AI Summary
5 min readIn a recent experiment, tech journalist Joanna Stern strapped an iPhone to her head and filmed herself folding laundry for hours. She was paid about $20 an hour for the footage, which went to a German AI company called Micro AGI. The company uses such videos to train robots—not by showing them what laundry looks like, but by capturing the precise 3D motion of human hands performing tasks. This is one small piece of a much larger shift underway in artificial intelligence: moving beyond text-predicting large language models toward "world models" that can understand and act in the physical environment.
Why text-based AI isn't enough for the real world
The core argument of the episode is that the kind of AI powering ChatGPT and Claude—models that mathematically predict the next most likely word in a sequence—will not work for robots that need to navigate a messy, unpredictable physical world. As Stern explains, a self-driving car cannot rely on text training alone because the road environment changes constantly: a bicycle cuts in, a road closure sign appears, a pedestrian steps out. A home is even worse. "Someone moves something in the refrigerator. Somebody moves a plate. My kids have put their toys all over the ground." These are not problems of language; they are problems of spatial reasoning, object manipulation, and real-time adaptation.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Science Friday
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:02) **Why robots need a "world model," not just text-based AI** - Introduction to the episode's first topic: the limitations of large language models in the physical world.
- 2 (01:11) **Guest Introduction: Joanna Stern** - Tech journalist Joanna Stern is introduced to explain the emerging field of world models for robots.
- 3 (01:22) **Why text-based AI fails in the physical world** - The fundamental problem: the physical world is unpredictable and can't be learned from text or simple images alone.
- 4 (03:19) **Two main methods for training robots: Teleoperation vs. Video Data** - The debate over the best way to gather training data for physical-world AI.
- 5 (04:49) **Joanna's hands-on experiment with data collection** - Joanna tried being a paid data collector for Micro AGI, filming her own chores.
- 6 (06:09) **The current state of robot capability: A laundry-folding test** - A real-world example shows how far the technology still has to go.
- 7 (07:46) **Privacy concerns with home data collection** - The trade-off of sharing thousands of hours of home footage is a significant issue.
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
When you ask an LLM like ChatGPT or Claude a question, the model goes through its massive amount of training data and guesses the answer by mathematically predicting the word most likely to appear next in a sentence.
This model, experts say, will not work well for technology designed to navigate the physical world. Something like a robot that works in a warehouse will instead require a “world model” that can understand spatial surroundings, like the stuff we walk by or bang into.
But what is a world model, exactly? And how do you train AI to recognize what the real world looks like? Host Ira Flatow checks in with tech journalist Joanna Stern, who’s seen the early days of these models up close, even in her own home.
Then, we check in on the math world, where frontier AI models have made meaningful progress on decades-old problems. Mathematician Emily Riehl gives us the big picture on how significant these results actually are.
Guests:
Joanna Stern is a tech journalist who writes newsletters and creates videos for New Things Media.
Dr. Emily Riehl is a professor of mathematics at Johns Hopkins University.
Transcript will be available after the show airs on sciencefriday.com.
Subscribe to this podcast. Follow our show on Instagram, TikTok, Facebook, and Bluesky @scifri and sign up for our newsletters. Got a science question that’s keeping you up at night? Call us: 877-472-4374
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
More from this podcast
Science Friday →