Scientists have created an AI version of a monkey brain that recognizes images without requiring the massive computing power of existing AI systems.
JUANA SUMMERS, HOST:
A human brain consumes less power than a light bulb. An artificial intelligence system uses vastly more energy to do the same tasks. NPR’s Jon Hamilton reports on new research that hints at how living brains do more with less.
JON HAMILTON, BYLINE: The brain’s visual system takes in bits of light and transforms them into something we recognize, like grandma or the Grand Canyon. Ben Cowley of Cold Spring Harbor Laboratory has been studying how this happens.
BEN COWLEY: Getting at the question, for example, how do you recognize a cat? How do you recognize a dog, etc.?
HAMILTON: There’s no good way to watch a human brain do this, so Cowley has been looking at artificial intelligence systems. But he says there’s a problem.
COWLEY: We’re very impoverished in our understanding of how these AI systems work, much like our own brain.
HAMILTON: Cowley’s team created an AI model they could understand. It simulates the neurons in just one part of the visual system.
COWLEY: They encode colors and textures and curves and very complicated proto-objects. So you wouldn’t say, oh, this cares about a dog, but it might care about a weird nose and an eyeball or something like this.
HAMILTON: Existing AI systems can do this using large models that consider every possibility. Cowley’s team was after something more efficient.
COWLEY: So we want to take these big, clunky models and try to compress it down into a much smaller, compact form.
HAMILTON: They started with a model trained on data from macaque monkeys. It worked, but it included about 60 million variables. Next, Cowley’s team pruned the model and applied statistical techniques like those used to compress digital photos. The result, a version with only about 10,000 variables.
COWLEY: And that is incredibly small. This is something we could send in a tweet or an email.
HAMILTON: Yet, it still does most of what monkey neurons do. And because the model is so small and simple, the team was able to get a glimpse of what its artificial neurons were doing. For example, Cowley says cells called V4 neurons were responding to shapes with strong edges and lots of curves.
COWLEY: The best comparison I could find is when you go into the supermarket and you see the arranged fruit, your V4 neurons love that. They love arranged fruit. They love all the curves of the apples, oranges and things like this.
HAMILTON: Cowley says the AI model also included a group of V4 neurons that seem to be looking only for small dots in an image.
COWLEY: And this was quite interesting to us because, especially for primates, we are very drawn to eyes. You can see a big scene, and if there’s a little human there with little eyes, you are directly going to that face, and you’re looking at their eyes.
HAMILTON: Cowley says the results, which appear in the journal Nature, suggest how human brains may have found efficient ways to make sense of what they see. He says the findings also have implications for artificial intelligence.
COWLEY: If our brains have less-complex models and yet can do much more than these AI systems, that tells us something about our AI systems.
HAMILTON: Namely, that they could probably be smaller and simpler, yet do a better job interpreting what they see. Self-driving cars, for example, could run on smaller computers and still not confuse a plastic bag with a pedestrian. Mitya Chklovskii of New York University and the Simons Foundation Flat Iron Institute says the new study shows one way AI can become more like its natural counterpart. But he says a living brain still outperforms an artificial one in some tasks.
MITYA CHKLOVSKII: You can recognize a face of your friend, regardless of the distance to the face, the orientation of the face, maybe a suntan, maybe a new haircut. You still get the face of your friend.
HAMILTON: Chklovskii says that’s because the neurons in biological brains aren’t limited to snapshot images. Instead, they take in videos that show the same face at different times in different places. Jon Hamilton, NPR News.
(SOUNDBITE OF MUSIC)
Copyright © 2026 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.
Accuracy and availability of NPR transcripts may vary. Transcript text may be revised to correct errors or match updates to audio. Audio on npr.org may be edited after its original broadcast or publication. The authoritative record of NPR’s programming is the audio record.
