Brain-like computers explained

A brain-like computer is a computer whose design borrows ideas from the brain. The field is also called neuromorphic computing. People in the field also use the words brain-inspired computing.

Most computers in use today follow a layout often called the von Neumann architecture. Memory sits in one place and processing sits in another, and data moves back and forth between the two.

The brain works in a different way. Memory and processing sit close together. Many parts work at the same time. Nerve cells pass short signals to each other, often called spikes.

A brain-like chip tries to follow some of these ideas in silicon. Many such chips run spiking neural networks: networks of artificial neurons that pass short signals to each other. Some chips are digital, some are analog, and each takes its own path.

The projects below are public research work by other organisations. Each description keeps to what the linked page says. Each link opens the makers' own page.

TrueNorth is a brain-inspired chip from IBM Research. IBM describes it as a fully digital chip for spiking neurons, with an event-driven design that departs from the von Neumann architecture. IBM Research: TrueNorth

NorthPole is a later chip from IBM Research, presented in 2023 as an extension of TrueNorth. IBM describes it as a digital chip for neural inference, the step where a trained network is used. IBM says all of its memory is on the chip itself, and that it blurs the boundary between compute and memory. IBM Research: NorthPole

Loihi is a line of neuromorphic research chips from Intel Labs. Loihi 2 is the second generation. Intel describes it as using event-based spiking neural networks, with memory and computing integrated. Intel provides an open-source software framework for it, called Lava. Intel Labs: neuromorphic computing

SpiNNaker, short for spiking neural network architecture, is a neuromorphic computing platform at the University of Manchester. The university describes it as completely digital and built from a very large number of small ARM processors. It is used to model spiking neural networks, for research in brain science and in new computer architectures. University of Manchester: SpiNNaker

BrainScaleS is an analog neuromorphic system built at Heidelberg University. In it, spiking neural networks are physically emulated as analog circuits rather than computed as numbers. Researchers can use it as a cloud service through the EBRAINS research infrastructure. Heidelberg University: BrainScaleS

All of them take ideas from the brain, and they differ in how. Some are digital and some are analog. Some serve brain science, some run trained networks, and some do both. This page does not rank them or compare them. The linked pages describe each one in its makers' own words.

Hyperloop is designing a brain-like computer, a computer with a human mind. It is a design, not a product. Design complete on paper; nothing built yet. Hyperloop is not connected to any of the projects above. More about the work is on the about page.