Summarising several "scientific" studies where:

[they] reduce a human task into an oversimplified game that, at its core, involves producing some plausible-looking text, only to conclude that LLMs can, in fact, generate some plausible text. They have committed one of science's cardinal sins: They designed an experiment specifically to validate their preexisting belief.

The pandemic showed us that undermining the public's trust in science can cost human lives, but the harm here goes further. These so-called studies are purposefully, almost explicitly designed to reach the result that workers are dispensable.