Orca: China's AI world model matches specialized robotics systems without action labels

The Beijing Academy of Artificial Intelligence developed Orca, an AI model predicting abstract world states from 125,000 hours of video without action labels.

Published 2j1 sourceNotable

Lire en français

≈ 25sSpeed

The fact

Orca matches the performance of the specialized π0.5 robotics system on five tasks, despite lacking labeled data.

This method could help address the chronic data shortage in robotics.

Explore this topic

Beijing Academy of Artificial Intelligence #world model #robotique #IA générative #apprentissage non supervisé #vidéo

Cross-sources — 1 source

Dev.to

Auto-synthesis from 1 media source · identified on July 11, 2026

← Back to home

The fact

Read more

Nadella criticizes OpenAI and Anthropic over data policies and distillation bans

Over 200 experts warn of AI's imminent economic impact

Rich Sutton (Turing Award) founds Oak Lab to develop self-learning AI agents

Anthropic extends free Claude Fable 5 access amid AI pricing war

Soofi S: Open 30B AI model excelling in German and English released

Google SensorFM: AI model turns wearable data into a health intelligence layer