Since it's possible to train your AI bot for whatever area of expertise you want, what would happen if someone took an AI and pointed it at all the Q posts, this site, ATS, BIN, conspiracy.win, John Birch, Bill Cooper, Alex Jones, David Icke, Project Camelot, on and on.
Put ALL the "conspiracy" stuff in there and see what happens.
An algorithm that sees all the ideas and facts and history and theory, maybe it could put it all together and get a Grand Unification Theory of sorts.
Edit-
I'm primarily thinking private or self hosted, not something someone else controls like ChatGPT which is already lobotomized with "woke" principles.
Edit 2 -
Saw this today
https://www.youtube.com/watch?v=ByV5w1ES38A
A locally hosted and pre-trained chatgpt type of chatbot, and, it can apparently be trained locally. Will look into.
I'm working on my own AI software right now. It's a bit of a slog. The problem with this is that you need hundreds of billions, maybe trillions of weights/variables depending on what kind of machine learning models you use. That's why this stuff is hosted in the cloud and sucks up computation cycles.
Might be a bit pricey to try to host it yourself, not to mention trying to train the model. I have a fairly hefty machine, but I imagine you'd want to use the cloud and thus Amazon AWS instances or something equivalent. That's pricey.
https://www.youtube.com/watch?v=ByV5w1ES38A A locally hosted and pre-trained chatgpt type of chatbot, and, it can apparently be trained locally. Will look into.
It costs $600k per day to host ChatGPT… that doesn’t even include the training time/cost.
That's pretty steep for a program developed initially by a not for profit company. I think the real cost is the training, getting enough data points for the supervised learning algo to do something interesting.
The data is easy - they use Wikipedia and a large book corpus. But the training time is 4ish days on a large cluster of extremely expensive hardware.
Microsoft bought OpenAI so they’re not really hurting for cash now. The numbers I’m quoting from are here: https://www.semianalysis.com/p/the-inference-cost-of-search-disruption?publication_id=329241&isFreemail=true
right, ChatGPT is incredibly diverse and they've thrown a MASSIVE amount of data from the internet at it, and may be one of the biggest projects done.
But a narrowly focused dataset, a fraction of what ChatGPT is looking at, just conspiracy stuff, and then having an algorithm go through it all to put pieces together and correlate this or that and answer questions about what all it gets, could be interesting and - most importantly - possible to do with normal human hardware and modest amounts of time and effort and skill.