How long will humans be needed for translation? (www.bbc.com)

Recent discussions among experts have highlighted the ongoing relevance of human translators in the face of advancing AI translation technologies. While AI systems like DeepL and Microsoft Translator have improved significantly, translating idiomatic expressions and context-sensitive phrases remains a challenge. François Yvon, a research director at CNRS, points out that AI struggles with context, such as misinterpreting legal terms or medical dosages, potentially leading to severe consequences. Ariella Baltié and Carol Bereuter's experiences illustrate how errors in AI-generated translations can jeopardize legal and medical communications, emphasizing the necessity of human oversight. John Zimmerman of the Open Knowledge Association notes that while AI can produce competent drafts, human translators are still essential for reviewing and correcting these outputs. In a study conducted by OKA, 27% of AI-generated texts required corrections, with some errors altering their meaning. The consensus among experts is that despite the advancements in AI, reaching 100% reliability in translations will be challenging. Yvon suggests that future developments may involve AI systems capable of self-assessment and recognizing their limitations, indicating a potential path toward more reliable translation tools.