Computing with Living Neurons: Chaos-Controlled Reservoir Computing with Knowledge Transplant

A team of researchers from institutions including the University of Illinois Urbana-Champaign has published a groundbreaking paper on arXiv titled 'Computing with Living Neurons: Chaos-Controlled Reservoir Computing with Knowledge Transplant.' The work introduces a novel framework that makes living neural cultures—dishes of brain cells—into viable, stable computational substrates. The core innovation is chaos-controlled Reservoir Computing (cc-RC), which first identifies each culture's unique, chaotic 'dynamical signature' and phase-portrait attractor. It then applies low-power optical control to stabilize this spontaneous and stimulus-evoked activity, creating a predictable regime where a simple readout layer can be trained for tasks like pattern classification.

Across hundreds of neural samples, this method proved robust, improving both computational accuracy and the functional longevity of the biological system by approximately 300% compared to standard Reservoir Computing approaches. The team's second major breakthrough, Knowledge Transplant (KT), addresses a fundamental bottleneck: training each new biological system from scratch. KT allows the reservoir map learned by a high-performing 'expert' culture to be transplanted into a new 'student' culture with an equivalent dynamical attractor. This transfer reduces training time from hours or days to mere minutes while also boosting performance, effectively enabling reusable, cross-substrate AI models.

By creating a method to stabilize biological computation and share learned functions between living systems, this research paves the way for knowledge accumulation that transcends the lifespan of individual cell cultures. It represents a significant step toward hybrid biocomputing systems that leverage the innate efficiency and adaptive potential of biological neurons for next-generation, low-power intelligent machines.