“… a growing number of articles appeared to be copy-pasted into Wikipedia by people using machine translators. They were riddled with elementary mistakes—from grammatical blunders to meaningless words to more significant inaccuracies, like an entry that claimed Canada had only 41 inhabitants..”
– Jacob Judah
Poor input is prompting concern about knock-on effects for fragile languages, reports MIT Technology Review. One example is textbooks generated by LLMs largely drawing on Wikipedia entries that were flawed to begin with.
They say that sometimes Wikipedia passages sourced for the rare language start with an AI translator, but the individuals involved lack language knowledge, making it impossible for them to correct the AI’s errors.
The result is compounded when the model goes on to generate textbook entries for others to learn from.
How AI and Wikipedia have sent vulnerable languages into a doom spiral | MIT TECHNOLOGY | September 25, 2025 | by Jacob Judah