TL;DR: Mozilla’s translation bot on Support Mozilla (that is currently overwriting user contributions is based on the closed source, copyright infringing LLM, Google Gemini. This is in spite of Mozilla claiming that they are at the forefront of open source AI, and belies their exhortations to choose to build open source AI and data sets. Although Mozilla has experience in attracting open contributions for data sets in projects like Common Voice, Mozilla is using a closed data set to overwrite open contributions. Since (paid) Gemini queries do not train the model, Mozillians can expect to correct errors every time the bot automatically updates an article.
‘Betrayal’ is a harsh word for optimizing trivialities with a closed source model. Whether it was a success or not, doesn’t really tell us why it’s a betrayal. I guess this has to do with the current wave of ‘AI-hate’ reactions we see everywhere.
It has to do mostly with overwriting community (human) translations with (worse) machine translations. Paying google to redo the work your community did in good faith but badly with tech that goes against everything they’ve been saying they believe in. Doesn’t really bode well for the “opt-in” and “ethical” LLM features they claim are coming.