Just posted to my Forbes column about why the future of robot competitions shouldn’t look anything like the human Olympics.
We already have robot marathons, boxing matches, and humanoid games that test athleticism. But former Google roboticist Benjie Holson argues we’re missing the point. What actually matters isn’t whether robots can run fast or punch hard, it’s whether they can open a door, fold laundry, clean a countertop, or use basic tools in messy, unpredictable environments.
In other words, whether they can do the boring, everyday work humans actually want help with.
Holson’s proposed “robot Olympics” focuses on tasks that are trivial for people but brutally hard for machines: manipulating fingertips, handling flexible objects like clothes and plastic bags, wet cleaning, and coordinating multi-step actions like opening a door without getting stuck or hit. Physical AI company Physical Intelligence recently tested its latest models against these challenges, with impressive results … but even then, some real-world shortcuts (like replacing dog poop with a wooden block) show how far there still is to go.
The deeper point is Moravec’s paradox: the things humans find effortless are often the hardest for robots.
These proposed humanoid games make that gap visible and measurable. If we really want robots that are useful in homes, offices, factories, and job sites, these are the benchmarks that matter. Not medals for speed or strength, but quiet competence at getting real work done.