Ken Goldberg
speaker
221 appearances
3 recordings
3 series
first heard Jan 2025
last heard 7 May
Ken Goldberg’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in May 2026 with 1.
Appearances
Yeah, robot, which means worker or forced worker in Czechoslovakia.
And it's programmable. So then you can make a whole bunch of them over and over again. And that's still used, by the way, very heavily. And then what's the next big leap forward? There's a lot of fear around that time that robots are going to take over. In the newspaper, there's all these articles that they're going to do all the work, but that doesn't happen.
And the first robotics conference is in 1984. Then there's a big research field that starts to grow around robotics. But then it started really taking off in factories, especially automotive. The big thing that it's used for is welding. And spray painting. The welding's awesome. Yeah, the welding's fun because you get those sparks. It's like little pinches just come in, boom, boom, boom, right?
And welding sheet metal is very hard to do. For humans. You burn right through it so easy. Right, so it's very delicate, but then you're just basically doing the same thing over and over again. So it's repetitive, and that is very good for factories and also some of the assembly, putting together various devices and appliances and things like that. That's a big wave.
And that's also happening in Japan and other places. So that's growing, the industrial robotics. And the biggest breakthrough is now in 2012 in the breakthrough of deep learning and AI.
Let me back up a little bit, which is that when I did my PhD, I was interested in this incredibly simple problem of just trying to pick up objects just to grasp. It's something everybody does. Babies do it. I was actually clumsy as a kid. I later thought maybe that's why I wanted to study that. But it's still an open problem.
Robots to pick things up.
What is it? Oh, yeah, yeah. Moravec. There we go. There we go. Okay. So Moravec was actually at CMU when I was there. He was this very eccentric guy. And he wrote this book and he was saying it's a paradox that what's easy for robots, like lifting a heavy car, is hard for humans. But what's easy for humans, like stacking some blocks, that's hard for robots. And that's still true today.
Yeah. It's very counterintuitive because humans, it's so easy. But we've sort of evolved over millions of years, just like dogs and crows. Crows are able to pick up things amazingly. They can put coins in slots and they can do eight-step problems.
For sure.
Exactly. Robots, there's a lot of uncertainty in the environment. And even if you tell a robot to go to one specific spot, because of the motors and levers and gears that are in it, it won't go to that exact spot. You want it to put its jaws or something at a specific point to grasp this cup, it'll be slightly off. And that will cause it to miss, drop the object.
Exactly, and then the other is sensing. So we can take a high-resolution picture of an environment like this room, but there's no sensor that can give me the depth, the three-dimensional part of this room. What if you used a 3D camera, so you had bilateral? There's errors. There's little noise in those things.
If you look at the result of that, there'll be a depth map, which is like a 3D camera image, but you'll see there'll be lots of noise and imprecision and mistakes in those. And those are inevitable. There's no camera that really works reliably for 3D.
That's such a good way to put it. And that's exactly right. And so if you look at that history, hundreds of millions of years of evolution to see to mobility and being able to manipulate just the opposable thumb and all of those things. And so all these other things like math is relatively very recent.
That I think is helpful for people to understand why we've made all this progress in these, quote, hard problems like playing chess and go.
But we really haven't made much progress in just being able to like clear the coffee table.
You're right. The muscles in the human body, there are hundreds of muscles and bones, and they pull in all these nuanced ways. And we have the skin that's very complex. What's amazing is how much we don't know about human biology. We don't understand how touch works. Touch is incredibly complex. Like we can feel things that are so small. They're much smaller than human hair.
We can perceive up to very complex vibrations and other things. You add in temperature we can feel.
When I was in undergrad, I tried to use electricity to do that, and it failed. It did not work. Okay. But what people are using now is light. Okay. And they transform the touch into light. And so imagine that you have a little camera in your fingertip. It's looking inside at a pad from the bottom. And so when the pad gets indented, you see the pattern of what it's touching. Okay.
So there's like a membrane, and above that, the membrane's being observed? That's exactly what it is. But what happens is the membrane gets rubbed off, or over time, those sensors get deformed, and so it doesn't solve the problem. It's just the latest method that we're trying. Great.
Showing 101–120 of 221 · page 6 of 12
← Previous
Next →