At the Materials Artificial Intelligence Laboratory of the Shenzhen Institutes of Advanced Technology of the Chinese Academy of Sciences, a dual-arm robot named Dr.Mindis independently completing an entire centrifugation experiment, from opening the lid and picking up test tubes to balancing samples, recapping, and resetting, without any human intervention.

Developed independently by Professor Yu Xuefeng's team, China's first embodied AI robot scientist system for materials research overcomes the limitation of conventional AI systems, which can perform simulations but lack the ability to conduct physical experiments. Powered by the integration of a dual-mode brain-inspired AI agent and self-learning embodied algorithms, the system has established a fully autonomous research loop, significantly shortening the development cycle for new materials. It has successfully developed key materials including magnetic microspheres, black phosphorus and lithium iron phosphate, areas long dominated by overseas companies. The project stands as a typical example of Guangdong's AI Plusinitiative driving new quality productive forces.

The technology system is now being gradually expanded to high-risk experimental environments such as cleanrooms for cell preparation. It can take over repetitive and demanding laboratory procedures, allowing researchers to move away from routine operations and focus on original scientific exploration. Industrial outcomes are tangible: over the past five years, the team has undertaken more than 250 million yuan worth of commissioned R&D projects from enterprises, established eight innovation consortia with investment over 10 million yuan each, and holds 120 invention patents and 14 software copyrights. Twenty‑nine core technologies have been commercially transferred, with total transfer value exceeding 40 million yuan.

Going forward, the team will continue to broaden the robot's applications into life sciences, precision inspection, and other fields; refine standardized and modular autonomous laboratory solutions; leverage national pilot-scale testing platforms to expand industrial services; and collaborate with upstream and downstream enterprises to build a complete industrial ecosystem. The goal is to persistently tackle critical material bottlenecks and advance self‑reliance in the new materials industry through human‑machine collaborative innovation.

深圳造出“机器人科学家”！AI亲手做实验，数十倍提速攻克材料“卡脖子”难题

在中国科学院深圳先进技术研究院材料人工智能实验室，一台名为明德博士（Dr.Mind）的双手机器人正独立完成整套离心实验，开盖、夹取试管、对称配平、合盖复位全程无需科研人员介入。

这套国内首个面向材料研发的具身智能机器人科学家系统，由喻学锋团队自主研发，一改传统AI只会推演、无法实操的短板，依靠“双模类脑智能体+自主学习具身算法”搭建完整自主科研闭环，大幅压缩新材料研发周期，接连拿下磁性微球、黑磷、磷酸铁锂等长期被海外垄断的关键材料，是广东落地“人工智能+”行动、激活新质生产力的典型样板。

目前，该技术体系逐步拓展至细胞制备等超净间高危实验场景，替代科研人员完成重复、严苛的实验操作，让科研工作者从繁杂实操中抽身，专注原创科学探索。产业转化成果直观体现技术落地成效：近五年，团队累计承接企业委托研发项目金额超2.5亿元，落地8个千万级创新联合体；手握120项发明专利、14项软件著作权，29项核心技术完成市场转化，转化总额超4000万元。

下一步，团队将持续拓宽机器人应用边界，向生命科学、精密检测等领域延伸，完善标准化模块化自主实验室方案，依托国家级中试平台扩大产业开放服务，联动上下游企业搭建完整产业生态，持续攻克高端材料卡脖子难题，以人机协同创新推动新材料产业自主自强。

文、图｜记者 沈婷婷
译｜魏薇
英文审校｜曾敏