ParsingphaseOne of the dirtiest secrets of the "AI boom" is that it will depend entirely on refuseniks to have any sort of long-term future. LLMs require a continuing supply of high-quality, human-created data to train and update on, or they'll devolve into ever more evident slop, to the point of failure. 1/
This means that those of us with the decency, ethics and long-termism *not* to use the damned things will be ever more intensively parasitized, while being constantly aggrieved as reactionary technophobes, particularly by the industries that become ever more dependent on us. 2/
Sadly, while it's morally, ethically, and practically essential to keep doing things the "old, inefficient, human way", without some very unlikely social and legal changes, we're not going to be able to stop our work from being used to prop up the "AI" industry, until it collapses. 3/
Much like those who worked to avert the Y2K bug, we can expect this work to be thankless, ignored and essential, but all the more abused. 4/4