Discovery Loop is a new venture focused on automating the scientific discovery process through frontier AI models and large-scale computational infrastructure. The traditional scientific method, while powerful, is often slow and labor-intensive due to repetitive experimental loops. Discovery Loop aims to accelerate progress by automating these cycles, enabling rapid proposal, execution, and learning from thousands of experiments in parallel.

Initially, the company will focus on automating machine learning research and engineering, using these capabilities to optimize its own technology stack before expanding to other scientific and engineering domains. Their ultimate goal is to tackle Grand Challenges outlined by the National Academy of Engineering (NAE), such as developing better medicines and making solar energy economical.

The founding team comprises Jeff Dean, Sanjay Ghemawat, Quoc Le, and Oriol Vinyals, highly cited researchers with extensive experience in AI and distributed systems. They have a history of pioneering large-scale computing and creating foundational technologies at Google, including TensorFlow and Gemini. This team's expertise spans from chips to AI models, positioning them to build systems capable of solving complex problems and driving significant advancements for humanity.