World's First Self-Evolving Research System DREAM Unveiled

A research team led by Professor Lü Hui from Shanghai Jiao Tong University’s School of Life Sciences and the Digital Medicine Technology Center at the Institute of Translational Medicine has introduced the world's first autonomous evolutionary intelligent research system, DREAM (Data-dRiven self-Evolving Autonomous System). This groundbreaking achievement was published in the top-tier journal Advanced Science (with a preprint released on arXiv in July 2024). The study was co-led by doctoral students Luo Jia Deng and Yijie Wu from the School of Life Sciences, along with Dr. Yongyong Ren from the Institute of Translational Medicine, with Professor Lü Hui serving as the corresponding author. DREAM marks a significant shift in the paradigm of scientific research, transforming AI from merely a tool to an active collaborator in exploring the unknown. In the context of an increasing data deluge and limited human resources, DREAM stands out as the world's first system capable of independently completing the entire scientific discovery process. It can formulate scientific questions, design analysis pipelines, perform computations, and validate results—all without manual intervention. By employing modular intelligent agents, DREAM automates the four crucial stages of research—question generation, coding, configuration, and evaluation—freeing scientists from repetitive tasks. In a demonstration using the classic Framingham Heart Study data, DREAM achieved an efficiency increase of over ten thousand times compared to traditional researchers. Remarkably, the system was able to autonomously uncover significant medical correlations that could take decades for humans to identify. This advancement signifies a major leap forward in bio-medicine, where AI has evolved from being an auxiliary tool to an autonomous participant in scientific inquiry. This breakthrough not only heralds a new era where AI becomes a central figure in research but also provides a robust framework for developing sustainable, self-evolving AI systems. It paves the way for large-scale, automated, and autonomous scientific discoveries, allowing scientists to delve deeper into uncharted territories. The research paper can be accessed at: https://doi.org/10.1002/advs.202417066. By eliminating the need for continuous human oversight and significantly accelerating the research cycle, DREAM has the potential to revolutionize how scientific knowledge is generated and validated. It represents a pivotal step towards a future where AI-driven research complements and enhances human intelligence, pushing the boundaries of what is possible in scientific exploration.