| Mobile Web

Unconventional AI unveils new AI architecture aiming to cut inference power use 1,000-fold

Unconventional AI, founded by former Databricks AI head Nabeen Rao (나빈 라오), unveiled a new oscillator-based computing architecture it says could sharply reduce power use for AI inference. The company said it aims to cut energy consumption to as low as one-thousandth of existing AI systems over the long term. It also released its first AI model, the image generation system Un-0, and a research paper, with results shown so far in software simulation rather than a physical chip.