University spin-off startup Applied Brain Research finds success commercializing deep science

In 2011, entrepreneur Peter Suma knocked on Chris Eliasmith’s door at the University of Waterloo, equipped with a fascination for neuroscience, business experience and a hunch. Eliasmith is the Canada Research Chair in Theoretical Neuroscience at the University of Waterloo, and the creator of Spaun, the world’s largest functional model of the human brain. Suma, who had advised and funded some early artificial intelligence companies, saw commercial potential in Eliasmith’s research and reached out.

Now, eight years later, Eliasmith and Suma are co-CEOs of Applied Brain Research (ABR), which is using neuromorphic, or brain-inspired, computing to power AI applications in the real world. “Neurons, and by extension circuits, and by extension brains, live embedded in time and space. And our current AI does not,” said Suma in an interview with RE$EARCH MONEY. “Fundamentally, neuromorphics is about getting closer to the secrets of intelligence, as shown us by the brain, by adding time to the understanding of artificial neural networks. But that requires a completely new set of software and hardware.”

ABR has both. Its products include Nengo, a set of software tools that dramatically improve the efficiency of AI computing, speeding up computation and vastly reducing energy consumption. Nengo works with traditional CPUs or GPUs or with neuromorphic chips, such as Intel’s Loihi. According to the company, Nengo currently has more than 3,000 users, including DARPA, the US Airforce and others.

ABR is also commercializing a neuromorphic chip, which boasts a new architecture designed to emulate the brain. It's one of only a handful of companies in the world doing so. Having a chip in its arsenal will allow ABR to bring more efficiency to edge computing devices, including cellphones, drones and driverless cars, which depend on data moving quickly in a network and on adequate power.

As an entrepreneur, Suma saw commercial potential where most others saw pure scientific exploration. When he reached out to Eliasmith, computers were no longer following Moore’s Law, and the limits of machine learning, such as its inability to generalize, were becoming apparent. Suma felt new, brain-inspired tools were needed. “I said to Chris, I don't know much, but I know AI can't be about single networks. True intelligence has to come from the dynamic integration of multiple networks we see in the brain,” he said. “We sort of fake it with current AI and we get by with speech and voice recognition, but the holes in it show immediately. It doesn't really understand what's going on.”

This year, ABR was named to the CIX Top 20 as one of Canada’s most innovative startups and showcased at VentureBeat’s Transform AI conference in San Francisco. Just this month, it was named European automotive startup of the year, in recognition of its work to build more efficient AI systems for automakers. But ABR’s recent accolades belie the long and sometimes winding road to commercializing the kind of deep science—long-range research with unpredictable outcomes or applications—that Eliasmith does.

Still, Suma is confident the company, which has 16 employees and is growing, will succeed. Currently, ABR has contracts with a leading cellphone manufacturer, automaker and energy company to demonstrate the commercial benefits of its software and hardware.

“We'll take neuromorphics and fully commercialize it, and we hope to build a massive company, powering edge chips all over the world in devices running AI faster, cheaper, better,” he said. "Our second act, now that we have that, is getting to labour-saving autonomy, safer devices, smarter devices."

R$