Spaces:
Running
World Model Requirements for the Field environment.
Could I get some feedback on what models might fit into this system outside of tools for the sensors? If you need a reference for what I've been building lookup the ATAK suite. imagine that but as a real world assistant and advancement platform which is customized for each user.
I would like it to be a platform for people to deploy their own custom assistants but I'm not 100% how.
So the Field device foundations are established as:
Local Evidence ledger (Stores checkpoints, choices, tracks and as much sensor data as is reasonable for proof of work and future system improvement)
User Base Profile (Contributes to Users public profile, System state and contains preferences and is updated where needed)
Endeavor packages(Contains endeavor specific assets ie: tools, tasks, models etc. Activated through start, step and finish qualifiers)
Mapping and plotting software (with field events/objects scored on the fly to gauge and announce opportunities for advancing the users objectives.)
To be more specific I am curious if anyone has developed unique types of Model Architectures.
I have developed a system for a low resource, infinite context adaptive/learning ML system which should work but I'd rather avoid going down 50 different rabbit holes if I don't have to. I have a pretty good system for out of domain development but it hurts throwing billions of OpenAI tokens down rabbit holes and I'd rather work with an expert if there's one out there not yet beholden to the cube or poisoned by it's shade...
Thanks Meow