Most researchers developing artificial intelligence are focused on creating machines that simulate the human brain. To accomplish this, they are using silicon-based chips for processing power.
The biggest obstacle the team must overcome is keeping the neurons alive. In the lab, neurons have been kept alive for up to two years.
Emulating human brain capability requires an incredible amount of computing power in a relatively small space. Chip manufacturers are working to reduce the size of silicon-based chips, but many believe we’re approaching the physical limitations to how small we can make a transistor.
Oshiorenoya Agabi, a Nigerian neuroscientist, had a bold idea—rather than mimic the brain, why not use actual brain cells (neurons) to manufacture a chip. “We believe biology is the most advanced technology platform on the planet. Instead of copying the neuron, why don’t we take the neuron and put it in a chip?” Agabi said in the very first session of TEDGlobal, held in Tanzania in 2007.
Putting together a team of scientists, including geneticists, physicists, other bio-engineers, and molecular biologists, Agabi applied his Ph.D. in bioengineering to form a startup company, Koniku. The team reverse-engineered biology to develop the “Koniku Kore,” a modem-sized computer consisting of a neuron and silicon processing core, sensors that recognize smells, and an electrode. The electrode reads and writes information inside the neurons. The prototype of this biological chip stores 64 neurons. Ten years after Agabi’s first announcement, he unveiled the prototype at the August 2017 TEDGlobal conference, fittingly held once again in Tanzania.
To demonstrate the effectiveness of the device, the team chose a problem difficult for silicon-based technology to solve—smelling a gas. The Koniku Kore “breathes” in air and “smells” it to determine whether the air contains a particular volatile substance.
Significance of the Neuron-based Chip
The merging of biological and digital systems is unprecedented and has tremendous possibilities. Think of situations currently using trained sniffer dogs. Airport security could quickly scan for explosive substances, eliminating the need for long lines at the security checkpoint. Law enforcement could quickly scan crowds, buildings, or vehicles for explosives and drugs. In the medical field, the device (or a smell-enabled robot) could detect diseases such as cancer. Drones could hover over pipelines and natural gas processing plants to detect methane leaks.
In addition to the security industry, the technology has garnered much interest from a diverse group of large corporations like AstraZeneca, Boeing BA, and Cisco. Agabi was able to raise $1 million startup money for his Silicon-Valley based company and currently claims he has made $10 million in profits. He expects $30 million in profits by 2018 through deals he has made with the security and other industries and projects the actual size of the market for his technology to be $145 billion.
Agabi’s ultimate goal is to “build a truly cognitive system” using artificial neurons (a combination of a neurological and biological material and silicon) within the next five to seven years. To create such a system, Agabi believes his team will be able to expand the number neurons in their device from 64 to millions of neurons per chip.
The biggest obstacle the team must overcome is keeping the neurons alive. In the lab, neurons have been kept alive for up to two years. However, in the device, the neurons last for at most two months.
Agabi has no doubt his team will overcome the neuron longevity problem. He knows that by achieving the first major step of his bold idea, he’s made a new leap for AI: computer chips that can smell.