Autonet
Infrastructure-as-a-Service Economy for Distributed Deep-Learning Applications
Demo
About
The problem it solves Commercial AI models cost a lot to train and, as a result of that, the revenue generated by them is concentrated among a few companies. With an increasing number of tasks being automated, several professional categories run a real risk of becoming economically irrelevant. Autonet provides a technological and logical framework for crowdsourcing the development of large-scale AI models, while also ensuring a trustless distribution of revenue among stakeholders. It effectively decentralizes the ownership of AI services.
How it's made
Challenges we ran into The logic by which to validate the training work done by individual nodes needs to be added to the core map/reduce mechanism which is natively present in TensorFlow 2.