Skip to content
No PriorsNo Priors

No Priors Ep. 48 | With Covariant CEO Peter Chen

Building adaptive AI models that can learn and complete tasks in the physical world requires precision but these AI robots could completely change manufacturing and logistics processes. Peter Chen, the co-founder and CEO of Covariant, leads the team that is building robots that will increase manufacturing efficiency, safety, and create warehouses of the future. Today on No Priors, Peter joins Sarah to talk about how the Covariant team is developing multimodal models that have precise grounding and understanding so they can adapt to solve problems in the physical world. They also discuss how they plan their roadmap at Covariant, what could be next for the company, and what use case will bring us to the Chat-GPT moment for AI robots. 00:00 Peter Chen Background 00:58 How robotics AI will drive AI forward 03:00 Moving from research to a commercial company 05:46 The argument for building incrementally 08:13 Manufacturing robotics today 12:21 Put wall use case 15:45 What’s next for Covariant Brain 18:42 Covariant’s customers 19:50 Grounding concepts in Ai 25:47 How scaling laws apply to Covariant 29:21 Covariant’s driving thesis 32:54 the Chat-GPT moment for robotics 35:12 Manufacturing center of the future 37:02 Safety in AI robotics

Sarah GuohostPeter Chenguest
Jan 23, 202440mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Covariant’s Peter Chen Builds Data-Driven Foundation Models For Real-World Robots

  1. Peter Chen, CEO and co‑founder of Covariant, explains how the company is building a foundation model for robotics by deploying AI-powered robots in warehouses and logistics centers.
  2. He contrasts today’s “dumb,” pre-programmed industrial robots with the next wave of intelligent, adaptive systems that can handle diverse objects and environments with high reliability.
  3. Chen argues that success in robotics will hinge on collecting massive amounts of real-world, embodied interaction data, not just simulation or internet-scale text and images.
  4. Looking ahead, he expects industrial manipulation to lead the way, with humanoids and consumer robots following once hardware economics and safety constraints are met.

IDEAS WORTH REMEMBERING

5 ideas

Real-world data is the core strategic asset for robotic intelligence.

Chen’s central bet is that the winner in robotics will be whoever gathers the most high-quality, embodied interaction data from robots operating in production environments, because internet and simulated data miss crucial physical nuances.

Today’s robots are precise but fundamentally “dumb” and inflexible.

Over 99% of industrial robots are pre-programmed to repeat fixed motions and cannot adapt to changing items, layouts, or tasks, leaving huge classes of real-world problems—like e-commerce fulfillment—unsolved.

Foundation models for robotics need both generality and high reliability.

A “ChatGPT moment” for robots requires not just broad task coverage but failure rates low enough to avoid costly or dangerous physical errors, which demands dense data coverage and rigorous real-world training.

Warehouse manipulation is a high-value proving ground for robotic AI.

Logistics centers face booming e-commerce demand, labor shortages, and >100% annual turnover, making them ideal environments to deploy robots that learn from diverse manipulation tasks and generate valuable training data.

Grounding in the physical world requires precision beyond internet multimodal data.

Image–text pairs teach high-level concepts, but manipulation requires sub-centimeter understanding of object shapes, contact forces, and dynamics, plus action–outcome data that current online datasets largely lack.

WORDS WORTH SAVING

5 quotes

“When we started Covariant, there was no AI that was good enough to make robots do useful things commercially.”

— Peter Chen

“99+% of the robots that are deployed in the world are dumb robots… doing the same thing again and again.”

— Peter Chen

“We believe the future of robotics would be built by whoever has most robotics data.”

— Peter Chen

“You cannot just go straight to full general physical AGI… you have to build something that is valuable that you can ship to customers, and from that process you get more data.”

— Peter Chen

“The bar for the ChatGPT moment for robotics is high… you need to solve the generality, but you need to solve it with a high level of reliability.”

— Peter Chen

Peter Chen’s research background and motivation for starting CovariantLimitations of current industrial robots and the need for intelligenceCovariant’s strategy: foundation models for robotics and data collection via real deploymentsWarehouse and logistics manipulation use cases (e.g., put walls, pick-and-pack)Grounding, multimodal understanding, and what’s missing from internet-scale dataScaling laws, real-world data vs. simulation, and emergent capabilitiesFuture of robotic applications, humanoids, industrial vs. consumer robots, and safety

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome