An Open Source humanoid trained collaboratively by a community of builders.
AI technology has advanced enough to speculate that within a decade most people will have their own humanoid buddy. By some estimates humanoids will become $100 Trillion market (5B humanoids * $20,000 per unit).
Today's leading closed source humanoid is trained on 100,000 GPU farm with real world data collected from millions of cars labeled by able human drivers. This is an enormous scale of compute and data that is hard to compete with as a centrazlied entity. However it would be interesting to see if a decentralized approach might produce useful results over time. On the chance that proprietary humanoids ever go rogue, it would be nice to have open source alternatives.
zk0 is composed of several major building blocks:
- Generative AI:
- HuggingFace LeRobot for the Open Source 3D printed robot parts and end-to-end vision language action models.
- Federated Learning:
- Flower for collaborative training of AI models
- Zero Knowledge Proofs:
- EZKL for verification of contributed model checkpoints trained on local data.
Here is a complete example demonstrating federated learning with the LeRobot PushT dataset. Shows client-server architecture, data partitioning, and model update aggregation.
Following is the high level directory structure of the code repository. Jump in, try the example and explore. Contributors are welcome!
zk0
│
├── lerobot # clone of remote lerobot repo:
│ # https://github.com/huggingface/lerobot.git
│
├── federate # federated learning layer
│ │
│ └── lerobot_example/
│ # Federated Learning Example with Flower and LeRobot Diffusion PushT task
│
└── README.md # This README file