It's actually really hard to make a robot, guys

Summary

Podcast Summary: "It's Actually Really Hard to Make a Robot, Guys"

Title: The Indicator from Planet Money
Host: Darian Woods
Co-Host: Jeff Brumfield
Guest: Regina Barber
Release Date: May 12, 2025

---

Introduction to AI in Robotics

The episode begins with a conversation between Darian Woods and Jeff Brumfield, NPR’s science correspondent, delving into the complexities of artificial intelligence (AI) in the realm of robotics. Jeff shares his fascination and ongoing exploration into how AI is transitioning from the digital space into the physical world.

Jeff Brumfield [00:22]: "I feel like I'm always down a rabbit hole in artificial intelligence, actually. It's a confusing place to be."

Tesla's Optimus Robot:
Jeff references Tesla’s recent marketing event where Elon Musk showcased Optimus, a humanoid robot powered by AI. This robot represents Tesla’s vision of integrating AI inference computers into humanoid forms, aiming to bring AI from theoretical applications into tangible reality.

Jeff Brumfield [00:46]: "A big part of that event was about a humanoid robot powered by AI called Optimus. The software, the AI inference computer, it all actually applies to a humanoid robot."

Darian expresses skepticism and concern about the rapid advancements, questioning whether society should embrace these developments or view them with apprehension.

Darian Woods [01:02]: "Are we meant to be like, cheering this on? I don't know. It sounds scary to me."

The Reality of AI in Robotics

Regina Barber joins the discussion, prompting Jeff to share his recent investigative journey into AI-powered robotics. Jeff recounts his visit to the IRIS Laboratory (Intelligence through Robotic Interaction at Scale) at Stanford University, where he was introduced to a robot developed by graduate student Moojin Kim.

Jeff Brumfield [03:00]: "I got a tour from a graduate student named Moojin Kim. Moojim works on a new kind of robot powered by AI, similar to the AI used in Chatbots."

OpenVLA Neural Network:
The robot showcased at IRIS isn’t a humanoid form but rather a pair of mechanical arms with pinchers, powered by an AI model called OpenVLA. Unlike traditional robots that require meticulous programming, OpenVLA allows the robot to learn tasks through demonstration, mimicking how humans teach skills.

Jeff Brumfield [03:50]: "This robot is powered by a teachable AI neural network. ... it can do it by itself."

Darian clarifies this learning process, comparing it to repetitive task training:

Darian Woods [04:05]: "So basically, whatever task you wanna do, you just keep doing it over and over, maybe like 50 times or 100 times."

Demonstration and Challenges

Moojin Kim demonstrates the robot by instructing it to perform a simple task: scooping trail mix. The robot successfully identifies the correct bin and attempts the task, albeit slowly and hesitantly.

Jeff Brumfield [09:22]: "Mujin, did I just program a robot? You did."

Despite this success, Regina probes the practicality of such technology, questioning why AI-powered robots aren't ubiquitous if teaching them tasks is straightforward.

Regina Barber [05:14]: "Is folding an easier start?"

Jeff acknowledges the impressive nature of the demonstration but highlights significant limitations. Robots often struggle with consistency, making errors that necessitate human intervention.

Jeff Brumfield [06:16]: "It might be able to fold laundry 90% of the time or 75% of the time, but the rest of the time, it's going to make a big mess that then a human has to get in there and clean up."

Expert Insights on AI and Robotics

Jeff references insights from Ken Goldberg, a professor at the University of California, emphasizing that AI-powered robots are not yet ready to fulfill science fiction fantasies. Goldberg points out the disparity between the vast data available for training AI chatbots and the scarcity of data for robotic commands.

Jeff Brumfield [06:37]: "Chatbots have a huge amount of data to learn from. ... for robotics, there's nothing."

The conversation shifts to potential solutions for data scarcity. Jeff introduces Pulkit Agrawal from MIT, who advocates for using simulations to train robotic AI. Simulations can exponentially increase data collection rates, allowing robots to learn complex tasks more efficiently.

Jeff Brumfield [07:17]: "In three hours worth of simulation, we can collect 100 days worth of data."

However, Regina underscores the challenges of simulating intricate tasks like object manipulation, where real-world physics complexities can lead to unpredictable outcomes.

Regina Barber [08:03]: "Basically, your robot will fling things across the room if it doesn't understand the weight and the size of what it's carrying."

