AI, Humanoid Robots and the Next Leap in Human Imagination

As the PCB industry absorbs AI’s impact on nearly every aspect of business, the latest humanoid robots are showing how the technology can multiply our ability to imagine.

AI is one of my favorite subjects, as you may have noticed, and I make no apology. It’s already transforming our lives, and this will continue for years to come. Fixating on those two letters is easy, as they are currently evoking many emotions, including both excitement and fear. In the PCB industry, we are seeing how AI server demand is drastically reshaping the board market and the supply chain for substrate materials. So much so that, within Ventec, we feel the market will never be the same again.

Instead, I’d like to look at how this technology is giving us new tools to realize innovations that have driven our fantasies for generations. Humanoid robots, for example. Humanoids were probably the first types of robots people imagined. They have existed in stories since long before the word “robot” was coined, even in Greek mythology. And of course, we have seen early humanoids, such as Asimo, with its endearing, close-to-authentic human gait. Some may have found it comforting to focus on that awkwardness, feeling reassured that robots will never beat us.

Well, now, something big is happening, and the outlook has changed. Suddenly, humanoid robots are performing complex dance routines in perfectly coordinated troupes, balanced, fast and with fluent movements. They are also running, and the best are already faster than human athletes. More than 100 humanoid robots from institutions across China took part in a robot half-marathon held recently in Beijing’s E-town. The winner’s time, 50 minutes 26 seconds, was almost seven minutes faster than the current human record. I’m expecting that margin to have grown significantly by the next event.

Footage on YouTube, showing a humanoid robot chasing a group of boars in an urban street, is another impressive display of speed and balance while running. What’s also interesting in this clip is that the robot seems to decide for itself when the boars have been seen off, waving them away with a gesture as if to say, “And don’t come back!”

This is a noticeable – and sudden – shift from Asimo’s cute baby steps just a short time ago to the bad-to-the-titanium-bone machines we see today. And AI is the key to the breakthrough. This is partly enabled by robotics moving beyond constraining bottlenecks in areas like computing, sensing, actuation and algorithms. More fundamentally, however, AI now lets robots learn to move rather than following explicit instructions that tell them how to move. Leveraging techniques such as reinforcement learning with model-based perception and control, they can dynamically balance, coordinate their limbs, adjust their position and compensate for errors. It’s the reason humanoid robots can now move so fluidly, adapt, coordinate in groups, and run faster than human beings – even if that’s partly due to the optimized physiology of the winning machine, with its large hip motors, liquid-cooled power modules and lightweight upper body.

Some observers have noted that the dancing humanoids are following a prescribed choreography and that the half-marathon winner was running on a pre-laid course. This means the robots cannot generalize to fit their learning to changing or unfamiliar surroundings.

Although this is true, robotics recognizes multiple types and levels of generalization. While the early Asimo had no ability to generalize and could perform only basic, prescribed actions in a controlled space, today’s humanoids can adapt in limited ways to handle small variations in their environment and unexpected events. Research is now aiming for much broader generalization, and as these goals are reached, humanoids will begin to challenge human abilities. We’re not there yet, but we can expect rapid improvements now that simple generalization is complete.

At some time in the future, then, we know robots will approach our position as the planet’s apex generalist. We need to address how we task these machines so we can stay in control. Isaac Asimov’s Three Laws of Robotics, from the “I, Robot” stories of the 1940s, constrain robots and impose ethical standards that protect and grant authority to humans. They are often cited in these discussions, but are these laws already being pushed aside as robot technology is increasingly pervading military activities? Drones, for example, are transforming today’s conflicts by using techniques such as loitering over a target and dropping instantly when commanded. It contravenes Asimov’s entire construct in one action.

We are probably only at the very beginning of military robotics. Its relatively low cost – in terms of procuring equipment and in the risks to trained soldiers – raises the prospect of wars becoming increasingly affordable and faster and easier to launch.

We have known for some time that working out ethics will be the most difficult aspect of coming to terms with AI’s influence on our world. Governments, while having the power to legislate, are also keen to leverage the opportunities AI can provide.

But let’s not dump Asimov just yet. He was not the first, by thousands of years, to conceptualize what we now call humanoid robots. But he and similar writers show us the role of talented imaginations to make our fantasies visible and desirable. This is a powerful catalyst for scientists and engineers to begin exploring the possibilities, using their skills and knowledge to bring them to life in the world.End of article content 

Alun Morgan is technology ambassador at Ventec International Group (venteclaminates.com); alun.morgan@ventec-europe.com. His column runs monthly.