Inception, a startup, has raised $50 million in seed funding to advance the development of diffusion-based AI models for software development. The funding round was led by Menlo Ventures, with contributions from Mayfield, Innovation Endeavors, Nvidia’s NVentures, Microsoft’s M12 fund, Snowflake Ventures, and Databricks Investment. Angel funding was also provided by Andrew Ng and Andrej Karpathy.
The company is led by Stanford professor Stefano Ermon, known for his work in diffusion models. These models refine outputs iteratively, unlike traditional auto-regression models. While diffusion models are prevalent in image-based AI, Ermon aims to extend their application to a wider range of tasks through Inception.
One of Inception’s key achievements is the introduction of the Mercury model, designed specifically for software development. Mercury has been integrated into various development tools, including ProxyAI, Buildglare, and Kilo Code. Ermon states that the diffusion approach enhances model efficiency and optimizes metrics like latency and compute cost.
“These diffusion-based LLMs offer superior speed and efficiency compared to existing approaches, presenting substantial room for innovation,” Ermon said.
Source: TechCrunch