Baidu Famo Self-Evolving Research Agent Discussed in Podcast
A Baidu podcast featured Li Annan (Baidu Famo product lead) and Prof. Lin Haifeng (Nanjing Forestry University) discussing AI for science. Prof. Lin's team studies early detection of pine wilt disease using spectral data, vegetation indices, and textures – a labor-intensive process. Students accidentally used Baidu Famo's self-evolving agent to generate experimental plans, achieving good results without understanding why. Li Annan explained the agent iteratively explores new solutions, either improving existing ones or trying new directions. Both agreed the agent helps find solutions humans wouldn't think of. Li envisions a 'research factory' where humans oversee multiple AI-run experiments, while Prof. Lin stressed that results must be explainable and human responsibility remains. The podcast also noted the importance of clearly defining goals and constraints for AI tasks.