Why do diverse ANNs resemble brain representations? Check out our new paper with Colton Casto, @nogazs.bsky.social , Colin Conwell, Mark Richardson, & @evfedorenko.bsky.social on “Universality of representation in biological and artificial neural networks.” 🧠🤖
https://tinyurl.com/yckndmjt
https://tinyurl.com/yckndmjt
Comments
Understanding why could reveal shared computational principles in 🧠 & 🤖. (2/n)
1. If ANNs share similar brain-like representations, separating them would also distance their representations from the 🧠.
2. If some ANNs are inherently better models, separation will highlight these, keeping the best aligned with the 🧠.(4/n)
Our question: how does reducing agreement across ANNs affect their alignment with the brain? We tested this in language. (5/n)
𝗔. Low-agreement stimuli robustly reduced ANN agreement compared to previous stimuli.
𝗕. When ANNs disagree, their representations are less predictive of fMRI responses, supporting representation universality and the link between model agreement and brain alignment. (6/n)
If representation agreement across ANN models captures the universal part of representations (shared between 🤖 and the 🧠), modulating representation agreement should also modulate brain alignment. (7/n)