Dr. Fei-Fei LiCo-founder & CEO

Growing up in Chengdu during the eighties and nineties, my dad was a curious soul who got me interested in nature. I spent a lot of time chasing bugs, running after puppies, and sketching in the mountains. I cut my hair short, loved aerospace, and developed a strong identity around physics and science.

Moving to New Jersey was tough because shifting language, culture, and life as a teenager is incredibly difficult. You are trying to find who you are, and suddenly you do not even understand the world. I learned English, finished high school, and helped run my family's dry cleaning business on weekends.

It was a beautiful period because artificial intelligence was a niche research field. There was no money, no power, no fanfare, and no hype. We were driven purely by scientific curiosity, similar to astronomers looking at the sky or oceanographers exploring the depths of the ocean.

Humans use eyes to capture sensory signals that the brain computes. For computers, the process is similar. We teach them to learn patterns, understand colors, and recognize objects. Ultimately, seeing is not just about recognition, but about preparing humans to act and navigate through spatial intelligence.

At the time, the entire research field was focused on developing algorithms, and nobody was paying attention to data. My students and I had an epiphany that machine learning algorithms needed to be trained on massive amounts of structured data to get smarter and more accurate.

In 2012, his team entered our ImageNet competition with an algorithm called AlexNet. They combined our massive dataset of pictures with neural networks and NVIDIA graphics processing units. That combination became the golden recipe for modern AI, proving that data-driven deep learning worked.

I do not naturally think of myself as a godmother of anything, and it is not my personality to use that term. However, if I reject the title, it perpetuates a situation where women's contributions in science are not recognized. I want more women to get recognition for their discoveries.

I saw an opportunity to advance AI beyond chatbots. While the rest of the industry focused on language, I wanted to build systems that understand the physical, spatial world. I run the shop with high standards, and because our engineers and scientists are young and talented, I feel like a tiger mom.

We define spatial intelligence through three functions: rendering, simulation, and planning. Rendering outputs pixels for human consumption, simulation captures the geometric and physical structures of the world for machines, and planning tells a robot how to navigate and interact with physical objects in a de risk environment.

Language models cannot interact with the physical world. Words alone cannot put out fires, cook an omelet, or control a robot. Building world models and spatial intelligence is not about opposing language models, but about entering the next frontier of physical AI.

Marble generates explorable, editable three dimensional worlds from text or image prompts. Filmmakers use it for virtual production to shoot actors in any environment, game developers use it to reduce creation time, and robotics companies use it to augment physical training simulations.

At a massive technology corporation, your hands cannot change the trajectory of the business because the scale is too large. In a fifty person startup like World Labs, every researcher can directly change the company's trajectory, which is a beautiful opportunity.

Policy must be rooted in actual science, not science fiction. Conversations about humanity's extinction or robotic overlords distract from real policy work. We must also fund public research and STEM education to build human capital, rather than focusing solely on machines.

World models and generative systems can be used to create deepfakes and spread disinformation. Physical robots can be weaponized by bad actors if they are not properly governed. Additionally, students might use AI as a lazy crutch instead of an empowering learning tool.

It is dangerous for any individual to think they know what is best for everyone else. Our role as CEOs and scientists is to contribute to society and empower people, not to make decisions for them. We need democratic governance and collective input to guide this technology.

History shows that humanity collectively works to make life better for the next generation. We have collective moral compasses that bend the arc of technology toward benevolence. While we are in a messy transition, my ultimate hope resides in humanity, not in machines.