Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
From Bits to Volts: Achieving Faster and Greener AI Inference with Plug-in Analog Neural Hardware
This talk introduces a plug-in analog hardware module that performs neural network inference using voltages and currents, reducing power by 100–1000x and latency by up to 100,000x.
The AI infrastructure of tomorrow demands breakthroughs beyond power-hungry GPUs. This talk unveils a revolutionary plug-in hardware module that performs neural network inference using analog computation, slashing power consumption by 100–1000× and latency by up to 100,000× compared to digital accelerators. By encoding weights and biases as voltages and currents, our device computes at the speed of physics, enabling near-instantaneous inference with minimal energy. We’ll showcase SPICE-model validation proving digital equivalence, a roadmap from PCB prototype to ASIC implementation, and real-world use cases like edge AI, drones, and data-center acceleration. Discover how analog neural hardware redefines efficiency, making low-latency, low-power AI deployable anywhere, from IoT to defense. Join us to see the future of inference infrastructure.