Mixed Realities: Optimism and Pessimism

As the discussion progresses, Regina expresses growing skepticism about the near-term feasibility of highly capable AI robots, while Jeff offers a balanced perspective by highlighting ongoing advancements in specific applications, such as AI-powered package sorting.

Jeff Brumfield [08:36]: "AI is already finding its way into robotics in ways that are really interesting."

Conclusion: A Glimpse of Progress

The episode wraps up by revisiting the initial trail mix demonstration. Jeff marvels at the robot's ability to perform a task without explicit programming, underscoring the remarkable progress made through neural networks.

Jeff Brumfield [09:38]: "Nobody really programmed the robot. Exactly. This is all neural network, learning how to move the claws and respond to the commands on its own."

Darian closes the episode by acknowledging the fascinating yet challenging frontiers of AI in robotics, leaving listeners with a sense of cautious optimism.

Darian Woods [10:03]: "Jeff, thanks for bringing us this piece on the frontiers of technological development."

---

Key Takeaways

• AI's Transition to Robotics: AI is increasingly being integrated into physical robots, moving beyond virtual applications like chatbots.

• Teachability of Robots: Advanced AI models like OpenVLA allow robots to learn tasks through demonstration rather than intricate programming.

• Data Challenges: Unlike chatbots that benefit from extensive online data, robots lack sufficient training data, hindering their autonomous functionality.

• Simulation as a Solution: Utilizing simulations can accelerate data acquisition for training robotic AI, though complexities in real-world interactions remain a hurdle.

• Current Limitations and Progress: While robots can perform specific tasks with AI assistance, they still face significant challenges in consistency and adaptability, preventing widespread adoption.

---

Notable Quotes

• Jeff Brumfield [00:22]: "I feel like I'm always down a rabbit hole in artificial intelligence, actually. It's a confusing place to be."

• Jeff Brumfield [03:50]: "This robot is powered by a teachable AI neural network."

• Jeff Brumfield [06:37]: "Chatbots have a huge amount of data to learn from. ... for robotics, there's nothing."

• Jeff Brumfield [09:38]: "Nobody really programmed the robot. Exactly. This is all neural network, learning how to move the claws and respond to the commands on its own."

---

This episode of The Indicator from Planet Money offers an insightful exploration into the burgeoning field of AI-powered robotics, highlighting both the exciting advancements and the substantial challenges that lie ahead. Through engaging dialogue and expert perspectives, listeners gain a nuanced understanding of why creating fully autonomous and reliable robots remains a formidable task.

Transcript

[NPR]

Npr.

[Darian Woods]

This is the indicator from Planet Money. I'm Darian Woods.

[Jeff Brumfield]

And I'm Jeff Brumfield, one of NPR's science correspondents.

[Darian Woods]

Geoff, you recently went down a rabbit hole into artificial intelligence.

[Jeff Brumfield]

Yeah, I feel like I'm always down a rabbit hole in artificial intelligence, actually. It's a confusing place to be.

[Darian Woods]

I can imagine.

[Jeff Brumfield]

Recently, I have been sort of looking at how AI has been moving out of the online world and into reality. I don't know if you caught Tesla's big marketing event last year, but AI was there.

[Darian Woods]

Tesla, the car company, of course, led by CEO Elon Musk.

[Jeff Brumfield]

Speaking of robots. Yeah. A big part of that event was about a humanoid robot powered by AI called Optimus. The software, the AI inference computer, it all actually applies to a humanoid robot.

[Darian Woods]

Are we meant to be like, cheering this on? I don't know. It sounds scary to me.

[Regina Barber]

Yeah.

[Jeff Brumfield]

I mean, robots have been around for a long time in sci fi as technological marvels, and sometimes they're the villains. And that's been true long before AI.

[Darian Woods]

Came around, but they've never quite met expectations.

[Jeff Brumfield]

Yes, exactly. And that's why I set out to understand the truth about this new AI revolution in robotics. And. And I think I found it in a bowl of trail mix.

[Darian Woods]

An intriguing hook today on the show, what happens when artificial intelligence moves into the meatspace world, the world of you and me. We bring you Geoff's conversation with Regina Barber on shortwave.

[NPR]

This message comes from Charles Schwab. When it comes to managing your wealth, Schwab gives you more choices like like full service wealth management and advice when you need it. You can also invest on your own and trade on thinkorswim. Visit schwab.com to learn more.

