This was a fun discussion - thanks BIU for the invite!
LLM introspection revisited. if we do the controls properly we may not have strong enough evidence just yet
arxiv.org/html/2605.26...
To many of us, myself included, the greatest puzzle in neuroscience is how our brains give rise to our very rich perceptual experience of the world. But studying the neural correlates of rich perception is difficult, so we usually resort to very simple stimuli. 1/3 #neuroskyence
Yes, I agree! And some emergent capabilities of the models are just hard to dismiss as "fancy auto-complete", e.g. this amazing paper from Jack Lindsey arxiv.org/abs/2601.01828
who wants to come work with us, on whether AIs can introspect? e.g. something similar to this - transformer-circuits.pub/2025/introsp... - but maybe especially in the perceptual domain?
we just got some new funds to collaborate with our friends at Yale (on foundational models for fMRI data), & ...
(YouTube) Steve Fleming on "Sensory horizons and the functions of conscious vision" www.youtube.com/watch?v=euGC... with @smfleming.bsky.social
Last call to apply! ⏰ Applications close 10 June!
⬇️⬇️⬇️
cifar.ca/next-generat...
#NeuroscienceResearch #ConsciousnessStudies #PostdocOpportunities #AIConsciousness #NeuroscienceOfConsciousness