Future of Translation: How Long Will Humans Be Needed?
Machine translation accuracy is rapidly improving, but experts question its sufficiency for all situations. Errors in legal and medical documents, in particular, can have serious consequences.
The capabilities of machine translation are advancing at an unprecedented pace. Google Translate alone handles one trillion words per month, and numerous other AI systems like DeepL and Microsoft are also prominent. The central question has shifted from whether machine translation is good enough, to whether human translators can be entirely removed from the process.
Direct sales of machine translation were estimated at $1.4 billion last year. For many applications, such as internal documents or product listings, raw AI translation is now considered a rational choice. For instance, the Open Knowledge Association (OKA) uses AI to draft Wikipedia articles, significantly reducing costs compared to traditional methods.
However, experts emphasize that human input remains crucial in specific fields. Legal translator Ariella Baltié points out that errors in legal documents, such as mistaking "plaintiff" for "requester," can lead to serious misunderstandings. Similarly, in medical translations, a minuscule error in dosage, like confusing microgram with milligram, can be life-threatening.
Experts note that as AI translations become more sophisticated, detecting errors grows harder, underscoring the need for human oversight. While machines enhance efficiency and reduce costs, human judgment, cultural nuance, and critical accuracy are indispensable where stakes are high.