineering is just a random walk? Not at all! 💡 @rpatrik96.bsky.social @wielandbrendel.bsky.social Randall Balestriero did an amazing job clarifying where theory can help practice—and where practice should inspire theory. 🤝

Kudos to the great team @ninamiolane.bsky.social @rpatrik96.bsky.social @charlesoneill.bsky.social Harald Maurer 🙏

How can neural nets extract *interpretable* features from data—& uncover new science? 👉 Discover our mathematical framework tackling this question w/ identifiability theory, compressed sensing, interpretability & geometry!🌐 By @david-klindt.bsky.social @rpatrik96.bsky.social C. O'Neill H Maurer

David Klindt@david-klindt.bsky.social · 2y ago

🔵 New paper! We explore sparse coding, superposition, and the Linear Representation Hypothesis (LRH) through identifiability theory, compressed sensing, and interpretability. If you’re curious about lifting neural reps out of superposition, this might interest you! 🤓 arxiv.org/abs/2503.01824

New preprint!🚀 Decoding neural representations is a challenge in neuroscience & AI. 👉 Learn how identifiability theory, compressed sensing & interpretability research -w/ a dash of geometry- can help! @david-klindt.bsky.social @rpatrik96.bsky.social C. O'Neill H. Maurer @ninamiolane.bsky.social

David Klindt@david-klindt.bsky.social · 2y ago

🔵 New paper! We explore sparse coding, superposition, and the Linear Representation Hypothesis (LRH) through identifiability theory, compressed sensing, and interpretability. If you’re curious about lifting neural reps out of superposition, this might interest you! 🤓 arxiv.org/abs/2503.01824

🚀 We’re hiring! Join Bernhard Schölkopf & me at @ellisinsttue.bsky.social to push the frontier of #AI in education! We’re building cutting-edge, open-source AI tutoring models for high-quality, adaptive learning for all pupils with support from the Hector Foundation. 👉 forms.gle/sxvXbJhZSccr...

Check out our @neuripsconf.bsky.social spotlight on showing how language models learn to extrapolate and compose language rules. See you on Fri 13 Dec 4:30 p.m. PST — 7:30 p.m. PST at poster #2702 in East Hall. neurips.cc/virtual/2024...

Congrats to the recipients of the 2024 ELLIS PhD Award! Co-winners: @koloskova.bsky.social (efficiency in decentralized learning) & Luigi Gresele (identifiable representation learning) Runner up: @schwarzjn.bsky.social (sparse parameterizations) Read more about them: bit.ly/4fg4jkg #AI #ML

1/ Hi all, I am at #NeurIPS2024 and I will be hiring a postdoc in probabilistic machine learning starting asap. Research interests: amortized, approximate & simulator-based inference, Bayesian optimization, and AI4science. Get in touch for a chat or come to our posters today 11AM or Friday 11AM!

📄 New Paper: "How to Merge Your Multimodal Models Over Time?" arxiv.org/abs/2412.06712 Model merging assumes all finetuned models are available at once. But what if they need to be created over time? We study Temporal Model Merging through the TIME framework to find out! 🧵

How to Merge Your Multimodal Models Over Time?

Model merging combines multiple expert models - finetuned from a base foundation model on diverse tasks and domains - into a single, more capable model. However, most existing model merging approaches...

arxiv.org

We are looking for a PostDoc on human-AI interaction to work with @mirizilka.bsky.social The project studies systems used in law enforcement or legal decision-making, when humans act on recommendations/predictions made by ML algorithms. Email for details. www.jobs.cam.ac.uk/job/49538/

Check out our @neuripsconf.bsky.social spotlight on showing a useful inductive bias in language models to extrapolate language rules. See you on Fri 13 Dec 4:30 p.m. PST — 7:30 p.m. PST at poster #2702.

Ferenc Huszár@inference.vc · 2y ago

Can language models transcend the limitations of training data? We train LMs on a formal grammar, then prompt them OUTSIDE of this grammar. We find that LMs often extrapolate logical rules and apply them OOD, too. Proof of a useful inductive bias. Check it out at NeurIPS: nips.cc/virtual/2024...

Next week I will be in Berkeley. Reach out if you'd like to talk about OOD/compositional generalization, causality, SSL, or inductive biases.

The ✨ML Internship Feed✨ is here! @serge.belongie.com and I created this feed to compile internship opportunities in AI, ML, CV, NLP, and related areas. The feed is rule-based. Please help us improve the rules by sharing feedback 🧡 🔗 Link to the feed: bsky.app/profile/did:...

Join the Interdisciplinary Postdoc Fellowship Program at the European Molecular Biology Laboratory (EMBL), one of the best places to do research in modern biology and develop your career. Great opportunities for statisticians, comp. biologists, AI experts, mathem. modelers! www.embl.org/eipod-linc