The startup Unconventional AI presented Un-0, a new model of artificial intelligence developed on a computing architecture different from the one currently used. The company claims that this technology could reduce energy consumption by up to 1,000 times the energy consumption of AI systems during inference.
The project is led by Naveen Rao, former head of the artificial intelligence area at Databricks. The company has already demonstrated the operation of its first model and announced that it is working on the development of the hardware necessary to bring this technology to market.
What is Un-0, the new artificial intelligence model
Un-0 is the first model presented by Unconventional AI. It is an image generation system designed to demonstrate that its new architecture can deliver results comparable to the most advanced diffusion models.
According to the company, the performance achieved is similar to tools like Stable Diffusion and GPT Image 1. The difference lies in the way it processes information.
How the new architecture for artificial intelligence works
The technology developed by Unconventional AI replaces the traditional chip-based architecture with a system built on oscillators.
Currently, Un-0 operates through a software simulation of those future chips. However, the company announced that it will soon publish the first designs of the hardware that will use this architecture.
