It's an increasingly less heard argument (in my experience at least) that AI companies would eventually run out of training data and, with the ever degrading quality of synthetic data, some kind of brick wall of development would be hit. It was never a good argument, relying as it did on domain specific imaginations (well they will run out of online art to steal) and naivety (they can only scrape so much online). Which is probably why it's heard less and less now as the voracious hunger for data of AI companies begins to evolve and show just how abyss like its gut is.For a while now people have known about the monitoring of workers for AI training purposes, from Amazon tracking workers to the point of constant, harassing surveillance to factory workers being recorded to build their dead labour into the speculative form of robots. Resistance to that, especially of the everyday sort, can be a hard one to gauge. Whether any given worker or community has thought to engage in go slows, or…
No comments yet. Log in to reply on the Fediverse. Comments will appear here.