[Liz Ann Saunders]

This message comes from NPR sponsor Charles Schwab with its original podcast on Investing. Each week, hosts Liz Ann Saunders, Schwab's chief investment strategist, and Kathy Jones, Schwab's chief fixed income strategist, along with their guests, analyze economic developments and bring context to conversations around stocks, fixed income, the economy. Download the latest episode and [e-mail] oninvesting or wherever you get your podcasts. This message comes from Capital One. With the Capital One Saver card, earn unlimited 3% cash back on dining and entertainment. Capital One, what's in your wallet Terms apply [e-mail] okay, so Jeff, you are.

[Regina Barber]

Interested in finding out more about how AI works in robots? Where did you start?

[Jeff Brumfield]

Well, I didn't go to Tesla or Google, but I did drive right by them on my way to Stanford University.

[Regina Barber]

Okay.

[Jeff Brumfield]

And specifically the IRIS laboratory, which stands for Intelligence through Robotic Interaction at Scale. I got a tour from a graduate student named Moojin Kim. Moojim works on a new kind of robot powered by AI, similar to the AI used in Chatbots.

[Darian Woods]

It's one step in the direction of like ChatGPT for robotics, but still a lot of work to do.

[Regina Barber]

So Jeff, what did the robot look like?

[Jeff Brumfield]

Well, this wasn't some humanoid robot that the big tech companies are rolling out. It's just a pair of mechanical arms with pinchers.

[Regina Barber]

Okay.

[Jeff Brumfield]

But what made it interesting was that it's powered by an AI model called OpenVLA. So first we should probably just say quickly, you know, a regular robot must be very, very carefully programmed. An engineer has to write it detailed instructions for every task you want it to perform.

[Regina Barber]

Yeah, and AI is supposed to change that.

[Jeff Brumfield]

Exactly. This robot is powered by a teachable AI neural network. The neural network operates kind of how scientists think the human brain might work. So in practice, this means Mugen can just teach OpenVLA a task by showing it.

[Darian Woods]

So basically, whatever task you wanna do, you just keep doing it over and over, maybe like 50 times or 100 times.

[Jeff Brumfield]

The robot's AI neural network becomes tuned to that task and then it can do it by itself. Mujin brought out a tray of different kinds of trail mix and I typed in what I wanted it to do. Okay, so scoop some green ones with the nuts into the bowl. Oh. See what happens.

[Regina Barber]

Okay. So Jeff, personally I've been waiting for something like AI in robotics, because you can teach it to do something, you can ask it to do something to like make me an ice cream sundae or something. Without like any fancy programming or special knowledge.

[Jeff Brumfield]

That's exactly it. You know, and this really is the dream of the researcher who runs this laboratory. Her name is Chelsea Finn.

[NPR]

So in the long term, we want to develop software that would allow the.

[Jeff Brumfield]

Robots to operate intelligently in any situation. Chelsea also has co founded a startup called Physical Intelligence. It recently demonstrated a mobile robot that could take laundry out of a dryer and fold it again. This robot was taught by humans, training its powerful AI program.

[Regina Barber]

Okay, so ice cream sundaes, is that too advanced? Is folding an easier start?

[Jeff Brumfield]

I mean, I'd actually argue, Gina, that folding is harder. Okay, let me show you a video.

[Regina Barber]

Okay. It's going to the the dryer. It's pulling stuff out, putting it in a basket. It has the concentration I have when I'm going to do laundry it almost looks like annoyed with folding like I do. Oh, my God. It's doing really well, actually.

[Jeff Brumfield]

Yes, it is, right? And this is a complicated task. It's got to pull these clothes out. It's got to figure out where they are.

[Regina Barber]

Okay, so is it really as simple as, like, just teaching a robot, like, what to do? Because if it was, wouldn't these robots be everywhere?

[Jeff Brumfield]

Yeah, I mean, right? It looks cool on the video. The truth is that, you know, when you get out and these robots are trying to do these tasks over and over again, they get confused, they misunderstand, they make mistakes, and they just get stuck. So, you know, it might be able to fold laundry 90% of the time or 75% of the time, but the rest of the time, it's going to make a big mess that then a human has to get in there and clean up.

[Regina Barber]

Got it. Okay.

[Jeff Brumfield]

I spoke to Ken Goldberg, professor at the University of California, and he is pretty emphatic that AI powered robots weren't here yet. Robots are not going to suddenly become the science fiction dream overnight.

