21 Aug How ChatGPT Impacts Translation: A Few Thoughts
Barbara McClintock, MA, Certified Translator and Certified Terminologist
Issue
We know that Artificial Intelligence will have a significant impact on the translation industry. ChatGPT can make human-sounding translations and is constantly improving. Billions of dollars are being spent on AI research, including our tax dollars. For the moment, it is free and available to everyone. But where does ChatGPT get its information? Marcin Frackiewicz wrote in a blog post, How ChatGPT Is Revolutionizing the Translation and Interpretation Industry, on June 14, 2023, that ChatGPT is a “text-generating system that is trained on millions of conversations from social media, books, and other sources.” In addition, it should be noted that “To produce accurate translations, ChatGPT must be trained on a large number of examples of human-translated texts.”1
Call to action
Members of the Writers Guild of America and Screen Writers Guild have been on strike since May. As translators work with writers, and we sometimes translate books, plays and scripts, the writer strike is of importance to us. They are complaining that copyrights are not being respected by artificial intelligence models which use copyrighted materials to train AI without permission or compensation. Over 8,000 authors, including Canadian Margaret Atwood,2 recently signed a letter addressed to tech giants because their work is used for training AI despite copyrights. They are asking for AI-generated work to be labelled as such. The public and academia need to know what is AI generated. Translators should also demand labelling, e.g., “This is AI generated.” It’s time for language professionals and their associations to take a stand on the issue and make our voices heard. The best way to do this is through our professional associations and elected officials. What can we do? Send letters and emails to all your associations, MPs and MNAs.
Some hope
An article by Mihad Vlad, general manager of Language Weaver, in the RWS blog dated December 19, 2022,3 offers hope for language professionals. He pinpoints some major negative aspects of ChatGPT impacting translation: quality and data privacy. Mr. Vlad states that ChatGPT (Generative Pretrained Transformer) should not be used by enterprises or organizations where data security and privacy are important. “There is just one single ChatGPT model owned by Open AI, shared with all customers.” This may change in the future, of course. In the meantime, he suggests that the postediting work of translators “will expand into linguistic validation, cultural adaptation, tone adjustment, fact checking and bias removal.” Although he recommends the use of machine translation (MT) models rather than AI, Mr. Vlad thinks that there will continue to be a need for linguistic services for revising, adapting and certifying AI output.
Note: This article does not reflect the opinions of ACJT. It was not AI generated! Please send any comments to Juriscribe in care of ACJT.