Archive — US — techcrunch
Inception raises $50 million to build diffusion models for code and text
Inception, a startup focused on diffusion-based AI models, raised $50 million in seed funding led by Menlo Ventures. Founded by Stanford's Stefano Ermon, Inception aims to apply diffusion models—known for efficiency and speed—specifically in software development. These models show potential to outperform traditional auto-regression models while managing large datasets and codebases effectively.
Key points
- Inception raised $50 million in seed funding led by Menlo Ventures.
- The startup focuses on diffusion models for code and text, championed by Stanford professor Stefano Ermon.
- Diffusion models are different from auto-regression models, processing outputs holistically and incrementally.
- The new Mercury model is integrated into tools like ProxyAI and KiloCode, enhancing software development efficiency.
- Diffusion models can operate much faster (1,000 tokens per second) and with lower compute cost than auto-regressive models.
Why it matters
- The rise of startups in AI, similar to trends observed in 2024 with increased investment and innovation.
- Diffusion models are becoming more prominent in AI research, marking a shift in how AI models are developed and applied.
- The growing interest in efficiency and speed in AI processing highlights the importance of new model architectures.
Inception's funding and focus on diffusion models illustrate a significant shift in AI development strategies, showcasing the potential for these innovative approaches to reshape industries, particularly in software development.