A 'Fair Use' Distributed Language Model

This is going to be a four part series of posts that contribute towards this idea that personal language models can exist and contribute to a larger model from our machines. Circumventing the need for data-centers, while generating citations, and providing personalized experiences that ambiently assist our lives day-to-day. Sounds too good to be true... as if it just magically solves most of the problems of the status quo.

Maybe it does, maybe it does not, maybe its naive, but someone has to try and at least fail fast enough for others to move onto the actual solution that does reach the equilibrium we desire. I am certain about 1 thing, that an equilibrium is indeed coming, it's just a question of where, when, & how.

This will be a 4 part series.

1. The Path Towards 'Fair Use' in Generative A.I.
2. Ablations: Mamba, Falcon H1, & more — coming soon
3. A Framework Around Fair Use — coming soon
4. A Protocol to Decentralize Inference — coming soon

The first of this series will be a high level overview of Braid that is the starting architecture of a Mamba language model that cites at the token level preparing nodes for a distributed language model paradigm. It is currently being tested for completion prior to Q/A capabilities, grounding our understanding around fair use behind a practice that can be defended in court.

So... let's begin. Here's The Path Towards 'Fair Use' in Generative A.I.