Skip to content
MIT Technology Review·

🤖Danijar Hafner's New Startup Aims to Revolutionize Robot Navigation

The future of robots in human spaces just got a lot closer

TL;DR

Danijar Hafner's stealth startup in San Francisco is working on humanoids that can navigate unfamiliar environments using advanced AI. This could be a game-changer for integrating robots into human spaces.

Danijar Hafner's new startup is developing humanoids that can navigate environments they've never encountered before, thanks to advanced AI techniques. This breakthrough could finally make robots a practical reality in human spaces. Hafner's approach uses model-based reinforcement learning, where agents learn to predict and react to new situations without real-world trial and error. His work on PlaNet and Dreamer 2 has already shown this can work in video games, and now he's applying it to real-world robots. This could be a major step forward for robotics, but it's still early days.

Danijar Hafner's New Startup Aims to Revolutionize Robot Navigation — MIT Technology Review

Key Points

1

Danijar Hafner's startup, founded in fall 2025, is in stealth mode in San Francisco's SoMa district.

2

The startup's office is sparse except for humanoids of various shapes and sizes hanging from racks.

3

Hafner's technique uses model-based reinforcement learning, allowing agents to navigate unfamiliar environments.

4

His work on PlaNet and Dreamer 2 has shown this can work in video games, hitting human-level performance.

5

The startup aims to make robots practical in human spaces, a major step forward for robotics.

Why It Matters

Danijar Hafner's work on model-based reinforcement learning could make robots practical in human spaces. If successful, this could transform industries like manufacturing, healthcare, and service robotics. However, the technology is still in its early stages, and real-world application remains to be seen.

Danijar HafnerroboticsAIreinforcement learningnavigation

Frequently Asked Questions

Why does this matter?

Danijar Hafner's work on model-based reinforcement learning could make robots practical in human spaces. If successful, this could transform industries like manufacturing, healthcare, and service robotics. However, the technology is still in its early stages, and real-world application remains to be seen.

What happened?

Danijar Hafner's stealth startup in San Francisco is working on humanoids that can navigate unfamiliar environments using advanced AI. This could be a game-changer for integrating robots into human spaces.

Comments

Subscribe to join the conversation...

Be the first to comment

Enjoyed this article?

Get it daily. 7am. Free. Reads in 5 minutes.

Join 3,464 builders reading daily.

Also get