Princeton builds brain-cell computer mesh
- Princeton researchers said April 27 they built a 3D bioelectronic device that grows roughly 70,000 living neurons around embedded electrodes for computing. - The system recorded and stimulated activity across multiple planes for more than six months, then learned to distinguish spatial and temporal pulse patterns. - The paper extends biocomputing beyond flat cell cultures toward instrumented 3D networks. (nature.com)
Computers usually run on silicon, but Princeton researchers built a device where living neurons grow through a soft 3D electronic mesh and process signals. (engineering.princeton.edu) (nature.com) The team reported the work in *Nature Electronics* on April 23, 2026, in a paper titled “A three-dimensional micro-instrumented neural network device.” Princeton published its explainer on April 27. (nature.com) (engineering.princeton.edu) The hardware is a scaffold, like a trellis for vines: microscopic metal wires and electrodes coated with a thin epoxy layer that is flexible enough for soft cells to grow around. The researchers cultured about 70,000 neurons into that mesh, creating a three-dimensional network instead of a flat dish. (engineering.princeton.edu) (nature.com) That “inside-out” design differs from earlier biocomputing setups that used two-dimensional neuron cultures in petri dishes or 3D clusters measured mostly from the outside. The new device let the team record action potentials from multiple planes over six months and map how connections changed over time. (engineering.princeton.edu) (nature.com) The researchers also used chronic electrical stimulation to tune those connections, strengthening some links and weakening others. In the paper, they describe turning the network into a reservoir computer, a system that classifies patterns by reading out the changing activity of a dynamic network. (nature.com) In tests, the device distinguished pairs of spatial pulse patterns and pairs of temporal pulse patterns. Princeton said those experiments showed the living network could be programmed to recognize recurring electrical inputs rather than only fire spontaneously. (engineering.princeton.edu) (nature.com) The project was led by Tian-Ming Fu, an assistant professor of electrical and computer engineering, James Sturm, the Stephen R. Forrest Professor of Electrical and Computer Engineering, and postdoctoral researcher Kumar Mritunjay, who was first author. The device was fabricated at the Princeton Materials Institute. (engineering.princeton.edu) (ece.princeton.edu) Princeton framed the work partly around energy use in artificial intelligence. Fu said the human brain uses about one millionth of the power of today’s artificial intelligence systems for similar tasks, though this device is still an early lab platform, not a replacement for conventional chips. (engineering.princeton.edu) What Princeton has now is a long-lived, instrumented 3D neuron culture that can be trained on simple pattern-recognition tasks. The next step, according to the university, is scaling the system to handle more complex computation. (nature.com) (engineering.princeton.edu)
