Recogni reveals new AI computing method to cut costs and power use

Successfully tested on AI models from firms such as Meta Platforms, Recogni is working with a partner to make Pareto widely available and plans further announcements soon.
AI,water

Recogni, an AI chip and software startup backed by BMW, Bosch, and Mayfield, has unveiled a new computing method that could revolutionise the efficiency of AI systems. The patented system, called Pareto, uses a logarithmic approach to outperform current methods in running large AI models, potentially making AI chips smaller, faster, and less costly to operate. This new method significantly reduces power consumption by converting multiplication operations into simpler additions, all while maintaining accuracy.

Recogni has already tested Pareto on AI models from companies like Meta Platforms and Stability AI. The company’s first chip was developed using Taiwan Semiconductor Manufacturing Co’s seven-nanometer process. Recogni is now collaborating with an undisclosed partner to make Pareto more widely accessible, with further details expected in the coming months. The startup is also considering offering its technology through data centres, allowing broader access to its innovative AI computing solution.