Engineering · Electrical and Electronic Engineering
Neuromorphic computing is moving toward larger, more programmable systems that process information in ways inspired by the nervous system. A review of recent work suggests these computers may be useful for low-power, real-time tasks at the edge, while major challenges remain in training, testing and manufacturing.
A preprint reports measurable patterns for how capabilities emerge and are forgotten during language-model training. The study says emergence depends on several model parts becoming useful together, while forgetting depends on how widely and strongly a capability is stored.
Researchers developed a light-based AI navigation architecture and tested part of it with a laser array. Simulations reported an 80% ± 8.3% success rate, while the tested laser activations used an estimated 0.78 nJ of energy and took 191.20 picoseconds under ideal parallel operation.
