York University Study Finds AI May Not Mirror the Brain After All
Artificial intelligence may be less brain-like than scientists have assumed, according to a new York University study that questions how closely these systems resemble biological vision.
Artificial intelligence may be less brain-like than scientists have assumed, according to a new York University study that questions how closely these systems resemble biological vision.
Researchers have long used artificial neural networks to study how the brain processes visual information, since the models can reproduce patterns of brain activity. But the team asked whether the relationship works in both directions. Using a method called reverse predictivity testing, they checked whether recorded neuron activity could predict what happens inside the AI models.
AI models predicted neuronal responses well, but the reverse failed: brain activity was much less effective at predicting many of the models' internal features. The same problem did not appear when comparing neurons from different brains.
The team tested responses to 1,320 natural and synthetic images of ten familiar subjects, plus 300 images in different visual styles. The findings, published in Nature Machine Intelligence, suggest AI can mimic brain output while relying on different internal strategies. The researchers released a public testing toolkit for AI developers.
Source: Science Daily (Popular)
Ready to Modernize Your Business?
Get your AI automation roadmap in minutes, not months.
Analyze Your Workflows →