Georg Keller

and the idea of internal models for motor control (Jordan, Rumelhart, Kawato, Wolpert) – combined with the beautiful music box spacetime attractor ideas coming out of @behrenstimb.bsky.social ’s lab are the most promising approach we currently have to try to understand how cortex works.

We suspect, a model that combines the self-supervised learning of JEPA (@yann-lecun.bsky.social and team), the credit assignment and capacity to work on arbitrary graphs of predictive processing (work of the lab of Rafal Bogacz),

We think cortex might function like a JEPA. It looks like prediction errors in layer 2/3 are not computed against input (as is the idea in predictive processing), but against a representation in latent space (i.e. like in a JEPA arxiv.org/abs/2301.08243 or RPL doi.org/10.1101/2025...).

The Neurobiology of Mental Health 2026 from May 17 – May 21 in Thun, Switzerland.

Join leading scientists for a 3.5-day #LakeConference discussion on: "The Neurobiology of Mental Health" at the level of genes, cell types, organoids, circuits, networks and brain systems. Learn more & apply by January 31: https://lakeconferences.org/conf/1ed63fdb-de0c-4a56-a8b3-1f72f93d30ea

Join us this May to explore the biological mechanisms of mental health challenges and treatment strategies. Apply by January 31: lakeconferences.org

1/6 New preprint 🚀 How does the cortex learn to represent things and how they move without reconstructing sensory stimuli? We developed a circuit-centric recurrent predictive learning (RPL) model based on JEPAs.
🔗 doi.org/10.1101/2025...
Led by @atenagm.bsky.social @mshalvagal.bsky.social

Understanding cortical computation through the lens of joint-embedding predictive architectures Tracking prey or recognizing a lurking predator is as crucial for survival as anticipating their actions. To guide behavior, the brain must extract information about object identities and their dynami...

One of the most promising approaches to making headway in understanding the cortical algorithm that I have seen in a long time! www.biorxiv.org/content/10.1...

Inside the tunnel: Experiencing the brain’s response to sensory mismatches In this first-person account, FMI’s senior communications manager describes taking part in an early human trial that adapts previous experiments in mice to explore how the human brain responds when vi...

In this first-person account, FMI’s senior communications manager describes taking part in an early human trial that adapts previous experiments in mice to explore how the human brain responds when visual and auditory information suddenly fall out of sync. www.fmi.ch/news-events/...

Applications are open for the #LakeConference on the Neurobiology of Mental Health in Lake Thun, Switzerland.

📆 May 17-21, 2026
🏔️ All career stages welcome
⏳ Apply by January 31, 2026

Learn more and apply: https://t.co/WcuXHdWnFa

@lakeconferences.bsky.social

Very much looking forward to this collaboration with @georgkeller.bsky.social at @fmiscience.bsky.social, using a cross-species approach in humans and mice to investigate the effects of antipsychotics on cortical circuit function.

Join us for the second Neurobiology of Mental Health conference (May 2026) that will explore the biological mechanisms underlying mental health challenges and their treatment. Information and application on: lakeconferences.org. The deadline for applications is January 31st, 2026.

Stellenangebot Postdoctoral Researcher (PostDoc m/f) 100% Translational Psychiatry Lab for 2 Years bei Universitäre Psychiatrische Kliniken Basel UPK Forschung in Klinik für Erwachsene (UPKE)

Please share: Postdoc position available in a new collaborative SNSF project with @georgkeller.bsky.social at @fmiscience.bsky.social using a cross-species approach in humans and mice to investigate the cortical circuit mechanisms underlying schizophrenia. Apply at karriere.upk.ch/Postdoctoral...

I’m super excited to finally put my recent work with @behrenstimb.bsky.social on bioRxiv, where we develop a new mechanistic theory of how PFC structures adaptive behaviour using attractor dynamics in space and time!

www.biorxiv.org/content/10.1...

This year's Ruth Chiquet Prize goes to @solygamagda.bsky.social, who sent a video message, for her work on how the brain detects sensory mismatches. Read more at: www.fmi.ch/news-events/...

🚨 We're hiring, please share! The FMI seeks a tenure-track Group Leader (Assistant Prof) in Structural Biology 🔬
Innovative scientists in genome regulation, RNA metabolism, or protein homeostasis—especially using cutting-edge approaches—apply now at www.fmi.ch/education-ca...

I often get asked “do you think visuomotor mismatch responses are a mouse thing?” – it looks like at least humans have them as well. very excited by this!

If so, this would mean - as you also propose - that the responses indeed won't always be easily interpretable.

My current best guess is that cortex implements something like a JEPA and uses local error computations to implement credit assignment (or backprop if the system is hierarchical) the way Rafal Bogacz has suggested.

I also think the predictive processing model is wrong, but for slightly different reasons. These pertain to the way superficial and deep layers in cortex appear to interact that are not consistent with predictive processing (@loghyr.bsky.social‬ will have a preprint on this soon).

i.e. if the coding space of your population can represent bimodal distributions (in a non-trivial way), so can the prediction errors. And regarding the role of prediction errors in driving plasticity, the problem is the same for gradient descent in backprop.

Good point, but I don’t think that is different from other models of cortical function? Regarding the role of prediction errors in updating internal representations, the coding space of predictions, prediction errors, and internal representation just needs to match.

Intuitively, I think this would make sense – imagine opening a cookie jar to find a mouse having eaten your last cookie, I suspect your brain will prioritize the unexpected presence of the mouse, to only later be irked by the unexpected absence of the cookie.

My interpretation of this is that there likely is an inhibitory competition between negative and positive prediction errors that always ends up favoring positive prediction errors.

What I think Sonja’s paper very nicely shows is that presenting both positive and negative prediction errors (i.e. a stimulus substitution), drives only a positive prediction error response.

A negative prediction error neuron for stimulus surround (a la Rao&Ballard) might not respond to a visuomotor mismatch, etc.

There could be dedicated prediction error neurons to the different predictions (e.g. a visuomotor prediction, an audio-visual prediction, a stimulus surround prediction). It is still unclear how different predictions are combined.

Re 2) Also agreed. But there are multiple viable interpretations here. Assuming we let go of the idea of cortex as hierarchy (cortex is a hierarchy, like the earth is flat) – then there must are multiple predictive inputs.

The notion of ‘generalized predictive coding’ sounds similar to the assumption that if something is predictable in principle, the brain must predict it. I don’t think there ever was good evidence to support that assumption?

The two instances we have solid evidence for prediction errors are stimulus surround (i.e. statistics of natural images) and visuomotor coupling (i.e. physics of the world) – both likely behaviorally relevant.

Re 1) Correct, and why would it? We can easily construct (artificial) examples of perfectly predictable sequences of images we would intuitively agree the brain likely does not predict (think e.g. sequences of random visual noise that repeats every X s or so).