While the eventual goal of artificial intelligence is to develop computers that can exhibit intelligence equal or greater than that of the human brain (quite possibly the latter), there are a few benchmarks we’ll need to achieve first. One of these is to be able to replicate the less complex brains of other creatures inside a computer. Achieving this means first being able to understand exactly how the brains of said creatures work on a cellular level — including how neural circuitry triggers specific behavior.
That’s what researchers at the Howard Hughes Medical Institute are helping contribute to with a new project, with the stated goal of creating a brain-wide “atlas” of fruit fly behavior.
“While the fruit fly brain is much simpler than the human brain, it still contains about 100,000 interconnected neurons that combine in complex ways to produce behavior,” project lead Dr. Kristin Branson told Digital Trends. “Our study made use of a few very new technologies. We used recently developed genetic techniques to activate sparse populations of neurons. We also made use of new machine vision and learning algorithms for analyzing the data we collected.”
The results the team has achieved are the most in-depth neural map of fruit fly behavior yet. The project involved studying 2,204 populations of flies to find the neurons involved with 14 different behaviors, ranging from wing-flicking to attempted copulation. Were humans to have had to do the project’s “behavior labeling” work instead of machine learning algorithms, the task would apparently have taken 3,800 years. Even in the field of long-term research projects, that’s considered excessive!
“We have mapped the regions of the fly brain that are involved in a variety of locomotion and social behaviors,” Branson continued. “We have done this at the resolution of individual neurons across the entire brain. We hope that the behavior-anatomy maps resulting from our study will enable other biologists to understand the precise computations that the brain performs to produce these behaviors.”
The researcher’s work isn’t just limited to fruit flies, however. “As we start to decipher the ways that the fruit fly brain implements behavior, we hope to find common principles and motifs of neural computation that generalize beyond fruit flies,” she noted. “Understanding circuit computations does involve simulating our models of those circuits in the computer to prove to ourselves that we understand the system, and may enable us to understand why that particular implementation of behavior is advantageous.”
While currently artificial neural networks are only an approximation of how the brain works, hopefully research like what has been conducted by the Howard Hughes Medical Institute will help brain-inspired computation advance to the next level. A paper on the research was published in the journal Cell.
