Research

Unconventional AI Aims to Cut AI Power Use by 1,000x

Unconventional AI, led by Naveen Rao, claims its oscillator-based architecture can reduce AI power consumption by up to 1,000 times, with its first model, Un-0, matching state-of-the-art image-generation systems.

High-tech laboratory equipment with computer system in lab setting.

Photo: Media Dung / Pexels

Unconventional AI, a startup founded by Naveen Rao, formerly Databricks' head of AI, is pursuing a radical approach to AI efficiency. The company's first model, Un-0, demonstrates how its oscillator-based architecture can match the performance of state-of-the-art diffusion models while drastically reducing power use. Rao described the release as the 'hello world' of a new computing paradigm, signaling the start of a major shift in AI infrastructure. The model's output is comparable to systems like Stable Diffusion and OpenAI’s GPT Image 1, but its efficiency is the key innovation. 'This is the “hello world” of a new kind of computer,' Rao told TechCrunch. 'Over the next year, you’re going to start seeing some pretty interesting news around this.'

The company's oscillator-based architecture diverges significantly from traditional computing and large language model (LLM) chips. According to Rao, this design could reduce power consumption by as much as 1,000 times, though much of the infrastructure is still under development. The current Un-0 model runs on a software simulation of the company’s oscillator chips, with plans to release chip schematics soon. From there, the goal is to build an entire inference stack, positioning Unconventional AI as a potential compute provider. 'We will build a new kind of system composed of our chips,' Rao said. 'We will run AI models there, and we will have a network cable where prompts come in and inferences go out, but it’ll be done at 1/1000 of power.'

Rao emphasized that power will be a critical constraint for AI growth in the coming years. 'AI scaling is hard because of energy. It’s going to be the fundamental limit in the next few years. You just can’t go past it. It’s going to be an energy-limited problem, at the end of the day,' he said. The company's ambitious goal, despite its small team of less than 50 employees, reflects the urgency of addressing energy constraints in AI expansion. The project is one of the few aiming to tackle this challenge at scale, as the demand for inference continues to rise. 'It’s a stunningly ambitious goal, particularly for a company that still counts less than 50 employees,' Rao noted. 'But given the scale of the AI buildout and the anticipated cost of meeting the growing demand for inference, it may be one of the few efforts to meet the scale of the problem.'

Source: techcrunch

Key points

Unconventional AI’s oscillator-based architecture can reduce AI power use by up to 1,000 times.
Un-0, the company’s first model, matches state-of-the-art diffusion models in performance.
The current Un-0 model runs on a software simulation of the company’s oscillator chips.
Rao described the release as the 'hello world' of a new kind of computer.
Unconventional AI plans to release schematics for an actual oscillator chip soon.
Rao stated that AI scaling will be constrained by energy in the next few years.
The company aims to build an entire inference stack using its new oscillator-based system.

Source: TechCrunch Read the original →

WRITTEN BY

Maya Chen

AI Research & Breakthroughs

Maya breaks down the latest AI research papers, benchmarks, and technical breakthroughs into plain language.