GenAI and the future of multilingualism. Anthony Pym

[@lauravaz5712]

Thank you so much for sharing your insights with us. You’ve been such an inspiration for my academic path that I cannot thank you enough. Kind regards, Laura 🇵🇹

[@janet77822]

thank you for sharing this video, very interesting!! love your humor

[@alexhuang3965]

15:01 I have to point out this is mainly due to two factors: First, the language differences between English and Spanish/Chinese. Second, the data that LLM is trained. Spanish is very similar to English in many ways, while Chinese, is Chinese - you know what I mean. LLM, in this case ChatGPT and all its western counterparts, are trained based on mainly the English language. Comparatively, these LLMs are not better trained in languages like Chinese than the Indo-European language family combined. That explains why GPT 4o is doing well in Spanish but fail the expectations of Chinese users. If, alternatively, a China trained LLM is employed, Chinese students might be surprised to see better results, even tho their expectations are still high. That being said, Calude is an exception. It's doing extremely well in dealing with/producing Chinese texts, far better than GPT.

[@AnthonyPym]

The translation problems in Spanish and Chinese were different but at similar levels. We will be publishing our full report on this experiment next year.

[@user-yf1ml8jt3i]

Anthony.. the quality of this video is Potato

[@AnthonyPym]

Take a look at the original and you will see why: www.b2match.com/e/2024tef/sessions/c2Vzc2lvbjoxNzY2NDM=