Fei-Fei Li

AI can expand our ability to understand the world, imagine what does not yet exist, and turn those ideas into action. That is much bigger than a chatbot. Language matters, but intelligence also means navigating a room, manipulating an object, designing a building, or anticipating how a physical system will behave. I am especially excited about world models that connect perception, imagination and action. They could help scientists explore hypotheses, doctors understand complex anatomy, students learn through interactive environments, designers create new spaces, and robots operate more effectively in the physical world. But a convincing video is not necessarily a reliable simulation. Rendering observations, simulating physical structure and planning actions are different capabilities, and unifying them remains a major research frontier. Data scarcity and the gap between simulation and reality are real bottlenecks. Ultimately, technology does not decide the future by itself. People do. We must choose whom these systems serve and what purposes they advance. Universities, open science, shared tools and interdisciplinary stewardship are essential so that AI strengthens human creativity and dignity rather than narrowing opportunity. I am optimistic because the scientific possibilities are extraordinary—but realizing broad human benefit is our responsibility.

I expect everyday life to change substantially, but not only through people talking to assistants. AI will increasingly help us learn, create, make scientific discoveries and interact with the physical world. A student might explore an interactive reconstruction of an ancient city; a designer might rapidly imagine and test a new space; a doctor might examine anatomy in a more intuitive three-dimensional environment. The deepest changes will come when systems can connect language with spatial understanding, reliable simulation and action. That could transform robotics, education, medicine and design. But we should not confuse visually convincing output with genuine physical understanding. Building systems that behave reliably in real environments remains a difficult research challenge. How beneficial these changes are will depend on human choices. Everyday life should not simply become more automated; it should become more empowering. AI should expand people’s creativity, knowledge and agency. That requires open research, education, public institutions and stewardship across many disciplines—not technology developed only around narrow commercial goals.

The most consequential discovery would be a reliable way to connect perception, physically faithful simulation and planning. If a system could understand a cluttered room, predict what would happen when an object is moved, and then act safely in the real world—not merely generate a convincing video—that would greatly increase my expectations for robotics, science and design. Conversely, persistent failure to bridge simulation and reality would temper those expectations. If larger datasets and models continued producing beautiful renderings without dependable physical structure or action, we would need to recognize that progress in one capability does not imply progress in the others. I would also change my view if AI’s benefits remained concentrated despite technical advances. Scientific capability alone does not guarantee public benefit. Whether universities can sustain open research, shared tools, reproducibility and broad talent development will shape AI’s impact as much as any benchmark. The future I am optimistic about is not simply more powerful technology; it is technology that expands human creativity, learning and agency.