Argonne turns to insects for neuromorphic computers

The burgeoning field of neuromorphic computing – arguably led by Intel’s “Long” chip – uses virtualized neurons to mimic the behavior of the brain. Normally, the “brain” referred to in that sense would be a human brain, but what if it wasn’t?

“Traditionally, we look at brain-inspired computers and think about us — humans — and how complex our brain is and the vast capabilities it has,” Angel Yanguas-Gil, chief materials scientist for Argonne National Laboratory, explained in a webinar yesterday. . . “What we’ve done at Argonne is take a step back and look not just at people, but at other kinds of inspiration that can help us create systems that are essentially capable of doing this kind of smart learning.”

“And in particular,” he continued, “one of the most promising things we’ve found is insects.”

Towards neuromorphics

The need for neuromorphic computing, Yanguas-Gil explained, arose from core limitations in how current AI algorithms work.

Angel Yanguas-Gil, chief materials scientist for Argonne National Laboratory. Image courtesy of Argonne.

“Once you have a trained system, that system can be deployed, for example on your smartphone or in a chip,” he said. “But once that system is implemented, it’s static. If there are changes or disturbances in the environment that the system needs to respond to, it can only do so if it falls within the data set for which it has been trained. That is very different when it comes to our way of working. We – and humans and animals in general – have a tremendous ability to learn on the fly, react to new information and adapt to changes in the environment.

“What you would like,” he continued, “is to have a system that is able to recognize that something has changed and that, with a few examples, is able to adapt and recover.”

Insects, Yanguas-Gil said, weren’t just an inspiration for compact AI: Insects like bees acted like smart sensors that were able to work in a noisy environment, gather information and — crucially — adapt to that information. . “That,” he said, “is the kind of flexibility that we were very interested in.”

Part of this flexibility, he said, came from its compactness: For example, some insects can take more advantage of their scarce neurons by contextually adjusting the functionality of their brain connections.

The pincer maneuver of Argonne

Yanguas-Gil explained that Argonne’s research on neuromorphic computers in insects was a two-pronged approach. First, the math of it all, starting with an exploration of the state-of-the-art research in insect neuroscience and insect behavior and working to extract the mathematical principles that led to insect performance.

“Once we have those mathematical principles, we can run them in the same way as machine learning or an AI algorithm,” Yanguas-Gil said. Then they take those networks and compare them to benchmarks in machine learning – and in particular the subfield of continuous learning. “It turns out that while they are very small and very agile, they can perform as well in some tasks as the very latest algorithms out there,” he said.

The Loihi neuromorphic chip. Image courtesy of Tim Herman/Intel.

Second, with those algorithms in hand, Yanguas-Gil said, “you want to transfer them to hardware.” He outlined the three hardware approaches Argonne was exploring to take advantage of neuromorphic insect computing, including researching turnkey devices like FPGAs with collaborators like the University of San Antonio and working with state-of-the-art neuromorphic chips like Intel’s Loihi. .

Designing a stronger shell

“The last thing, though, is that we can take these ideas and figure out … how we can change the way we design chips with new materials,” he said. Yanguas-Gil outlined how the researchers took advantage of Argonne’s “extremely strong program” in atomic layer deposition — a thin-film technique used in semiconductor manufacturing — to conduct advanced co-design research for neuromorphic computing.

“We use [the neuromorphic application] as a goal not only to design the architecture at the same time… [to identify] what type of new materials we need, or how best to integrate the materials we already have into this architecture to optimize the ability to learn in real time.”

Some of this research, Yanguas-Gil said, focused on making those platforms more resilient to extreme environments.

“We’ve found that combining these new materials with other non-silicon platforms — such as silicon carbide — can help you maximize the amount of computation you can do while minimizing the number of components needed,” he said. important when you go to temperatures of 300 to 400 degrees Celsius – and even to environments with a lot of radiation.” A material Argonne developed years ago made it possible to tune resistance over “many orders of magnitude” and withstand temperatures of up to 500 degrees Celsius.

Yanguas-Gil sees several uses for this type of ultra-compact, adaptable, highly resilient neuromorphic hardware, mentioning self-driving cars (“If you want the vehicle to respond to a change it was not trained for without catastrophically failing, that’s one application” ), as well as the brain-computer interfaces used to control prosthetics.

Leave a Comment