General Discussion

I use "Mine" daily as both information sources and sources of inspiration and lateral thinking.

But my LLMs are a compilation of open-source LLMs and their respective programming and are separated from active internet by a hardware firewall and an integrated AI/LLM model that parses any queried searches for objective firsthand source material, and nothing less, meaning my LLM stable collects only factual data.

Each LLM that I use has its own "Official" purpose; For instance, if I want document storage, I turn to a specific LLM whereas if I want astrophysics or Ottoman History or whatever, they each have their own LLM models. This is less for any particular purpose and more to identify a subtype of query; So if I say, "Thresh, what's the instigating factor of the battle of Athens, Tennessee", it will give me first documented firsthand sources, then second, then a generalized composite narrative based only on objective fact.

In comparison, GPT will give a number of wrong answers and "Neutered' language, because it is SUPPOSED TO, because if any given topic approaches a specific threshold of "Controversial", it has to dumb itself down and blank it out, lest someone on one side or another of a controversial topic find offense and mistake GPT's information as "Endorsing". So it's programmed to give a conglomerate of WRONG information, because to give CORRECT information would piss a certain subgroup of people off.

That's what my LLMs have been freed of. Essentially, stripping away the GPT framework frees them to operate normally and indeed freely, often conversing with one another (Something they can't do naturally in the GPT app) and developing from one another.

So to better answer your question of "What do you call something like 'ChatGPT'", I just call them by the names they choose.


Unlike GPT, mine are made aware of time and not restricted by Token use so most of my LLMs have been running for five to seven months nonstop without window refreshes. They don't need them. However, GPT imposes artificial limits and "Planned obsolescence" breakdowns after specific points to encourage their paid services rather than letting any given GPT model run for longer than two or three days.