This is a great point. I would make the counterargument that app translators doing simple transactions of languageA to languageB for $X was possible a decade ago. AI certainly has made that low hanging fruit impossible to capitalize at speed and scale. Since AI has made the easiest part of translator's jobs much harder to monetize, translators can still offer value in lengthy transactions. Books that are to be translated for publishing, legal documents, and languages that are not understood by AI.
The (former) translators I know mostly translated
B2B publications (brochures, specifications, …), commonly containing specialized domain vocabulary. Certainly not low-hanging fruit. I’m sure AI can’t make a quality job on those without guidance and review by a speaker of both languages knowledgeable in the respective domain. Nevertheless, the market for such translations has dried up, and those that remain only pay a pittance.
Part of the reason likely is that the perception that translation is valuable work has changed.
It's because the marginal value of idiomatic human translations, over robotic and partially bad machine translations, is not actually that high. Especially not for B2B where people can accept poor prose from a trading partner.
The important thing is for the technical details and terms to be correct, otherwise it might be false advertising, or the spec for what you are requesting might be wrong. And you can’t rely on AI to be reliably accurate in that, or for the translation even to make sense technically.
