A Wired piece discusses some of the dark underbelly of LLMs. Currently the current and next big thing from big tech, the models don’t train themselves (yet). So how do you get the labor necessary to label data? You invoke the “gig economy” and “outsource” the “micro tasks” to “contractors.” Another way of saying you find people who are willing to work cheap. This concept isn’t new (e.g. Mechanical Turks), and isn’t solely a digital phenomenon (e.g. moving textile manufacture to countries with lower labor costs – and less stringent environmental regulations but that’s another topic).
As we race towards an AI-driven future, most eyes are on the money to be made and the existential threat to humans. But there are very real impacts on humans now that seem to be lost in the hype.