Brad Theilman and Felix Wang, researchers at Sandia National Laboratories, are working in earnest on their beautiful aquarium. They are presently unpacking a new neuromorphic computing core, Loihi 2. Intel created this challenging hardware to transform the way we approach solving complex mathematical problems. It is at its most promising for the solution of partial differential equations. Their work sheds light on the fact that neuromorphic computing could provide even more capabilities than what was understood by the scientific community to date.
The Sandia team has already demonstrated that Loihi 2 can successfully solve these key equations. This expanded capability is a big deal for a lot of fields, including engineering and physics. In addition, neuromorphic hardware has the potential to model some very complex problems. This technology is the most closely aligned to the brain’s organic processing techniques, executing complex calculations at high efficiency with much less power.
Advancements in Neuromorphic Computing
Neuromorphic computing mimics the neural structures and functioning of the human brain, allowing it to process information in ways traditional computers cannot. The cognitive wealth of the human brain is amazing operating on just 10 watts of power much less than traditional computing architectures.
Theilman and Wang’s recent study demonstrates that neuromorphic systems such as Loihi 2 are highly superior in non-conventional hardware environments. This optimization is what makes them unique from general-purpose computational approaches developed for legacy architectures. That’s why the team has been successful in translating the finite element method (FEM) into a model that behaves like the motor cortex. They tested this model out on Loihi 2.
The implications of this research go further than just performance. As USF’s Bradley Theilman noted, “It’s a really hard problem.” He elaborated that the brain directs muscles in quick succession to allow an athlete to touch the ball. This analogy serves to introduce the potential for neuromorphic systems to provide functionality that parallels personal human cognition. In turn, they can make them more effective for some key work.
Collaboration and Insights from Experts
The story behind this incredible work at Sandia National Laboratories is collaborative. It connects with a bright and passionate community of computational researchers dedicated to improving and expanding computational methods. Craig Fritz is one of those working on this creative project. They recently shared their findings in a paper published in Nature Machine Intelligence. In it, they discuss the exciting potential for using neuromorphic hardware to solve complex differential equations.
Steve Furber, at the University of Manchester, has long added trenchant, often positive critiques that are always welcome as a valuable corrective to naiveté or over-hype. His team used ARM-based SpiNNaker hardware to model the diffusion of heat. This artistry highlights the innovative use of neuromorphic computing to study complex and beautiful physical phenomena.
James B. (Brad) Aimone has been central to this research. From his point of view, he just underscores the importance of approaching mathematical challenges from the viewpoint of neuromorphic computing. He claims, “There’s no given premise that you couldn’t do anything with neuromorphic computing.” These words capture the unlimited promise that allows scientists to dream big with this technology.
Looking Towards the Future
As neuromorphic computing becomes more advanced, it increasingly poses a challenge to the conventional paradigm of computational approaches. Arturo R. García The Sandia team’s work underscores the incredible progress that has been made in artificial intelligence. We have a long road ahead of us in terms of understanding how to unlock the complete promise of neuromorphic systems.
Aimone is careful to emphasize this potential when he states, “We’ve only begun to scratch the surface of advances in AI. Look, the world doesn’t need more power plants. This closing remark harks back on the importance of innovation, exploration and pushing boundaries within computing. Here’s how they can get us closer to a more efficient overall energy usage. They can better improve our creative juices to solve big problems.”
