Alexander Koch @alexkoch_ai · 22 May 2025
General-purpose robots will have to learn from trillions of trials. We have developed a new world model that is optimized for reinforcement learning in imagination. We scale it up to 1B parameters and train it on a large dataset of human and multi-camera robot data. https://t.co/KnvcEpCSOI
34 737Views
189Likes
29Reposts
20Replies
4Quotes
88Bookmarks
Is that a lot?
12.6×vs 10K–100K median2 751 views is typical
3.0× audienceReachviews ÷ followers
0.70%Engagement rateof viewers reacted