Shanghai's AI Scientist Can Now Run Experiments Solo

Scientists in Shanghai just gave researchers worldwide a powerful new teammate that never sleeps and can tackle multiple experiments at once.

The Shanghai Academy of AI for Science unveiled Dasheng on March 1, an AI assistant that can understand what scientists need in plain language, then independently design and execute research projects. Unlike previous AI tools that simply answer questions, Dasheng can split itself into multiple versions to handle different parts of complex studies simultaneously.

The academy itself emerged from a lightning-fast government decision. When Fudan University Professor Qi Yuan proposed building large-scale scientific AI models in 2023, Shanghai's leadership approved the plan within a week. By September, the academy officially opened its doors.

Since then, the team has been busy. They've integrated over 400 scientific models, compiled 22 petabytes of valuable research data, and fed the system more than 500 million academic papers and patents.

The academy created specialized AI models named after Chinese mythological figures, each targeting real-world challenges. The Suiren model built a library of 12,000 molecules to help discover better lithium battery materials. The Nvwa RNA model tackles bottlenecks in developing siRNA drugs, a promising class of medicines.

For climate research, they developed PI@Climate, trained on data from more than ten scientific disciplines. It's already supporting climate negotiations and policy decisions as China's homegrown climate science model.

Why This Inspires

The secret ingredient isn't just the technology. The academy's team averages just 33 years old, with 31 percent recruited from overseas universities and leading AI labs. Another 40 percent bring hands-on engineering experience from top tech companies.

They work in an environment designed for creativity. The institute runs a "research bar" where young scientists trade ideas for free drinks, fostering the kind of spontaneous conversations that spark breakthroughs.

Materials researcher Wang Wenli says the regular face-to-face interactions between scientists, AI experts, and engineers create something special. "What we create here naturally integrates the genetic strengths of multiple disciplines," she explains.

The setup works. The academy now collaborates with 43 teams at Fudan University and has launched 18 interdisciplinary centers. AI scientist Jiang Ruoxi admits the progress surprised even insiders: "The speed has exceeded all expectations."

President Qi Yuan aims to build a world-class scientific intelligence institution by 2030, and with young talent flocking to join this mission, that confidence seems well placed.