@martin Absolutely agree #opensource #distributedcomputing #inspectableAI models should be a developmental goal. Perhaps it could be built on or designed like https://en.wikipedia.org/wiki/Berkeley_Open_Infrastructure_for_Network_Computing Privacy settings would allow users to decide what to share with the model, and every user decides how much of their "spare" computing power to contribute to training the model. Even people whose only device is a phone can participate.