The drive to discover the next big thing in AI has funded some pretty ambitious projects, but one company sees this as an opportunity to rebuild its computing architecture from the ground up.
Unconventional AI, led by former Databricks AI head Naveen Rao, promises to significantly improve the power efficiency of inference processing. The secret weapon: a new kind of oscillator-based computer architecture.
On Thursday, the company released its first model AI, called Un-0. This is the first image generation system tool to demonstrate how the company’s technology can replicate traditional AI systems. The accompanying new paper details how the company’s research team built a fully functional image generation model using software simulations of a new architecture that performs on par with state-of-the-art diffusion models.
“This is the ‘Hello World’ of a new kind of computer,” Rao told TechCrunch. “Over the next year, we’re going to start seeing some pretty interesting news on this.”
The output from the new Un-0 model is similar to that of image generation models such as Stable Diffusion and OpenAI’s GPT Image 1. What’s impressive is how they arrive at that performance. This model is built on an oscillator-based architecture that is completely different from traditional computing and the chips that drive traditional LLMs. Although the benefits of oscillator-based computing are complex, Rao believes it can ultimately reduce power usage by as much as 1,000 times.
Much of the infrastructure to get there is still being built. The current version of Un-0 runs on a software simulation of Unconventional’s oscillator chip, but the company plans to release a schematic of the actual chip soon. From there, the plan is to build the entire inference stack from scratch and eventually have Unconventional AI offer the same computing power as other providers.
“We’re building a new kind of system made of our chips,” Rao says. “We run the AI model there, and we use network cables to receive prompts and send inferences, and it runs on 1/1000th of the power.”
This is an incredibly ambitious goal, especially for a company with fewer than 50 employees. But given the scale at which AI is being built and the expected costs of meeting growing inference demands, this may be one of the few efforts to address the scale of the problem. In Rao’s view, available power supply will be one of the severe limitations for AI in the coming years. And Unconventional is one of the few projects that can address that.
“It’s hard to scale AI because of energy. This is going to be a fundamental limit in the next few years. We can’t go beyond it. At the end of the day, it’s going to be an energy-limited problem,” he says.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
