Tetsuwan Scientific, a startup based in San Francisco, is developing advanced artificial intelligence (AI) robotics capable of undertaking scientific activities. Co-founded by CEO Cristian Ponce and Chief Technology Officer (CTO) Théo Schäfer, the company emerged from stealth mode in November following a successful seed funding round. The firm’s goal is to create intelligent software that can seamlessly integrate with laboratory robotics to automate all aspects of scientific discovery, including hypothesis formation, experimentation, and conclusion generation.
Creation of AI-Driven Robotic Scientists
Established in 2023, Tetsuwan Scientific spent the previous year and a half developing its debut product — a robotic AI capable of conducting experiments. Now that it has unveiled its operations, the startup is collaborating with La Jolla Labs on RNA therapeutic drug development. The firm outlines its vision and initial product details on its website, although it has yet to release any public products.
The startup identifies a significant challenge in the current landscape of scientific automation: existing systems emphasize high volumes of experiments rather than diversity. Current laboratory robots often require extensive programming to follow specific protocols, effectively transforming research into assembly lines rather than fostering collaboration with scientists, according to the company.
Tetsuwan Scientific points out that existing robots struggle to comprehend scientific intent, rendering them incapable of independently conducting experiments. However, inspired by advancements in generative AI models, the company believes it is now possible to overcome this communication barrier by teaching robots to behave like scientists. This objective involves developing intelligent software in conjunction with flexible robotics hardware.
During an interview with TechCrunch, Ponce explained that large language models (LLMs) can help bridge the gap between software and robotics. These models enable developers to convey scientific intentions to robots without needing to write extensive code. He also mentioned that the Retrieval-Augmented Generation (RAG) framework could help minimize instances of AI hallucination.
According to reports, Tetsuwan Scientific is designing non-humanoid robots. Featured on its website, these innovative robots have large, square glass-like structures that allow them to assess results and adapt scientific experiments autonomously. They are equipped with sophisticated AI software and sensors that enable them to understand and adhere to technical standards such as calibration and liquid characterizations, among other properties.
Presently, the startup is in the early phases of its ambitious journey to create independent robotic AI scientists capable of fully automating the scientific process and driving innovation.
