I work from Spanish (and sometimes Catalan) into English.

Machine translation, like all aspects of machine learning/AI/LLMs can be a force for good, so long as it's used properly, its limitations are understood, and compensation is given to those whose work has been drawn on to train it (i.e. all professional translators working in at least the past 50 years, whose output has been digitised in bilingual corpora).

If it means providing access to more cultural and academic works across a broader readership, because they otherwise wouldn't be translated on cost-benefit terms, that's great. As translators we exist to let ideas flow around the globe, as has been the driver of human knowledge since time immemorial.

If it means undercutting human translators and/or generating substandard output, not so much.

And if it means undermining translation as a viable course of study/career, while simultaneously generating substandard translations with no one trained to correct them, so they then get fed back into the system and generate a kind of feedback loop, even worse.

We shall see.