Why I'm against "AI translation"

"AI translation" relies on a[n at least] two-part system of human exploitation to exist and be "successful" so a principled stance should not even take into account the quality of the final product. First, language models are trained unethically on copyrighted text (including at least one novel I translated according to a training data set that was revealed). There is no "artificial intelligence" without the exploitation of human intelligence and labor. In a utopia where no money was necessary to survive, we could probably abolish copyright or whatever, but humanity will not be reaching that level of enlightenment anytime soon. Second, even the people cheerleading for the machines expect to hire the very artists and professionals they are trying to cut out of the process to edit the "AI" output. (Anyone who tries to get me to do this will be quoted 10x my usual rate with the option to hire me to translate instead for my usual rate.) Some translators seem to want to promote "AI" as a tool that translators can harness for their work, and if it stopped there, as translators using a tool on their own terms (and the fundamental ethical issue were resolved), that would be all well and good, but we know it doesn't stop there, because it's already not stopping there. There was just a giant round of funding for a company that does "AI" translation of manga (and the company even admits that the model needs its bottom wiped by humans—why do we call it "AI translation" again?). Even if you don't care about arguments like, "Human literature deserves human translations," there are important reasons to oppose "AI" translation.

The above was originally posted in a FB discussion on May 16, 2024. I found myself making the same points often enough that I decided it might be useful to have a link…