LLMs and the Brain: Autoregressive Mechanisms of Human Cognition | Dr. Elan Barenholtz Keynote

​Dr. Barenholtz's lecture and Q&A outlines how key computational features of transformers, such as embeddings, attention weighting, and autoregressive updating, align with known neurophysiological signatures including temporal decay, graded residual activation, and recurrent cortical loops. Behavioral findings—such as priming without explicit memory, temporal recency effects, limits in short-term memory, and confabulatory drift—are presented as evidence supporting this autoregressive view of cognition over traditional memory- or prediction-based accounts.

​Finally, the talk highlights Dr. Barenholtz’s research using LLM frameworks to model both typical and atypical cognitive processes, demonstrating how transformer-style architectures may offer a unified computational perspective on human thought and behavior.

​👀 Join the livestream here at 6PM ET! https://www.youtube.com/watch?v=p_4dwIY1rIk

​— — —

​Follow Elan's work online:

​• His recent video with Prof. Michael Levin: https://youtu.be/Fp7cmIS7NFI

​• YouTube Channel: /Elan_Barenholtz

​• X: /ebarenholtz

​• Elan's Substack: https://elanbarenholtz.substack.com