An update for the Byron AI backers. We now have a sufficiently powerful box up and running with an open source LLM on it. We’re now ready to start the first stage of the actual training. The style module, which is being developed on a separate box, has made some actual advancements, and importantly, we’ve learned a lot about what authors are already in the data, which are partially in the data, and which are not.
This information is significant for how we train with them for the style module, although it’s of little significance for the genre module.
This is just the first step, but it is arguably the most important one. As it stands, we now have a moderately powerful AI running under our control without any external guardrails imposed by the mainstream AI corporations.
It’s also clear now that there will be other useful applications as well, but for the time being, we will stay focused on the mission.