Neuromorphic Computing: Living Neurons Learn Pong and the Future of Brain Inspired Computing

The video explores neuromorphic computing, focusing on how living human neurons in DishBrain learn to play Pong with no keyboard or controller. It explains how random electrical noise paired with predictable feedback can improve performance, illustrating the brain like learning and neural plasticity. It also introduces Cortical Labs CL1, a commercial platform where neurons grown from stem cells sit on a multi electrode array and learn from inputs and feedback, signaling a shift toward biology based processors for neuroscience research, drug discovery, and memory studies. The piece situates these technologies within a broader landscape that includes silicon based neuromorphic efforts and photonics, while noting challenges in reward signaling and remote access to a living computer.

Introduction to neuromorphic computing

The video examines a new paradigm in computing that mimics brain function and even uses living neurons to process information. Neuromorphic systems aim to be energy efficient, highly adaptive, and capable of real time learning, offering an alternative to traditional silicon based architectures.

Brains versus traditional computers

Traditional computers operate with a von Neumann architecture, separating memory and processing and relying on a central clock. In contrast, the brain processes information through interconnected neurons with memory and computation co located and a focus on spike driven communication. This leads to lower energy requirements and parallel, event driven processing that excels in pattern recognition and learning from small data samples.

DishBrain and the Pong experiments

In 2022 scientists taught a living network of human neurons to play a video game without a keyboard or controller. They tested learning under a theory called the free energy principle, feeding random noise as feedback when the cells erred and providing predictable feedback when they succeeded. The results showed rapid adaptation and highlighted the plasticity of neural networks under real world tasks.

The CL1 platform and commercial potential

Cortical Labs has moved beyond experiments to a commercial platform CL1, where living neurons grown from stem cells reside on a multi electrode array and learn from inputs and feedback signals. Learning emerges without traditional code, and the device can be accessed remotely like cloud computing, albeit the cloud is alive. The platform promises applications for neuroscience research, drug discovery, and exploring learning memory and disease, though challenges remain in designing positive and negative feedback and reward cycles.

Broader context and future directions

The video surveys other players including silicon neuromorphic chips and photonic brain inspired systems, emphasizing a shift from faster clocks to more intelligent, biology inspired computation. It highlights potential benefits for edge devices, robotics, and space exploration while underscoring hurdles such as unpredictable biology and the difficulty of translating living neural systems into robust processors. The takeaway is that neuromorphic computing invites us to rethink intelligence and how we build tools to understand memory, learning, and disease, potentially ushering a future where neurons in a dish teach us how to think.