Current Issue
This Month's Print Issue

Follow Fast Company

We’ll come to you.

2 minute read

Change Generation

Combining Human Neurons With Machines Makes A Truly Powerful Computer

Adding a little human biology to a computer can improve artificial intelligence.

Photos: GrayMark via Shutterstock

The silicon computer chip changed the world. Room-size computers shrank, and the smartphone in your pocket is now more powerful than all of NASA's computers, combined, in 1969. Now the chip might change again by adding a new material: human neurons.

"Silicon doesn't really scale," says Oshiorenoya Agabi, CEO of Koniku, a startup pioneering neuron-powered computation. "To simulate 204 neurons you need a supercomputer. Just to simulate 204! That gives you an impression of how powerful and how sophisticated neurons are. Now imagine if you could take this power and sophistication and put it in a device."

A brain has around 100 billion neurons and runs hyper-efficiently. "Our brains consume 10 watts of energy," says Agabi. "Our brains beat a supercomputer with 10 watts of energy. Imagine how revolutionary that could be, how much it could change computing, our interaction with computers, how sustainable it is."

While computers beat the human brain at things like straightforward calculations, there are other things that brains do much more efficiently—like noticing changes in a room. While traditional computing might take frame after frame of images, trying to calculate the difference between them, your brain watches for changes in the scene with little effort.

"We take the same principles and we put it on our chip," he says. "You can imagine, on our chip, it's like a single neuron has its own search engine, its own memory. It doesn't take a traditional microprocessor."

Adding a little human biology to a computer can improve artificial intelligence. "The kind of computation that we're looking for in the future, a neuron does this better," Agabi says.

After over a decade of work in academia—where others are also working on neuron-embedded chips—Agabi decided to try to bring it to market. His team, now part of the Indie Bio accelerator in San Francisco, has a working prototype of a chip that can amplify signals from neurons and do calculations.

The first application will be a "lab on a chip," a device that can help drug developers mimic how human organs work on a tiny scale. The addition of neurons makes it more precise.

"Our system is helping researchers understand how Alzheimer’s affects brain signaling with clarity," says Agabi. "They can develop drugs which can halt these effects. But we are already going further proving that we are a platform technology."

He thinks the entire field of computing could ultimately move to this type of chip. "We are at the stage where this field is about to blow up," he says.

"Remember the transition from slates to paper, paper to mechanical systems, mechanical systems to the vacuum tube, vacuum tubes to silicon. Now is the transition from silicon to biological neurons."

loading