Interested in doing a PhD with me and lacns.github.io? Or with any of the incredible fellows in the IMPRS School of Cognition www.maxplanckschools.org/cognition-en - apply before Dec 1st at cognition.maxplanckschools.org/en/application

In our forthcoming paper, John Hummel and I ask what it would mean for a neural computing architecture such as a brain to implement a symbol system, and the related question of what makes it difficult for them to do so, with an eye toward the differences between humans, animals, and ANNs.

Large Language models (LLMs) do not simulate human psychology. That's the title of our new paper, available as preprint today (1/12):

arxiv.org/abs/2508.06950

What does it mean if pure prediction fails? If you get 100% pure prediction you still don’t know how the model predicted unless you run an experiment that manipulates independent variables. A digital clock 100% predicts the time of a cuckoo clock but it works totally differently.

What's the deal with negatively accelerating search functions in relational searches? The CASPER model of visual search shows how emergent features may allow for parallel processing in searches we'd expect to be steep and linear. Visit me Tuesday afternoon in the Banyan Breezeway! #VSS2024

I've got an exciting #Visionscience and #STEMed announcement: My textbook "Practical Vision: Learning through Experimentation" is now in production at Routledge Books! The book focuses on hands-on, analog exercises to identify and discuss key mechanisms of human vision. Coming October 2024!

Clear writing is (imperfect) evidence of clear thinking. The use of LLMs for writing is IMO often inexcusable, substituting for one’s own voice the median voices of the past and deceiving one’s audience with an incorrect picture of one’s understanding (corrupting their training data, so to speak).

Building larger LLMs to get AGI is like linearly accelerating towards light speed

Excited to share @rgast.bsky.social's new PNAS paper from the lab, with Sara Solla.

We ask: does cell type heterogeneity affect what neural networks can compute? How might different brain regions leverage heterogeneity to achieve different things?

www.pnas.org/doi/10.1073/...

We then all agree in one respect: There is a lot of hype.  But the problem is with researchers who take the hype seriously?That we should focus on the strengths rather than weaknesses of DNNs? That there is little "confusion that deep neural networks (DNNs) are ‘models of the human visual system’”?🤷‍♂️

This is so cool:
www.pnas.org/doi/10.1073/...
Bacteria store a memory of swarming proficiency (measured as the time lag to start swarming on suitable media) in the form of intracellular iron levels. This memory can be passed down for 4 generations!

A lot of the neuroscience work, particularly in hippocampus, has been focused on content-addressable memory, where data and address are the same. But this might not be the right way to think about memory in the brain. Maybe we have an addressing system that is separate from stored content.