@JasonrShuman Generalization is the true frontier now because it is capital intensive and also because the architectures aren't quite there yet for human like systems
@JasonrShuman Data is getting cheaper as is. Oversupply + novel sources like human data, sim, UMI etc.
Policies are the real bottleneck atm but there are players (China, NVIDIA) who have an incentive to open-source policies as well. (Cosmos3 as an example)
Congrats to the team at @safe_sentinel_ 🚀
Great to see our dexterous hand being used in real-world experimentation. Every iteration like this moves embodied intelligence one step forward.
Sharing our recent breakthrough in dexterous manipulation. We are hiring across the board to build general purpose robots: policy learning, mechanical engineering, industrial design.
website: https://t.co/tM6SJY4iYz