2026 Symposium

14th Annual Marshall M. Weinberg Symposium

The symposium will take place in the Rackham Amphitheatre and Assembly Hall on Friday, February 20th, 2026, 1-6 PM.

Neuro AI: Can Generative AI Help Us Understand the Human Brain?

Generative AI systems display remarkable cognitive abilities once thought unique to humans. Can these models—built from silicon rather than neurons—help us understand how the brain produces perception, thought, and action? This symposium gathers leading experts in psychology, neuroscience, and AI to explore the emerging field of “NeuroAI”, a research program that promises to transform our understanding of the human brain.

2026 Featured Speakers

Our featured speakers are renowned experts in their respective disciplines, offering diverse perspectives in cognitive science.

Please click on their names below to read more about them.

Jeffrey Bowers

Jeffrey Bowers is a Professor in the School of Psychological Science at the University of Bristol. Jeffrey received his PhD in psychology from the University of Arizona in 1993. He was an Assistant Professor at Rice University from 1994-1998 and then moved to the University of Bristol in 1998, where he has been full professor since 2006. His work includes cognitive, neuroscience, and computational studies of language, memory, and perception. For past decade he has focused on the relation between artificial and natural intelligence.

Current Research Practices in NeuroAI Do Not Support the Many Strong Claims of ANN-Human Alignment

Abstract: Artificial neural networks (ANNs) developed in computer science are successful in a range of vision, language, and reasoning tasks. They can also predict behavioural responses and brain activations of humans better than alternative models. This has led to the common claim that ANNs are the best models of biological intelligence. However, most prediction studies are correlational, and accordingly, do not support causal conclusions. Furthermore, researchers are incentivized to identify ANN-human similarities, as reviewers and editors are more likely to publish studies that report similarities rather than differences. Accordingly, researchers rarely carry out “severe” tests of their claims that are more likely to falsify their conclusions (if indeed the conclusions are false). I show when the relevant experiments are carried out, ANNs do a poor job in explaining human intelligence. The field of NeuroAI needs to change its methods to better characterize ANN-human alignment and build better models of minds.

Ev Fedorenko

Evelina Fedorenko is a cognitive neuroscientist who studies the human language system and its relationship with other brain systems. She received her Bachelor’s degree from Harvard in 2002, and her Ph.D. from MIT in 2007. She was then awarded a K99R00 Pathway to Independence Career Development Award from NIH. In 2014, she joined the faculty at MGH/HMS, and in 2019 she returned to MIT where she is currently an Associate Professor in the Department of Brain and Cognitive Sciences and a Member of the McGovern Institute for Brain Research. Fedorenko uses fMRI, intracranial recordings and stimulation, EEG, MEG, and computational modeling to study language and cognition in adults and children, including those with developmental and acquired brain disorders, and structurally atypical brains. In 2025, her work was recognized with the Troland award from the NAS.

Neural Network Language Models as Models of Language Processing in the Human Brain

Abstract: A network of frontal and temporal areas in the human brain supports language processing. This “language network” a) is robustly dissociated from lower-level speech perception and articulation mechanisms, and from systems of reasoning (Fedorenko et al., 2024 NRN); and b) supports computations related to retrieving words from memory and building syntactic structures in the service of semantic composition (Shain&Kean et al., 2024 JOCN). However, a mechanistic-level understanding of how we extract meanings from word sequences, or express meanings through language has long remained elusive, in large part due to the limitations of human neuroscience approaches. Recently, a new candidate model organism emerged, albeit not a biological one, for the study of language—neural network language models (LMs). These models exhibit human-level performance on diverse language tasks, and their internal representations are similar to the representations in the human brain when processing the same linguistic inputs (Schrimpf et al., 2021 PNAS; Tuckute et al., 2024 NHB). I will talk about how we can use LMs to evaluate hypotheses about language processing, development, and impairments at an unprecedented granularity and scale. I will also touch on how neural networks can be used to investigate how the language system interacts with systems of thought.

Dan Yamins

Dan Yamins is a computer scientist and computational neuroscientist at Stanford University, where he's an associate professor of Computer Science and Psychology, faculty scholar at the Wu Tsai Neurosciences Institute, and an affiliate of the Stanford Artificial Intelligence Laboratory and the Stanford Vision Lab. His research group focuses on reverse engineering the algorithms of the human brain, both to learn both about how our minds work and build more effective artificial intelligence systems.

World Modeling with Probabilistic Structure Integration

Abstract: I will discuss Probabilistic Structure Integration (PSI), our new approach to world modeling. PSI has three basic components: (i) an efficiently-learnable probabilistic graphical model of the world, (ii) generic zero-shot structure estimation via causal inference in the underlying model, and (iii) a generic mechanism for re-integrating extracted structures as novel token types, continually improving the underlying model. I will discuss these core PSI principles, how they have generated state-of-the-art results in a variety of computer vision competencies, how they can apply beyond vision, and how PSI can help us toward improved brain and cognitive models.

Search: {{$root.lsaSearchQuery.q}}, Page {{$root.page}}

for