Abstract

Not included on Poster

NeuroAI aims to bridge the gap between biological and artificial intelligence by elucidating the specific mechanisms underlying emergent behaviors in behaving agents. Our study leverages the “thousand brains theory” through a preexisting platform called Monty, a sensorimotor learning agent that is structurally and functionally rooted in the mammalian neocortex. Monty’s architecture includes Learning Modules, analogous to cortical columns, which learn models of sensed objects, and Sensor Modules, representing sensory organs, which relay information about the external world into the neocortex. Conditions such as autism and schizophrenia are often associated with impaired sensory processing, yet the link between these impairments and specific neocortical circuitry remain poorly understood. To probe this relationship within a model, we introduce varying levels of sensory noise into Monty’s input streams. Our analysis examines how these different levels of sensory noise impact the agent’s object recognition accuracy. The results provide initial insights for investigating how alterations in sensory input might translate to cognitive impairments observed in neurological disorders.

Background

Active inference is a computational framework for action and perception and has been linked to various psychiatric disorders (Montague et al., 2012). From this point of view, psychopathological conditions can be understood as disorders of (active) inference. False inference could explain symptoms, like hallucinations or delusions in schizophrenia (Fletcher and Frith, 2009)

Methods

Results

Conclusions

Full Reference List

• Solomonoff, Grace. (2022). A Thousand Brains and a Million Theories. In Matthew Iklé, Ben Goertzel, & Alexey Potapov (Eds.), Artificial General Intelligence (Vol. 13154, pp. 250–260). Springer International Publishing AG. https://doi.org/10.1007/978-3-030-93758-4_26

• Clay, Viviane, Leadholm, Niels, & Hawkins, Jeff. (2024). The Thousand Brains Project: A New Paradigm for Sensorimotor Intelligence. arXiv.Org.

• Scharfen, Hans-Erik, & Memmert, Daniel. (2024). The model of the brain as a complex system: Interactions of physical, neural and mental states with neurocognitive functions. Consciousness and Cognition, 122, Article 103700. https://doi.org/10.1016/j.concog.2024.103700

• Zador, Anthony, Escola, Sean, Richards, Blake, Ölveczky, Bence, Bengio, Yoshua, Boahen, Kwabena, Botvinick, Matthew, Chklovskii, Dmitri, Churchland, Anne, Clopath, Claudia, DiCarlo, James, Ganguli, Surya, Hawkins, Jeff, Körding, Konrad, Koulakov, Alexei, LeCun, Yann, Lillicrap, Timothy, Marblestone, Adam, Olshausen, Bruno, … Tsao, Doris. (2023). Catalyzing next-generation Artificial Intelligence through NeuroAI. Nature Communications, 14(1), Article 1597. https://doi.org/10.1038/s41467-023-37180-x

• Brown, Harriet, Adams, Rick A., Parees, Isabel, Edwards, Mark, & Friston, Karl. (2013). Active inference, sensory attenuation and illusions. Cognitive Processing, 14(4), 411–427. https://doi.org/10.1007/s10339-013-0571-3

• Hartl, Benedikt, et al. “Remapping and navigation of an embedding space via error minimization: a fundamental organizational principle of cognition in natural and artificial systems.” arXiv preprint arXiv:2601.14096 (2026).

• Kawakami, Hajime. “Finding Similar Objects and Active Inference for Surprise in Numenta Neocortex Model.” arXiv preprint arXiv:2506.21554 (2025).

• Möller, T. J., Georgie, Y. K., Schillaci, G., Voss, M., Hafner, V. V., & Kaltwasser, L. (2021). Computational models of the “active self” and its disturbances in schizophrenia. Consciousness and cognition, 93, 103155. https://doi.org/10.1016/j.concog.2021.103155