Scientists have created a single artificial neuron that can mimic activity from different regions of the brain, marking a step toward machines that sense and respond to the world like humans.
The transneuron switches between roles linked to vision, planning, and movement. It also processes information through electrical pulses, bringing hardware closer to biological computation.
The research team, led by Loughborough University with collaborators from the Salk Institute and the University of Southern California, built the device to imitate real brain behaviour.
Traditional artificial neurons perform one narrow task. The new transneuron shifts between roles by adjusting its electrical settings.
“Is the human brain a mysterious device beyond our reach or could we one day recreate it with electronics – and perhaps even build something more powerful?” asks Professor Sergey Saveliev of Loughborough University.
He says the work shows one artificial neuron can mimic visual, motor, and pre-motor behaviour. He adds that this could support future chips that perform complex tasks with minimal hardware.
“Ultimately, this paves the way for more human-like robots.”
The team tested the transneuron by feeding electrical signals into it and comparing its pulses with recordings from macaque neurons.
The device reproduced pulse patterns from three brain regions with up to 100% accuracy. These patterns ranged from steady firing to erratic bursts.
Professor Alexander Balanov of Loughborough says small electrical changes let the unit act like different neurons. “We also know that our artificial neurons respond well to changes in the environment, like pressure and temperature,” he says.
He notes that this could lead to artificial sensory systems and faster, energy-efficient computers.
Computes like the brain
The researchers also showed that the transneuron processes information rather than just imitating activity.
When they changed the input, the device altered its firing rate, similar to biological neurons. When they fed it two signals at once, it responded differently depending on their timing.
This ability normally requires multiple artificial neurons.
The device’s flexibility comes from a memristor, a nanoscale component that physically shifts as electricity flows through it. Silver atoms inside form and break tiny bridges, producing electrical pulses.
Temperature, voltage, and resistance changes alter the pulse behaviour, allowing neural roles to shift without software.
Dr Sergei Gepshtein of the Salk Institute says conventional computers process information in rigid steps, unlike the brain.
“Our transneuron moves us closer to creating hardware that doesn’t just simulate brain-like activity in software but actually works in a brain-like way.”
Toward robotic nervous systems
The next step is building networks of transneurons into a “cortex on a chip.” The team believes such networks could help robots sense and adapt in real time.
“This work marks a small but significant step toward building robots with artificial nervous systems,” says Professor Joshua Yang of USC. He says such systems could support efficient, lifelong learning and cut energy use.
Dr Pavel Borisov of Loughborough adds that the devices may one day interface with the human nervous system or help scientists study consciousness.