summary: Researchers have developed an AI model of the fruit fly brain to understand how vision guides behavior. By genetically silencing specific visual neurons and observing behavioral changes, they trained an AI to accurately predict neural activity and behavior.
Their discovery revealed that a combination of multiple neurons, rather than a single type, processes visual data in a complex “population code.” This breakthrough paves the way for future research into the human visual system and associated disorders.
Key Facts:
- Scientists at CSHL have created an AI model of the fruit fly brain to study visually guided behavior.
- The AI predicts neural activity by analyzing changes in behavior after silencing specific visual neurons.
- The study revealed a complex “population code” in which multiple neurons combine to process visual data.
sauce: CSHL
It is said that the eyes are the windows to the soul. Windows work both ways. Eyes are also windows to the world. What we see and how we see determines how we move through the world. In other words, our vision helps guide our behavior, including social behavior.
Now, young scientists at Cold Spring Harbor Laboratory (CSHL) have discovered an important clue about how this works. He accomplished this by building a special AI model of the brain of a common fruit fly.
CSHL assistant professor Benjamin Cawley and his team refined the AI model through a proprietary technique called “knockout training.” First, we recorded the courtship behavior of male fruit flies chasing and singing after females.
They then genetically suppressed a specific type of visual neuron in male flies and trained an AI to detect behavioral changes. By repeating this process with different types of visual neurons, they were able to get the AI to accurately predict how real fruit flies would behave in response to the female's vision.
“We can actually computationally predict neural activity and see how specific neurons contribute to behavior,” Cowley says. “This is something we couldn't do before.”
Using new AI, Cowley's team discovered that the fruit fly brain uses a “population code” to process visual data. Rather than one neuron type linking each visual feature to her one action, as previously assumed, many combinations of neurons were needed to sculpt the behavior.
These diagrams of neural pathways look like an incredibly complex subway map and would take years to decipher. Still, it gets us where we need to go. This allows Cowley's AI to predict how real fruit flies will behave when presented with visual stimuli.
Does this mean that AI will be able to predict human behavior in the future? Not so soon. A fruit fly brain contains about 100,000 neurons. A human brain has about 100 billion neurons.
“This is what a fruit fly looks like. Now imagine what our visual system looks like,” Cowley says, referring to a subway map.
Still, Cowley hopes that his AI model can one day help decipher the math underlying the human visual system.
“This is going to take decades of work. But if we can figure this out, we have a leg up,” Cawley says. “By learning [fly] “Computation will help us build better artificial vision systems and, more importantly, provide a better understanding of disorders in the visual system.”
How much better? You'll have to see it to believe it.
About this AI and neuroscience research news
author: Sarah Jarnieri
sauce: CSHL
contact:Sara Jarnieri – CSHL
image: Image courtesy of Neuroscience News
Original research: Open access.
“Mapping model units onto visual neurons reveals population coding of social behavior” by Benjamin Cowley et al. Nature
abstract
Mapping model units onto visual neurons reveals population codes for social behavior
A wide variety of animal behaviors arise from the interplay between sensory processing and motor control. To understand these sensorimotor transformations, it is useful to build models that predict not only neural responses to sensory inputs, but also how individual neurons causally contribute to behavior.
Here, we present a new modeling approach to identify one-to-one mappings between internal units of deep neural networks and real neurons by predicting behavioral changes resulting from systematic variations in more than a dozen neuronal cell types. indicate.
A key element we introduce is “knockout training,” in which we perturb the network during training to match the perturbations of real neurons during behavioral experiments. Applying this approach, Drosophila melanogaster A male that engages in complex social behavior guided by vision.
Visual projection neurons at the interface between the optic lobe and the central brain form a series of separate channels, and previous studies have shown that each channel encodes a specific visual feature to drive a specific behavior.
Our model reaches different conclusions. A combination of visual projection neurons, including those involved in non-social behavior, drives male-female interactions and forms a rich population code for behavior.
Overall, our framework integrates the behavioral effects evoked from different neural perturbations into a single unified model, providing a map from stimulus to neuron type to behavior, and providing a map that will help the brain in the future. Allows you to incorporate wiring diagrams into your model.
