I did a prototype implementation of a “network of ML networks” - an internet-like protocol for federated learning where nodes can discover, join, and migrate between different learning groups based on performance metrics (Repo: https://github.com/bluebbberry/MyceliumNetServer). It’s build on Flower AI.
Want do you think of this? It could be used to build a Napster/BitTorrent-like app on this to collaboratively train and share arbitrary machine learning models with other people while keeping data private and only sharing gradients instead of whole models to save bandwidth. Would this be a good counter-weight for big AI companies or actually make things worse?
Would love to hear your opinion ;)
Can you help me understand the use case?
Is it intended to be a variant of Flower AI that can be used in an adversarial environment?
Since downloading copyright material is legal for training apparently, I’d be glad to help train a privacy respecting distributed LLM. 😉
/s of course. But you have a very interesting idea!
Btw, would it be legal to use a torrent client that uses an LLM to make up the outgoing packets so that you aren’t sending copyrighted material? ;)
What would a use case look like?
I assume that the latency will make it impractical to train something that’s LLM-sized. But even for something small, wouldn’t a data center be more efficient?
