Physical AI & humanoids: Hype vs reality

Humanoid robots sit at the intersection of engineering, myth, and design. While they promise to help us in so many ways, they also awaken very old anxieties: loss of control, deception, replacement, and the uncanny feeling of encountering something almost, but not quite, human. As humanoids move from our labs into real workplaces and public spaces, the challenge is no longer only technical. A robot can walk, lift, speak, and act—but if people do not understand it or feel safe around it, it has not truly succeeded. When argodesign developed the Apollo humanoid with Apptronik, we learned a lot of lessons about how to shape a machine that real people will understand and trust. Successful humanoids, like Apollo, must be more than technically adept. They must be legible, approachable, and honest. Their bodies, faces, screens, gestures, and actions need to help people understand who they are, what they are doing, and how to relate to them. Our humanoid robotic future will be shaped by more than our brilliant engineering—it will depend on whether we can design machines that real people are willing to live and work with. 

Ever since the explosion of AI due to the popularization of large language models, researchers have been accelerating their attempts to achieve artificial general intelligence (AGI). Today's AIs and their agentic applications have their foundations in machine learning techniques that have their origins in the ‘70s that rely on the availability of enormous amounts of energy and data. As we adapt these models to be more generalizable to create AGIs with the capability of interacting with the physical world, our requirements for more energy, water, and data further stresses the finite resources and infrastructure that we currently have in coexistence with the needs of society. This talk will explore a possible alternative future where compute is not powered by silicon but instead by biology with the goal of furthering our search for AGI but with less demands on energy, data, and water. 

Artificial intelligence is leaving the cloud and entering the world, not as abstract code, but as a property of physical systems such as robots themselves. This is the promise of physical AI: intelligence that is compact, adaptive, and embodied, inspired by the dynamics of living systems. Such AI could make our robots more efficient, trustworthy, and human-centered, but it also forces us to confront profound questions. What does it mean when intelligence no longer sits apart from the world, but is woven into its fabric? Will physical AI become a foundation for resilience and care, or will it bind us to technologies we cannot escape or control? 

Humanoids are poised to have a revolutionary social impact, and yet we still don’t know if we should love or hate them. While we’ve been building humanoids for over 2,000 years, there is a huge difference between one or two experimental or entertainment androids and commercial deployments of fleets of humanoids. 

Here are the facts and figures from the front row of the robot revolution, in a context that goes beyond asking what is technologically possible and considers what is socially valuable, where investment is leading robotic development, and how humans can still steer the technological tsunami. It’s our story and we can write a happy ending.