Skip to main content
World Today News
  • Home
  • News
  • World
  • Sport
  • Entertainment
  • Business
  • Health
  • Technology
Menu
  • Home
  • News
  • World
  • Sport
  • Entertainment
  • Business
  • Health
  • Technology
New Study Reveals AI Models May Not Process Visual Information Like the Human Brain

New Study Reveals AI Models May Not Process Visual Information Like the Human Brain

November 21, 2026 Technology Desk Technology

Artificial intelligence models frequently cited as “brain-like” may process visual information using internal logic that the human brain does not employ, according to a study published on Saturday.

Researchers at York University found that while these artificial neural networks (ANNs) can successfully predict neuronal activity, the reverse relationship—using brain activity to predict the internal states of AI—does not hold true.

The Tech TL;DR:

  • Current AI models achieve high predictive accuracy for brain activity despite relying on fundamentally different internal computational strategies.
  • The study introduces a “reverse predictivity” diagnostic metric to test whether AI mechanisms truly align with primate biological processes.
  • Researchers warn that using black-box models for clinical or psychological research may lead to misleading conclusions about human neural function.

For over a decade, neuroscientists have utilized ANNs to map how visual information moves through the brain. These systems are often categorized as brain-like because they demonstrate a capacity to mirror the neural responses observed when humans recognize objects. However, the York University team, led by Assistant Professor Kohitij Kar, questioned whether this predictive power implies a shared underlying mechanism or merely a mathematical coincidence.

“Artificial intelligence systems are often described as ‘brain-like’ because they can predict activity in parts of the brain that help us recognize objects,” said Kar. “Until now, scientists mostly tested this in one direction. They asked whether AI models can predict brain activity.”

To investigate the symmetry of this relationship, the team developed a “reverse predictivity” testing method. They examined neural responses to 1,320 natural and synthetic images, featuring 10 distinct subjects including animals, faces, and inanimate objects. The researchers also introduced 300 images in varying styles, such as sketches and artistic interpretations, to challenge the models’ categorization capabilities.

The findings, as reported by ScienceDaily, indicate a clear disconnect. While the AI models accurately predicted recorded neuronal responses, the biological data failed to predict the models’ internal features with equal precision. This asymmetry was not observed when comparing neural activity between different biological brains, suggesting that the AI is reaching its outputs through non-biological pathways.

# Example of a basic diagnostic check for model-brain alignment
def check_reverse_predictivity(neural_data, model_internal_states):
    # Testing if internal model states map back to known neural activation patterns
    correlation = compute_correlation(neural_data, model_internal_states)
    if correlation < threshold:
        return "Mismatch: Model utilizes non-biological internal strategy"
    return "Alignment: Potential biological plausibility"

The implications of this study, noted by DailyArtifact, suggest that researchers relying on these models for psychological or neurological insights may be interpreting “mathematical artifacts” as biological principles. If an AI achieves a correct result through a mechanism the brain does not use, the model fails as a digital proxy for human cognition. This presents a significant challenge for ongoing research into conditions such as autism, where models of the neurotypical brain serve as a baseline for understanding developmental differences.

“Our findings challenge how similar current AI systems really are with the primate brain,” said Postdoctoral Fellow Sabine Muzellec. “We show that models that were previously thought to be brain-like rely on internal components that the brain does not appear to use.”

This discrepancy highlights the necessity for more rigorous diagnostic standards in computational neuroscience. As AI continues to scale, the reliance on predictive power—without understanding the internal representational dynamics—risks creating an illusion of biological fidelity.

To address this, the York University team has released a testing toolkit designed to help developers evaluate their models for genuine alignment with biological mechanisms.

Moving forward, the research suggests that engineers may need to distinguish between raw computational efficiency and biological mimicry. While the former is sufficient for computer vision tasks, the latter is essential for clinical applications. The study does not advocate for the abandonment of ANNs, but rather emphasizes the need for architectural transparency and verification.

The question of how to bridge the gap between silicon-based logic and biological neural circuits remains a primary hurdle for the field. It is not yet known whether upcoming large-scale models will naturally converge toward biological processing or if they will continue to drift further from the mechanisms of the human mind.

The research team’s diagnostic framework is intended to provide a pathway for building more reliable, biologically grounded systems.

Always consult with certified IT and cybersecurity professionals before altering enterprise networks or handling sensitive data.

York University, established in 1959 and now Canada’s third-largest university with approximately 53,500 students, has developed significant research infrastructure including 32 research centres and the Lassonde School of Engineering. The institution began with just 76 students in 1960 and gained independence from the University of Toronto in 1965.

More on this story: Deep Learning Enables 3D Measurement From Single Image, Nature Reports · Apple Researchers Achieve Sub-50ms Interactive Latency in Video Generation With FastFrames

Share this:

  • Share on Facebook (Opens in new window) Facebook
  • Share on X (Opens in new window) X

Worth a look

  • Woman, 62, Dies After Neck Trauma on Catherine St.; Bloodied Man Handcuffed
  • Guillermo del Toro Says He Is Not Involved in Star Wars or Marvel Projects

Related

Search:

World Today News

World Today News is your trusted source for global journalism — breaking headlines, in-depth analysis, and reporting from around the world.

Quick Links

  • Privacy Policy
  • About Us
  • Accessibility statement
  • California Privacy Notice (CCPA/CPRA)
  • Contact
  • Cookie Policy
  • Disclaimer
  • DMCA Policy
  • Do not sell my info
  • EDITORIAL TEAM
  • Terms & Conditions

Browse by Location

  • GB
  • NZ
  • US

Connect With Us

© 2026 World Today News. All rights reserved. Your trusted global news source directory.
For contact, advertising, copyright, issues email: [email protected]

Privacy Policy Terms of Service