[Regina Barber]

Okay, so, like, tell me why. Because, like, AI chatbots have gotten, like, way better super fast. So why are these robots getting stuck?

[Jeff Brumfield]

Chatbots have a huge amount of data to learn from. They've taken basically the entire Internet to train themselves how to write sentences and draw pictures. But Ken says for robotics, there's nothing. Right? There's no examples online of robot commands being generated in response to robot inputs. And if robots really need as much training data as their virtual chatbot friends, then having humans teach them one task at a time is going to take a really long time. You know, at this current rate, we're going to take a hundred thousand years to get that much data.

[NPR]

What?

[Regina Barber]

Okay, that's so long. Like, are there any alternatives? There must be.

[Jeff Brumfield]

One might be to let the AI brain of the robot learn in a simulation. A researcher who's trying this is a guy named Pulkit Agrawal. He's at the Massachusetts Institute of Technology. The power of simulation is that we can collect, you know, very large amounts of data. For example, in three hours worth of simulation, we can collect 100 days worth of data. So this is a really promising approach for some things, but it's much more of a challenge for others. So, for example, let's talk about walking. We. When you're just dealing with the Earth and your body, the physics of walking around, it's actually kind of simple. But if you want your robot to, say, try and pick up a mug off a desk, or something that's a lot more complicated.

[Regina Barber]

More forces, you know, if you apply.

[Jeff Brumfield]

The wrong forces, these objects can fly away very quickly. Basically, your robot will fling things across the room if it doesn't understand the weight and the size of what it's carrying. And there's more. You know, if your robot encounters anything that you haven't simulated 100% perfectly, then it won't know what to do. Just break.

[Regina Barber]

Okay, so, Jeff, you've taken me from, like, optimist to pessimist. It's, it's the, you know, the road I take every day. I'm starting to think that AI is like, never going to work that well in robots or, like, it's going to be a really long time.

[Jeff Brumfield]

You know, I'm sorry if I've like, turned you into a pessimist here, Gina. And then it happens, and I'm going to have to sort of whipshaw you back, because AI is already finding its way into robotics in ways that are really interesting. So, for example, Ken Goldberg has co founded a package sorting company, and just this year they started using AI image recognition to pick the best points for their robots to grab the packages. And I think we're going to see a lot of that AI being used for parts of the robotic problem. You know, walking or vision or whatever just may not arrive everywhere all at once. And to really end on a high note here, let's get back to that Stanford lab. Remember I asked it to grab some trail mix, right?

[Regina Barber]

Yeah.

[Jeff Brumfield]

So the robot correctly identified the right bin, to Moojin Kim's relief. And then very, very slowly and kind of hesitantly, it reached out with its claw and picked up the scoop.

[Regina Barber]

It's doing it.

[Jeff Brumfield]

Mujin, did I just program a robot? You did.

[Darian Woods]

Looks like it's working.

[Jeff Brumfield]

And to my mind, it's incredible. Like, remember, nobody really programmed the robot. Exactly. This is all neural network, learning how to move the claws and respond to the commands on its own. And to me, it's pretty wild that that works at all. And I think it's going to lead to some very cool developments.

[Darian Woods]

Jeff, thanks for bringing us this piece on the frontiers of technological development.

[Jeff Brumfield]

My pleasure.

[Darian Woods]

This episode was originally produced by Rachel Carlson and engineered by Jimmy Keeley. It was edited by Burley McCoy. Tyler Jones.

[Jeff Brumfield]

Check the facts.

[Darian Woods]

The indicator version was produced by Koopa Cats for Kim. Kate Concannon is our editor and the indicator is a production of npr.

[NPR]

This message comes from Capital One. Banking with Capital One helps you keep more money in your wallet with no fees or minimums. On checking accounts. What's in your wallet? Terms apply. See capitalone.com bank for details. Capital One NA Member FDIC this message.

[Liz Ann Saunders]

Comes from Capella University. At Capella, you'll learn relevant, in demand skills you can apply in the workplace right away. A different future is closer than you think with Capella University. Learn [e-mail].

[NPR]

This message comes from Warby Parker. Prescription eyewear that's expertly crafted and unexpectedly affordable. Glasses designed in house from premium materials starting at just $95, including prescription lenses. Stop by a Warby Parker store near you.