Northwestern University engineers have crossed a major threshold: they've built artificial neurons that don't just mimic biological tissue but can directly communicate with it. The breakthrough opens pathways to implantable brain-machine interfaces and, more broadly, a radical rethinking of how computers could work.
The artificial neurons, constructed from printed electronic inks containing molybdenum disulfide and graphene on flexible polymer surfaces, produce electrical firing patterns that match the timing and shape of real neural signals. When researchers applied these signals to slices of mouse cerebellum, the artificial neurons successfully activated the biological tissue, triggering neural responses as if they were genuine brain signals.
The research, led by Mark C. Hersam at Northwestern's McCormick School of Engineering, challenges decades of conventional thinking about how neurons should be engineered. Previous attempts either produced spikes too slow or too fast to interact properly with living systems. This approach lands squarely in the biological window.
"You can see the living neurons respond to our artificial neuron," Hersam said. "We've demonstrated signals that are not only the right timescale but also the right spike shape to interact directly with living neurons."
The engineering trick lies in how the team handled a material scientists usually discard. Instead of completely removing the polymer binder from the printed inks, they partially decomposed it. When current flows through the device, further decomposition creates a narrow conductive filament that produces a sudden electrical response mimicking a neuron's action potential. The result is a single component that can generate multiple firing patterns: isolated spikes, continuous bursts, or rapid sequences.
That complexity matters enormously for computing efficiency. Traditional artificial intelligence relies on billions of identical transistors packed onto rigid silicon chips, each behaving identically in a fixed arrangement. The human brain operates on an entirely different principle, with diverse neuron types arranged in soft, three-dimensional networks that constantly rewire themselves.
"Silicon achieves complexity by having billions of identical devices," Hersam explained. "Everything is the same, rigid and fixed once it's fabricated. The brain is the opposite. It's heterogeneous, dynamic and three-dimensional."
This distinction carries enormous practical weight. Modern AI training demands staggering computational power. Large data centers now consume enough electricity to require dedicated nuclear power plants, with cooling systems straining water supplies. The human brain performs far more complex tasks while consuming roughly one hundred thousand times less power than a digital computer.
The new printing approach offers practical advantages beyond performance. The manufacturing process is straightforward and inexpensive, using aerosol jet printing to deposit materials only where needed. The flexible substrates could eventually enable implantable devices for neural interfaces, including prosthetics that restore hearing, vision, or motor control.
The research, appearing in Nature Nanotechnology, was supported by the National Science Foundation. Hersam collaborated with neuroscientist Indira M. Raman, whose team at Northwestern verified that the artificial signals could actually drive biological neural circuits.
The implications reach far beyond laboratory demonstrations. As companies build ever-larger AI systems, the energy problem becomes a hard ceiling. Scaling traditional architectures beyond current levels appears physically and economically infeasible. Brain-inspired computing isn't speculative anymore; it's becoming a necessity.
Author Jessica Williams: "This isn't just incremental progress in neurotechnology; it's proof that silicon-based thinking about computing is finally hitting its limits, and biology offers a proven alternative that actually works."
Comments