Emulating the brain’s behaviour requires us to develop new kinds of computer chips that more faithfully mimic the way neurons work and are arranged.2 One particular challenge is to replicate how brain cells communicate via spikes of electrical activity. Although there are similarities with AI, the details mean that there are limited opportunities for borrowing insights from that field. Current models of biological neurons and learning rules are also too simplistic, brushing over physiological details known to play an important role in the brain.
Many efforts at developing neuromorphic computing are underway. Researchers have, for instance, developed “Spiking Neural Networks” (SNNs), in which neurons communicate with each other with spiking, discrete electrical signals. The SpiNNAker project at the University of Manchester arranges general purpose silicon chips in a novel, highly parallel architecture3. IBM’s TrueNorth chip and Intel’s Loihi chip have both been custom-designed to run SNNs45. Memristors, simple transistor-like components that can retain information how much charge has flowed through them, also hold considerable promise.6 However, we still don't know what brain (and, possibly, embodiment) components we need to mimic, and in how much detail.