Kording Lab 🦖

Konrad kording, @Penn Prof, deep learning, brains, #causality, rigor, Transdisciplinary optimist, Dad, Loves outdoors, 🦖

[Ching-Lung Hsu @hiallen72]

https://t.co/DY7IvMhAem

[unusual_whales @unusual_whales]

Former Mayo Clinic research director and AI compliance lead Traci Tamiko Eto alleges that the globally recognized hospital retaliated against her for blowing the whistle on its disastrous implementations of AI health tools

Read more: https://t.co/5XIV4QnDCL

[Ilenna Jones, PhD @ ilennaj.bsky.social @IlennaJ]

What does synapse position buy a neuron? We trained a dendritic network to place its own synapses on a covariance task. It clustered correlated inputs onto shared branches, and shuffling that placement destroyed performance.
Preprint: https://t.co/3Zui4an2zK
0/

[Arjun Raj @arjunrajlab]

This is, in many ways, the whole bet: does superintelligence help us predict our way out of the causal complexity barrier? Or is it fundamentally unpredictable? I suppose a reasonable bet could be made either way, although I do think humans overestimate their own abilities.

[Kording Lab 🦖 @KordingLab]

@SergeyStavisky @esedaghatnejad I mean it’s your paper so you’d know better

[Kording Lab 🦖 @KordingLab]

@SergeyStavisky @esedaghatnejad I thought you meant more broadly for signals. But what about this? https://t.co/voG90XOSO0

[Reza Shadmehr @reziliusReza]

How does the brain know which neurons to blame when a movement goes wrong?
In machine learning, gradient descent solves this credit-assignment problem. We asked whether the cerebellum might do something similar.
https://t.co/SLSAfQwIPG

[Misha Teplitskiy | Science of Science @MishaTeplitskiy]

This makes sense

[Brokoslaw Laschowski @DrLaschowski]

Come join us in Toronto. @UofT is hiring an Associate Professor/Professor and endowed Hinton Chair in Artificial Intelligence.

https://t.co/BrrE3tusOV

[Kording Lab 🦖 @KordingLab]

Very well put. And a lack of causal identifiabilitynis one of the factors limiting how high intelligence can go.

[François Chollet @fchollet]

One of the biggest misconceptions people have about intelligence is seeing it as some kind of unbounded scalar stat, like height. "Future AI will have 10,000 IQ", that sort of thing. Intelligence is a conversion ratio, with an optimality bound. Increasing intelligence is not so much like "making the tower taller", it's more like "making the ball rounder". At some point it's already pretty damn spherical and any improvement is marginal.

Now of course smart humans aren't quite at the optimal bound yet on an individual level, and machines will have many advantages besides intelligence -- mostly the removal of biological bottlenecks: greater processing speed, unlimited working memory, unlimited memory with perfect recall... but these are mostly things humans can also access through externalized cognitive tools.

[Kording Lab 🦖 @KordingLab]

Tweetprint https://t.co/30zMkruN5D

[Turan Orujlu @TOrujlu]

UAI 2026 paper: A genetic perturbation can change the connectome, a new policy can rewire economic links between firms. In each case, interventions change the graph itself which standard SCMs treat as fixed.

We tackle this in Partially Observed Structural Causal Models (POSCMs).

[Kording Lab 🦖 @KordingLab]

Us trying to broaden the modeling of causality. With Turan orujlu.

[Turan Orujlu @TOrujlu]

UAI 2026 paper: A genetic perturbation can change the connectome, a new policy can rewire economic links between firms. In each case, interventions change the graph itself which standard SCMs treat as fixed.

We tackle this in Partially Observed Structural Causal Models (POSCMs).

[Kording Lab 🦖 @KordingLab]

Link https://t.co/tGgBVz5sDx

[Aran Nayebi @aran_nayebi]

I sometimes worry/wonder if Dune might end up predicting the long-term future we steady state to:

[Kording Lab 🦖 @KordingLab]

A truly embarrassing advertisement by https://t.co/9FgKNzNcr5 spotted in nyc: lots of obnoxious (and vacuous) ai-speak sentences. Vanilla Claude visuals. No one should put out anything that is not better than AI. Work with ai. Don’t have ai do your work. Only eat good dog food.

[Kording Lab 🦖 @KordingLab]

@vineettiruvadi But then the answer is a million a year or so?

[Sholto David @addictedtoigno1]

Today I'm posting a blog about another ridiculous antibody blunder. Dozens of papers in the senescence and ageing field have mistakenly used an antibody for E. coli β-galactosidase when trying to stain for the human protein with the same name. Link in next comment 👇🏼