The rapid expansion of modern artificial intelligence systems has led to a substantial increase in the energy consumption of computational infrastructure. Contemporary deep learning models are typically trained and executed on large clusters of graphics processing units (GPUs), whose power requirements pose growing economic and environmental challenges. Biological neural networks cultured in vitro represent an alternative computational substrate characterized by intrinsic synaptic plasticity, nonlinear dynamics, and extremely low energetic requirements relative to conventional digital hardware. This paper examines the architecture and computational potential of biological neural computing systems, with particular focus on the CL1 platform developed by Cortical Labs. The platform integrates living neuronal cultures with high-density microelectrode arrays and real-time runtime environments enabling closed-loop interaction between biological networks and simulated environments. Experimental evidence from in vitro neural systems is discussed together with comparisons to GPU-based and neuromorphic computing architectures. The analysis suggests that hybrid bio-digital architectures may provide promising directions for energy-efficient artificial intelligence and new experimental frameworks for studying neural computation.
Henrique Peixoto (Tue,) studied this question.