Humans have to be painstakingly taught everything from scratch — all humans must learn to walk, talk, then finish school and pick up specialized skills in college — but robots might be able to skip all these steps altogether.
In a compelling articulation of the future of robotics, Brett Adcock, the CEO of the robotics company Figure, has outlined a vision where humanoid robots could rapidly outpace human learning and efficiency. His central thesis revolves around the concept of a shared neural network, a collective consciousness for a fleet of robots, allowing a single robot’s experience to become the instantaneous knowledge of all. This, he argues, will create a market dynamic unlike any seen before in advanced hardware.

Adcock’s vision begins with the economics of scaling up production. “As you put more robots out into the world that can do something useful, people will pay you for the cost of that robot. The per-output of work will come down because you’ll just make more of them,” he explains. “You’ll be up on the experience curve of manufacturing.”
But the true game-changer, according to Adcock, isn’t just in the manufacturing. It is in the learning. “The robots will learn to get better every day in the market, and they will share that with the collective fleet,” he states. This is a fundamental departure from biological learning. “Not like kids, where the kids have to learn how to walk. My kids have to learn how to walk. But our robots can share a neural network and do the same use case we talked about today or you’ve seen here with any other robot that came off the manufacturing line today.”
This creates a powerful feedback loop. “It learns as you go, as it’s making mistakes, as it’s doing things successfully. It learns like a human,” Adcock elaborates. The implications for the market are profound, potentially leading to a “winner-take-all” scenario. “This is one of the first industries, I think maybe in the world for advanced hardware, that it could be a winner or winners, maybe even a winner-take-all market where one group has the cheapest robot and it’s the smartest.”
The desire for the most capable and safest robot will drive this consolidation, in Adcock’s view. “You’re never going to want to have the dumb employee. You’re not going to want the person at your house that’s not the smartest; it’s dangerous. You’re going to want the smartest employee, and you’re going to want the smartest person in your house. And that will also be the cheapest.”
He concludes by emphasizing the unique nature of this technological evolution. “It’s not like a computer or a car or a phone. It’s a living agent that’s getting better and cheaper over time. The group or groups that can get a large amount of robots out that are useful to the market will have a significant advantage and lead over everybody else.”
Adcock’s assertions tap into one of the most exciting and potentially disruptive developments in artificial intelligence and robotics: the concept of fleet learning. Unlike individual humans who must acquire knowledge independently through years of study and practice, a network of robots can share experiences and insights instantaneously. A mistake made by one robot in a factory in one part of the world can become a lesson learned by the entire fleet globally in a matter of seconds. This prevents the same error from being repeated, dramatically accelerating the learning curve.
This concept is not merely theoretical. We are already seeing nascent versions of this in the automotive industry with companies like Tesla and Waymo. Data from its fleet of vehicles is used to improve its Full Self-Driving capabilities, with learnings from millions of miles driven by its customers feeding back into the system to enhance its performance. What Adcock proposes is applying this same principle to a more complex and interactive domain: general-purpose humanoid robots.
The implications for the competitive landscape are staggering. If a company can deploy a significant number of robots and enable this shared learning, it could create an almost insurmountable “intelligence moat.” A larger fleet gathers more data, which leads to a smarter AI, which in turn makes the robots more useful and attractive to customers, further increasing the fleet size. This virtuous cycle could leave competitors with smaller or less intelligent fleets struggling to catch up.
Recent developments in the field underscore this trend. Figure itself has made significant strides, recently partnering with BMW to deploy its humanoid robots in the automaker’s manufacturing facility. This provides a real-world testing ground for Figure’s robots to learn and perform tasks, with every success and failure contributing to the collective intelligence of the Figure fleet. Similarly, competitors like Tesla with its Optimus robot and Agility Robotics with Digit are also racing to deploy their robots and gather the crucial data that will fuel their AI’s development.
The race to build the smartest and most capable humanoid robot is not just about sophisticated hardware. It is a race to build the most effective learning ecosystem. The company that can master the art of collective robotic learning will not only lead the market but could fundamentally reshape the nature of labor and productivity for generations to come.