Batch #1

Category: Microelectronics, AI

Today’s computer hardware is inefficient for artificial intelligence. Supercomputer centers require a lot of time, huge data sets, and the energy of an entire power plant to train a complex AI model, which results in high costs and is unsustainable. GEMESYS Technologies offers an analog chip design based on the same information-processing mechanisms as the human brain. This enables AI hardware vendors to distribute a novel chip that trains neural networks 20,000 times more energy efficiently than current technologies. This not only significantly reduces the cost, time, and amount of data required to train a neural network, but also increases overall quality and performance. Thanks to its small size and high energy efficiency, it can be installed in almost any device and enables decentralized training, data processing, and decision making in the field.