They’re small, efficient and capable of basic reasoning, and researchers want artificial intelligence tools to do the same.
The Pentagon’s research wing is trying to reduce the amount of computing power and hardware needed to run advanced artificial intelligence tools, and it’s turning to insects for inspiration.
The Defense Advanced Research Projects Agency on Friday began soliciting ideas on how to build computing systems as small and efficient as the brains of “very small flying insects.” The Microscale Biomimetic Robust Artificial Intelligence Networks program, or MicroBRAIN, could ultimately result in artificial intelligence systems that can be trained on less data and operated with less energy, according to the agency.
Analyzing insects’ brains, which allow them to navigate the world with minimal information, could also help researchers understand how to build AI systems capable of basic common sense reasoning.
“Nature has forced on these small insects drastic miniaturization and energy efficiency, some having only a few hundred neurons in a compact form-factor, while maintaining basic functionality,” officials wrote in the solicitation. “Understanding highly-integrated sensory and nervous systems in miniature insects and developing prototype computational models ... could be mapped onto suitable hardware in order to emulate their impressive function.”
The project comes as part of DARPA’s broader Artificial Intelligence Exploration program, which provides rapid, small-scale investments in efforts to create so-called “third wave” artificial intelligence. Unlike current AI tools, which must be explicitly trained to perform narrow functions, third wave systems would be instilled with the generalized reasoning and contextual awareness so as to better understand the world it describes.
Researchers often describe such tech as mimicking the human brain, but because our brains contain between 60 to 70 billion interconnected neurons, rebuilding it as software is next to impossible. By contrast, some insect brains contain less than 1,000 neurons, making them much easier map.
“Studying miniaturized insects may reveal fundamental innovations in architecture and computation analogous to their simultaneous simplicity, efficiency and complex functionality,” researchers wrote. “Compared with larger-brained organisms, these insects may hold the keys to strategies for AI that combine energy- , time- and space-efficient operation.”
DARPA will provide up to $1 million in funding to groups to create a physical model of insects’ neural systems, analyze how insects’ brains develop over time and design hardware platforms that mimic the neural structure of those brains.
The program will be divided into two phases, with participants first developing a proof of concept before building a prototype. Projects are expected to run roughly 18 months.
Responses to the solicitation are due Feb. 4, and the program is expected to launch April 3.