Neural Scaling Laws Have Been Identified
Researchers found power laws that allow brain sensory capacity to grow with neural populations.
Updated on Sept. 25, 2026 in Alzheimer’s
Live Poll
Does learning that biological brains scale indefinitely change your confidence in artificial intelligence development?
Scientists have identified invariant power laws governing how neural populations process information. These findings disprove a long-standing hypothesis that neural noise limits the brain's sensory capacity.
Why it matters
The discovery resolves a paradox in neuroscience regarding why the brain utilizes so many neurons. It suggests that sensory information capacity scales continuously rather than hitting a ceiling.
Researchers analyzed 18,000 to 21,000 neurons from the mouse primary visual cortex. The study identified two scale-invariant power laws that govern noise strength and signal alignment.
The players
Kyoto University
This is a leading Japanese research university that served as the primary institution for the study.
Harvard University
This private Ivy League research university in Massachusetts provided collaborative support for the neural mapping project.
UCLA
The University of California, Los Angeles is a prominent public research institution that collaborated on the findings.
The details
Using two-photon calcium imaging and electrophysiological datasets, researchers decomposed neural noise into mathematical patterns called eigenmodes. This methodology allowed the team to sample progressively larger neuron sub-ensembles to track information growth.
Timeline
1996-2026: Period during which neuroscience assumed a hard neural information ceiling.
September 25, 2026: The research results were published in Science Advances.
The Big Picture
This discovery marks a departure from the 1996 neuroscience hypothesis of neural noise correlation limits. By proving that information capacity scales with neuron counts, it updates fundamental figures previously established by legacy models.
These findings provide engineers with new scaling principles for designing fault-tolerant neuromorphic hardware. This breakthrough could eventually lead to more efficient artificial intelligence architectures that mirror human information processing.
The takeaway
The human brain is far more efficient at processing information than previously assumed by biological models. These new power laws offer a roadmap for developing next-generation AI systems that scale linearly with complexity.
Further reading
Learn more about the latest research in Alzheimer’s and brain health.
More information
Access the full findings in the Science Advances research paper.
Live Poll
Does learning that biological brains scale indefinitely change your confidence in artificial intelligence development?
