Scientists are teaching AI-powered robots to run lab experiments

Ginkgo Bioworks, a company founded by MIT graduates including Jason Kelly, has developed AI-powered robotic labs to automate complex biochemical experiments. The startup began nearly two decades ago with a focus on replacing manual lab work, initially struggling to secure funding but gaining momentum during the AI boom after a 2014 blog post by Sam Altman sparked interest. The company’s autonomous labs in Boston use robotic systems to handle tasks like pipetting chemicals and managing petri dishes, with AI translating experimental designs into robotic instructions. These labs serve pharmaceutical, agricultural, and government clients, conducting projects such as engineering microbes for fertilizer and creating proteins to combat ice formation. Recently, Ginkgo Bioworks advanced its approach by enabling robots to design experiments independently. Using OpenAI’s technology, including ChatGPT, the team challenged AI to generate protein designs—a task typically reserved for human scientists. The AI outperformed expectations, achieving a 40% cost reduction in protein synthesis and conducting over 30,000 experiments in six months, though the findings remain unpublished. Kelly highlights the lab’s efficiency, where robots operate alongside human oversight, executing tasks like transporting live-cell samples between stations. The collaboration with OpenAI represents a milestone in AI-driven scientific research, demonstrating the potential for machines to autonomously plan and execute experiments. Ginkgo Bioworks now operates as a leader in automated biotechnology, blending robotics, AI, and traditional lab work to accelerate discovery. The company’s work underscores the growing role of AI in transforming scientific processes, from drug development to agricultural innovation.