United States
United Kingdom
Canada
Australia
New Zealand
IrelandArt of the Problem's robot needs more than conversation
Brit Cruise's GrowBot experiment explores the gap between an AI that can interpret motion and a body that can recover from a push.
Art of the Problem
Story of Why this story? The channel's weekly summary: what went up this week, what is drawing views now, and where the channel stands among similar channels.
I Gave ChatGPT a Body
Things we can learn from the video
Press to jump
- 5:01A lightweight neural network learns motor control in simulation before running on GrowBot's onboard chip.
- 8:20Language models receive sensor information and generate higher-level responses and actions over Wi-Fi.
- 15:03Text memory preserves interactions, while a later consolidation cycle extracts behavioral lessons from the logs.
- 20:00Cruise explains how the cerebellum predicts movement to compensate for sensory delays.
- 22:53GrowBot's missing physical prediction architecture leaves it struggling with mimicry and recovery after a push.
- 23:24DayDreamer provides an example of a robot learning movement through real-world experience and prediction.
Why can GrowBot describe being tilted, yet struggle to recover when pushed over? That gap gives Brit Cruise's "I Gave ChatGPT a Body" its most interesting turn: language and physical competence are not the same thing.
Art of the Problem posted nothing new this week, but the May robotics film remains its busiest upload by a wide margin, with roughly 5.5 million views in total. Cruise trains a small neural network in simulation to make his two-legged creation walk, spin and balance, then connects language models to its controls. The robot interprets sensor readings, responds to spoken instructions and consolidates interaction logs through a process Cruise calls dreaming.
The stumbling block is physical prediction. Cruise turns to the cerebellum, explaining how anticipating movement helps people compensate for sensory delays. GrowBot's conversational abilities do not supply that missing machinery.
The later follow-up, "I Made Skynet?", takes the project in a different practical direction. Cruise says chip shortages disrupted his kit plans, before an old smartphone offered a substitute for the electronics he had been assembling. His wonderfully unglamorous prototype puts the phone in a pizza box with legs. That upload continues to draw a smaller audience, while "The Alien Signal That Looked Intelligent" is the channel's second-busiest video right now.
Across the week, Art of the Problem gained about 1,000 subscribers, although its current daily views are a little below usual. Among the nearby science channels, only Astrum Earth and Ziroth gained more weekly views. For Cruise, an upload-free week has left plenty of room for the robot's unfinished problem: making a thinking machine physically capable.
Published on Trari
Where next?
Is this your channel?For the channel's owners and team: a ready picture and text for your audience
Share the story with your audience. Everything is ready:
For an Instagram story or a WhatsApp status
For a community post or your channel's description
The videos drawing views on Art of the Problem right now
- Channel
- Art of the Problem
- Subscribers
- 395,000 ▲ 1,000 this week
- Channel views, last 7 days
- 215,936
- Videos published this week
- 0
The numbers
Here is how Art of the Problem's audience and older uploads moved this week.
💡 Test your knowledge
What becomes the robot's body in Cruise's smartphone-based follow-up?
In the description of "I Made Skynet?", Cruise says he put an old smartphone in a pizza box and added two legs.