Mike Powell wrote to KURT WEISKE <=-
AI companies seem unrelenting in training their models on any content
they can get their hands on, restricted or otherwise. I wonder when
we'll hear that enterprise content is being sanitized and used to train
their models? Do we trust Microsoft or Anthropic when they claim limits
on enterprise model training?
Define "enterprise content." I am thinking that some of the AI models have already been accused of, or actually caught, doing this?
My company uses Microsoft 365 and their enterprise copilot offering.
They claim that while it has access to company data in your tenant, it
does not use that data to train public models - only for internal use
in the company.
Jaron Lanier has some interesting ideas about creative works and the
internet. He imagined an internet with micropayments and a world where
creative works weren't free but reflected that there was a person
creating them. Imagine if you could like a song on YouTube and send the
artist a micropayment directly?
That would be nice, and would maybe be a way to get around
demonitization when a platform stops showing ads (and sharing revenue) with a creator.
It would also break the royalty system that rewards the music publisher
and barely compensates the creator. I remember the CEO of Pandora
talking about how much they paid in streaming royalties and comparing
that to artists complaining about their compensation.
... THE HEXAGONS OF AIM
--- MultiMail/Win v0.52
* Origin:
http://realitycheckbbs.org | tomorrow's retro tech (1:218/700)