HN user
Advertisers. Naturally someone who feels excluded or unable to compete on cleaner markets will offer the portal for people who don't have a regular id and if the ads on those portals do best if they are for toys then those are the ads they will sell.
Does the stand rent a cherry picker to clean their chimney? It may seem absurd, but it is probably fair that a restaurant has to take steps with it's externalities, and only absurd in the context of all the interest groups that don't have to take similar steps. Personally, I would suspect my health was damaged by living directly above restaurants in either of two places with terrible attention to air quality risk.
LLMs is where value capture is anticipated by the market. There's scant evidence they can capture value if they aren't relying on hardware scarcity and their own investment and contractual risks.
A model trained on all the data X was trained on should be improved to the extent that X is already out of date. A model trained on X itself has all the errors of X and all of it's own. Society itself seems to show that model collapse is entirely possible today and was presumably a problem in the past given the significance placed on citation and going to original sources that predates obsession with credit.
Personally, I find it rather humorous that we've moved from the fear that AI generated output would corrupt training to the idea that it is essential to training. Reality itself has not just a left bias but a bias to fundamentals. Bootstrap from fundamentals without introducing arbitrary error and you have the superior system; it just may not be highly compatible with a trash ecosystem.