Learning to compute with the language of neurons

The human brain has a remarkable capacity to receive, process, and store information with very little power consumption. Despite having 86 billion neurons forming over 100 trillion connections, the brain consumes only about 20 Watts of power—compare this to the 7 billion transistors in an iPhone 10 and the 60 Watts needed to power a lightbulb and you can begin to see the immense and efficient computational capacity of the brain. Modern microelectronics technology has given us the tools to observe the electrical signaling behavior of networks of neurons grown in the lab to better understand how they form connections with each other and propagate information through the network. With the insights we obtain by analyzing these signals, we are working to emulate targeted behaviors in models and physical substrates toward a new computing paradigm that will enable vast and scalable computing capabilities with a lower energy cost.