A biomimetic neural encoder for spiking neural network
As we tread the boundaries of impossible in manufacturing through digital twins, predicting human behavior through encoding neuronal spikes is becoming less of an impedance.
Early this month a paper published in Nature is bridging the gap between Artificial Neural Networks and Biological Neural networks through something called Spiking Neural Networks SNNs. How it works is that it exploits biologically plausible neurons that offer faster inference, lower energy expenditure, and event-driven information processing capabilities.
The study found out that the encoding energy was as frugal as ≈1–5 pJ/spike. It also demonstrated that fast (≈200 timesteps) encoding of the MNIST data set using their biomimetic device followed more than 91% accurate inference using a trained SNN.
Assume human body is an IoT device controlled by the brain with its processors analyzing sensory information from the furthest extremities of the body. If all these spikes can reliably be recorded and stored and fed to low-cost, high-throughput, and real-time predictive algorithms, free-will will largely become deterministically biological.
You can already create cyborg cockroaches and game-playing monkeys. Optimistically it can help us understand our pain points and triggers and work on them.
