Nathan Lepora

It's the hands that are the critical thing.

They're the things that come into interaction with the environment and do the work.

When I first started in the field maybe 15 years ago, it seemed like this kind of distant dream, and you couldn't really see how it was going to happen in the short term.

But things are moving so fast in robotics now with AI, new hardware, new fabrication methods.

We're getting close.

It's kind of becoming clear how this problem will be solved, but we're still not there to actually get it fully to the human levels of dexterity.

We can make machines do very repetitive things, you know, and that's how a lot of industry works.

You know, you've got a machine that's kind of tailor-made for a specific job, you know, screwing caps on bottles.

Whereas our hand, it's like a general purpose machine.

The most obvious way is to put a motor in every joint of the hand.

That's not how the human hand works.

We have, you know, tendons, muscles in our forearms.

We coordinate within our body how the hand moves.

And it's a really challenging design problem to crack that problem in terms of the hardware and the AI and then to simplify it down so that it's, you know, can be made more easily and is more affordable.

It's really hard.

With robots, there's multiple things that need to be got right before they become really practically useful.

And so like three things would be, you know, the body, design of the body, how it's actuated, you know, if you like the design of the robot hands, the brain, you need the AI, but also you need an AI that's appropriate for controlling robot hands and robots and dexterity.

And that can be quite different.

It's kind of a form of embodied AI.

And then you need the sensing as well.