York University Study Finds AI Models May Not Process Visual Information Like the Brain
- Published
- Source
- ScienceDaily.com
York University researchers report that artificial neural networks (ANNs) widely used to study visual processing may not work like the human brain, despite their ability to predict patterns of brain activity. The study’s senior author, Assistant Professor Kohitij Kar, says earlier tests mainly asked whether AI models could predict brain activity. The team instead tested whether recorded brain activity could predict what was happening inside the AI models.
The researchers developed a method called reverse predictivity testing to examine this relationship in both directions. Their approach addresses a limitation in conventional tests: a model’s ability to reproduce brain responses does not, by itself, establish that it processes visual information through similar mechanisms. The findings raise questions about how reliably these ANNs can serve as computational models for understanding biological vision, including object recognition and motion perception.
The team tested responses to 1,320 natural photographs and realistic synthetic images featuring 10 familiar subjects, as well as 300 additional images showing the same subjects in different visual styles. The reported results identify a significant disconnect between AI and biological vision. The article does not provide detailed results for each image category or disclose a timeline for further research; those details remain unresolved.
Original report
AI may not work like the human brain after all, scientists discover
ScienceDaily.comThis briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
-
4
RobotToday Weekly November 9–13, 2026 RobotToday · 1 week ago
- 5
- 6
- 7
- 8
- 9
-
10
RobotToday Weekly November 2 – 6, 2026 RobotToday · 2 weeks ago
- 11
- 12
Related Suppliers
Simple AI 深朴智能
Beijing embodied-AI startup building living-space service robots; its HiFi-UMI robot-free capture rig shipped with an open 2,000-hour dataset.
OneRobotics (Shenzhen) Co., Ltd. 卧安机器人(深圳)股份有限公司
HKEX-listed Shenzhen home-robot maker behind SwitchBot, the KATA Friends companion robot, the onero H1 home humanoid and the Acemate tennis robot.
BitRefine Group
AI and computer vision solutions company headquartered in Hong Kong, serving manufacturing, transportation, healthcare and other industries.
Xiamen Reconova Information Technology Co., Ltd. 厦门瑞为信息技术股份有限公司
Xiamen visual-AI company (HKEX: 07656) whose commercial robot division builds the Xiaoyi/AntOne airport baggage transfer robot.
BrainCo 浙江强脑科技有限公司
Hangzhou brain-computer interface firm founded 2015 at Harvard; makes FDA-registered bionic hands, a Revo2 dexterous hand for humanoids, and EEG wearables.
Neuracle 博睿康科技(常州)股份有限公司
Chinese brain-computer interface company behind NEO, the world's first commercially approved implantable BCI, and NeuSen EEG acquisition systems.
Sber Robotics Center
Sberbank's robotics unit building the GREEN humanoid (180 cm, 41 DOF) and WORKER semi-humanoid, controlled by its Green-VLA model.
Weizmann Institute of Science
Rehovot, Israel research university founded 1934, with five faculties including Mathematics and Computer Science; grants only M.Sc. and Ph.D. degrees.
Recognition Robotics Inc.
US robotic vision software company providing 2D/3D guidance systems using the patented CortexRecognition algorithm for automotive, aerospace, and industry.
KAIST Hubo Lab (Humanoid Robot Research Center)
KAIST humanoid research centre that built HUBO and won the 2015 DARPA Robotics Challenge with DRC-HUBO+; it now develops a 3.25 m/s next-gen humanoid.
Dahlia Robotics
German agricultural robotics startup developing fully electric autonomous robots for AI-driven mechanical weed removal to eliminate herbicide use.
IDSIA – Dalle Molle Institute for Artificial Intelligence (USI-SUPSI)
Swiss AI research institute (USI-SUPSI, Lugano) founded 1988; known for deep-learning work under Jurgen Schmidhuber.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.