somebody said once we are mining "low-background tokens" like we are mining low-background (radiation) steel post WW2 and i couldnt shake the concept out of my head
(wrote up in https://www.latent.space/i/139368545/the-concept-of-low-back... - but ironically repeating something somebody else said online is kinda what i'm willingly participating in, and it's unclear why human-origin tokens should be that much higher signal than ai-origin ones)
Since synthetic data for training is pretty ubiquitous seems like a novelty
every human generation built upon the slop of the previous one
but we appreciated that, we called it "standing on the shoulders of giants"
Low background steel is no longer necessary.
"...began to fall in 1963, when the Partial Nuclear Test Ban Treaty was enacted, and by 2008 it had decreased to only 0.005 mSv/yr above natural levels. This has made special low-background steel no longer necessary for most radiation-sensitive uses, as new steel now has a low enough radioactive signature."
https://en.wikipedia.org/wiki/Low-background_steel