Neural inference at the frontier of energy, space, and time.

Computing, since its inception, has been processor-centric, with memory separated from compute. Inspired by the organic brain and optimized for inorganic silicon, NorthPole is a neural inference architecture that blurs this boundary by eliminating off-chip memory, intertwining compute with memory on-chip, and appearing externally as an active memory chip. NorthPole is a low-precision, massively parallel, densely interconnected, energy-efficient, and spatial computing architecture with a co-optimized, high-utilization programming model. On the ResNet50 benchmark image classification network, relative to a graphics processing unit (GPU) that uses a comparable 12-nanometer technology process, NorthPole achieves a 25 times higher energy metric of frames per second (FPS) per watt, a 5 times higher space metric of FPS per transistor, and a 22 times lower time metric of latency. Similar results are reported for the Yolo-v4 detection network. NorthPole outperforms all prevalent architectures, even those that use more-advanced technology processes.

Artificial brains. A million spiking-neuron integrated circuit with a scalable communication network and interface.

Inspired by the brain's structure, we have developed an efficient, scalable, and flexible non-von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.

Shot-noise limited detection for surface plasmon sensing.

It is demonstrated that surface plasmon sensing can be performed in the shot-noise-limited regime to resolve index of refractive changes on the order of 10-10/√Hz at input powers of 1 mW. This improved resolution is achieved by using active electronic noise cancelling to suppress laser intensity noise and a wavelength that maximizes sensitivity to index of refraction changes occurring at an interface. The resolution of the system is experimentally demonstrated by measuring the refractive index change of air in response to pressure changes.

Enhanced autocompensating quantum cryptography system.

We have improved the hardware and software of our autocompensating system for quantum key distribution by replacing bulk optical components at the end stations with fiber-optic equivalents and implementing software that synchronizes end-station activities, communicates basis choices, corrects errors, and performs privacy amplification over a local area network. The all-fiber-optic arrangement provides stable, efficient, and high-contrast routing of the photons. The low-bit error rate leads to high error-correction efficiency and minimizes data sacrifice during privacy amplification. Characterization measurements made on a number of commercial avalanche photodiodes are presented that highlight the need for improved devices tailored specifically for quantum information applications. A scheme for frequency shifting the photons returning from Alice's station to allow them to be distinguished from backscattered noise photons is also described.