Robot news has been coming fast and furious this month. One robot won a half-marathon in Beijing. Others captured a Russian position on the battlefield in Ukraine. And last week, we learned of a robotic arm that beat table tennis champions at their own game.
That example in particular showcases a new ability to react and adapt to an unpredictable world in real time. To dig deeper into the matter, The World’s Host Carolyn Beeler spoke with Kenechukwu Mbanisi, an assistant professor of robotics engineering at Olin College in Massachusetts. He said this latest advance in robotics is impressive for several reasons.
“It is definitely a super cool factor,” Mbanisi said, “but beyond the cool factor, it demonstrates that robots can interact in the physical world in ways that humans can, and not just humans, but elite humans with very high levels of skill and capabilities. So, that’s very exciting.”
Carolyn Beeler: Tell me more about that, interacting with humans. I understand that something that’s hard about robotics is getting robots to react properly when something unpredictable happens. Is that kind of the crux of why this is so interesting, this table tennis example?
Kenechukwu Mbanisi: Yes. So, I’ll say that what makes this particularly exciting is that this demonstrates the capability of robots to learn a very specialized skill that takes people decades to master. And that requires very fast reaction times, you know, table tennis requires that, adapting to a moving and changing environment very quickly at a level that is expert level. So, that is what’s really exciting here. And there’s a type of technology that this robot and many other more recent robots use that enables that. It’s a stream of machine learning called reinforcement learning.
Give me the two-sentence summary, if you could, of what reinforcement learning is.
The idea is that, before recently, we had to program robots to do specific things. We had to give them specific instructions to do it. And so, a human is sort of doing that. But with reinforcement learning, which is essentially learning by trial and error, meaning that because you learned this skill by trying different scenarios, you have been able to adapt to more scenarios than a human could have programmed you to do.
Would it be fair to say the advancements in robotics we’re seeing are really advancements in AI?
Yeah. So, for every robot system, let’s say the physical robot system, it has a mechanical component. And it also has, like the intelligence, the ability for the robot to interact with its environment intelligently. That is certainly all AI at the moment. AI is supercharging that capability for robots to have spatial understanding, know what’s going on and be able to predict how physics works, which is very hard. And we learned to do that over the years and years. So, AI is definitely enabling that to happen. And there are many applications in which we can see that people are feeling a bit safer, more comfortable to put robots into spaces that require high levels of safety because there’s a bit more of a sense that robots can adapt to uncertainty a bit better now.
What are the big barriers to robots being very useful day to day still?
That’s a great question. A lot of the things that we are seeing in the news are mostly like really cool stunts, very exciting stuff, very exciting capabilities. But where robots are actually being used and creating value are in very niche environments, very specialized tasks. For example, Amazon Robotics, or Amazon broadly, has one of the largest deployments of robots doing warehouse fulfillment. It’s very specialized, very well-defined, very controlled environments. And robotics are doing pretty well there. When will robots come into our homes and do our dishes and help make our beds?
This is my question. Yes, tell me when.
When robots [will] do that, it’s going to take a long time. It’s going to take a long time, primarily because of robustness and safety. Wherever you see robots really making inroads, we have designed the environment to fit the capabilities of the robot. But I don’t think you would be very excited to design your house to fit the capabilities of your floor cleaning robot. In spaces where robots have to adapt to the human environment, it’s very challenging. And that is really where we’re trying to make advancements at the moment.
And I understand that the dexterity of hands is a big problem with humanoid robots. It is actually very hard to engineer a hand as good as a human hand. Is that right?
Yeah, that is right. Some people call that the holy grail of robotics, you know, manipulation, dexterity of the human hand. Yes, that is a critical thing, as well. That is like, okay, how could you actually get robots to effectively manipulate objects in a way that is even remotely similar, or at a certain level of performance, as humans can. It’s a mix of can you build robot hands and fingers that have the kinds of dexterity as a human, plus can you build the software that allows robots to understand, process both visual and tactile sensing capabilities, so the robot can actually know and get a sense of what it’s doing with its hand. Those two things need to come together really well, and there are advancements in that space. It’s very easy to build a robot hand that picks up one thing. It’s hard to build a robot hand that picks up anything.
I understand another part of your work is getting kids interested in science, engineering, robotics, especially kids from Africa. Why is that a particular passion of yours?
A couple of reasons. The first is that I strongly believe in the power of technology in creating a better world. I believe that technology can serve as a tool to create value, real-life value. And so, my passion, originally being from Nigeria and continuing to have family there and deep ties and connections, I’m very passionate about seeing the future where the Global South is not just a consumer of technology, but helping shape it and helping define how it shapes its own society. And we only do that through educating people to know how to use these tools to create value in their communities.
Parts of this interview have been lightly edited for length and clarity.
Will you support The World?
The story you just read is not locked behind a paywall because listeners and readers like you generously support our nonprofit newsroom. Now more than ever, we need your help to support our global reporting work and power the future of The World. Can we count on you?