Event description
This lecture presents a study evaluating the reasoning ability of machines in the context of translation. Can TAN and generative AI programmes pass the so-called 'Delisle test'? Do they show signs of an ability to reason, as adequately trained human translators do when translating from English into French?
To answer this question, TAN tools (MS Bing Translator, Google Translate and DeepL) and generative AI programmes (ChatGPT, NotionAI and Gaby-T) were tested using a selection of examples drawn from 30 objectives of Reasoned Translation (3rd ed.), with the results analysed in light of the explanations provided in the manual.
The lecture begins with an analysis of the ideas that inspired Jean Delisle’s handbook and an examination of the concept of 'reasoned translation'. It then presents the methodology and findings of both the pilot and main experiments. Finally, several avenues for future research are proposed, particularly in the field of translation didactics.
Join the session here.