ARLINGTON, Va. – U.S. military researchers are ready to approach industry to develop power-efficient computers that use neural cells or lab-grown living tissue, rather than silicon transistors, for artificial intelligence (AI) processing at the leading edge of the battlefield.
Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., issued a future program announcement (DARPA-SN-26-48) on Tuesday for the Organoid Cytomorphic Intelligence Resulting from Convergent Understanding and Information Transfer (O-CIRCUIT) project.
O-CIRCUIT will seek to use super-efficient biological processing units that use living biological material — typically neural cells or organoid tissue — as the core information-processing substrate, rather than silicon transistors for AI processing in austere environments on the battlefield.
Modern computation involving AI training and inference has significant energy requirements, DARPA researchers explain. O-CIRCUIT seeks to develop AI at the battlefield edge that uses minimal power draw measured in milliwatt-hours per day for sensing physical stimuli and outputting instructions like directing drone navigation.
Modeling fruit fly brains
Although this may sound rare and complicated, biology does this every day, DARPA researchers explain, all the way down to the level of fruit fly brains, which have about 140,000 neurons and consume less than six milliwatt-hours per day.
O-CIRCUIT will demonstrate that it is possible to build biological processing units comparable size, weight, and low-power consumption to naturally occurring neural structures to provide calibrated synthetic intelligence.
To accomplish these goals, the O-CIRCUIT program will develop a system using organoid and synthetic biological intelligence where cell compositions, such as neural, glial, and immune cells, can perform complex training and inference tasks.
Trainable biological processing
The challenge is to develop trainable biological processing units that provide sophisticated inference and drone guidance via neural olfactory sensing.
The 42-month O-CIRCUIT program has an 18-month first phase, a one-year second phase, and a one-year third phase, with two development approaches: architecture and action.
Architecture focuses on building a biological processing unit with enhanced learning capabilities by exploring biological processing unit architecture complexity, circuitry, health, and cell composition. This he biological processing unit should be able to play a video game close to human level proficiency, retain this proficiency, and consume energy at rates comparable to natural neural systems.
Detecting and following odors
Action will focus on building and testing a biological olfactory sensor system integrated into a biological processing unit for drone navigation systems. It should be able to detect tens of different smells and navigate an unmanned drone proficiently towards the source of the odors in limited time.
This announcement is neither a formal solicitation nor a request for information; it signals the intent to launch a new research without yet releasing full funding opportunities.
Email questions or concerns, or requests for future O-CIRCUIT program updates, to DARPA at [email protected]. More information is online at https://sam.gov/workspace/contract/opp/f8fd222be5f046b887fb4759b2d7d123/view.
