© 2024 SDPB Radio
Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations

A new AI model can predict some brain signals in fruit flies

JUANA SUMMERS, BYLINE: Scientists have created an artificial intelligence system that thinks like a fruit fly, and that's a good thing. NPR's Jon Hamilton reports that the approach could help make some technologies more efficient and provide new insights into how the brain processes information.

JON HAMILTON, BYLINE: The latest artificial intelligence systems can drive a car and even write fiction, but they rely on computers that contain billions of transistors and consume lots of electricity. Jakob Macke at the University of Tubingen says a fruit fly gets by with a lot less processing power.

JAKOB MACKE: And yet, with a small and energy-efficient brain, it's able to do so many computations. It's able to fly. It's able to walk. It's able to detect predators. It's able to mate. It's able to survive using just 100,000 neurons.

HAMILTON: Macke and a team of scientists wanted to know how a fruit fly does so much with so little, so they reviewed recent studies that have identified all the neurons in an insect's brain, as well as all the connections between those neurons. These pathways are known collectively as a connectome, and Macke says it gave the team a sort of road map.

MACKE: It tells you who's connected to who and how big is each road. That, in principle, tells you how information could flow, how you could travel from A to B, but it doesn't tell you which of these travels are actually taken by the system.

HAMILTON: And Macke says there's no good way to see all the information traveling through a living brain, even one as simple as a fruit fly's.

MACKE: Brains are so complex that I think the only way we will ever be able to understand them is by building accurate models.

HAMILTON: In other words, by simulating a brain or part of a brain on a computer. So that's what the team did. Srini Turaga is from the Howard Hughes Medical Institute's Janelia Research Campus in Virginia. He says they focused on brain circuits that allow a fruit fly to detect motion.

SRINI TURAGA: Our goal was not to build the world's best motion detector, but to find the one that does it the way the fly does.

HAMILTON: The team created a computer model based on the connectome of a fruit fly's visual system. Then they had an artificial intelligence system watch movies known to trigger a motion response in the brain of a living fruit fly. Eventually, the machine got really good at detecting movement, and tests showed that it was doing this the same way an actual fruit fly does. Turaga says that finding has big implications for brain scientists.

TURAGA: If you know the connectome and you know something about what the circuit is supposed to do, what it's supposed to compute, we can use AI to then predict neural activity for every single neuron.

HAMILTON: Because the model behaves just like a real fly brain. The research appears in the journal Nature, and Turaga says it should accelerate efforts to understand how brains process information in all areas, not just the visual system.

TURAGA: Now we can start with a guess for how the fly brain might work before anyone has to make an experimental measurement.

HAMILTON: Ben Cowley is a computational neuroscientist at Cold Spring Harbor who was not involved in the research, but he's already a convert.

BEN COWLEY: I'm using this model in my own work because it beautifully makes some predictions about this is what we should expect to see if one were to record from these neurons.

HAMILTON: And it shows how living brains are able to make so many calculations without consuming much energy. Cowley says that's one of the big challenges in the field of artificial intelligence.

COWLEY: When we think about AI, right now, I think the leading charge is to make these systems more power efficient.

HAMILTON: By doing a lot with a little, much like a fruit fly. Jon Hamilton, NPR News.

(SOUNDBITE OF MUSIC) Transcript provided by NPR, Copyright NPR.

NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.

Jon Hamilton is a correspondent for NPR's Science Desk. Currently he focuses on neuroscience and health risks.