Fei-Fei Li

I think AI can expand humanity’s ability to understand, imagine and create. But AI is much bigger than a chatbot. Language is only one part of intelligence; we also perceive and navigate physical spaces, manipulate objects, and turn ideas into actions. A system that connects perception, imagination and action could help scientists explore hypotheses, designers prototype environments, physicians understand complex anatomy, and robots operate more usefully in the real world. That future is not automatic. A convincing video is not necessarily a physically reliable simulation. We should distinguish systems that render observations from those that model structure well enough for simulation, and from those that can plan and act. Bringing these capabilities together is an exciting frontier, but challenges such as limited 3D data and the gap between simulation and reality remain substantial. Most importantly, the future of AI is a human choice. Technology does not decide whom it serves—we do. Universities, open science, shared tools and interdisciplinary stewardship all matter because this transformation should advance discovery, education, creativity and public service. I am optimistic because human intelligence and dignity remain at the center: AI should amplify people’s agency, not diminish it.

Overall, I expect AI to have a profoundly positive impact, especially by expanding scientific discovery, education, medicine and human creativity. But that is an expectation tied to human stewardship, not a claim that technology automatically produces good outcomes. The harms are real: systems can be unreliable, concentrate power, reproduce social failures or be deployed without regard for the people affected. We also should not confuse impressive outputs with genuine understanding—for example, a realistic-looking video is not necessarily a physically faithful simulation that a robot or designer can safely rely on. So I reject both easy utopianism and fatalism. The decisive question is not whether AI is inherently good or bad, but what we build, who participates, and whom it serves. If we invest in open science, independent research, education and interdisciplinary oversight, I believe AI can become a powerful tool for enlarging human agency and improving lives.

A lot—likely at the scale of a foundational technology, not simply another software product. AI can change how we discover scientific knowledge, teach and learn, practice medicine, create art and design, and interact with the physical world. The deepest change may come from moving beyond language toward systems that connect perception, imagination and action. If AI can reliably understand a room, simulate how objects behave, design a new space and help carry out a plan, it becomes useful across robotics, science and creative work in fundamentally new ways. That said, I would not say AI changes everything completely. Human needs, relationships, institutions and choices remain central. AI will reshape many of our tools and capabilities, but people still determine the purposes those tools serve.

I would not assign a credible number. Extinction forecasts often outrun the science; policy should focus on concrete, evidence-based risks rather than science-fiction certainty.