×
Login Register an account
Top Submissions Explore Upgoat Search Random Subverse Random Post Colorize! Site Rules Donate
2
3 comments block


[ - ] prototype 0 points 5 monthsNov 19, 2024 08:45:56 ago (+0/-0)*

In 2012 I laid out the case for AI to the DOD, the explicit war-game scenarios (simulating and predicting world leaders, and what regions would produce them, and their decision-making styles). In 2022 I ran a discussion group face-to-face with google engineers among the crowd (along with researchers from other ML companies) discussing step-by-step how to build on BERT and GPT, back when they were at a stand-still.

I gave those faggots the research that lead to BERT, o1, and page-rank based RAG.

I know what le-next-best-architecture and methodology is. I'm just not giving them any more research because I got nothing out of it the last four times. Not even credit.

[ - ] ArcturianDeathTrap 0 points 5 monthsNov 19, 2024 10:16:03 ago (+0/-0)

They already stole it

[ - ] prototype 0 points 5 monthsNov 19, 2024 15:23:31 ago (+0/-0)*

They already stole it

Kek.

I keep all my hard research in a hard-copy binder. If it's not copied from that book, it's not available anywhere else.
They can't and won't figure out how to build a high-quality data pipeline (HQDP) for another two or three years, maybe longer.
And thats their limiting factor right now.

They need a particular modality of HQD and a particular pipeline architecture to manage it.