Recent advances in computer technology, and in particular the powerful capabilities in both text generation and translation demonstrated by systems like DeepL or ChatGPT, have convinced many observers that speech-to-speech (S2S) translation is one of many functions in which AI systems will soon surpass human performance. To evaluate the strongest claim – that AI could one day instantly interpret natural, spontaneous, original, contextualised exchanges to a standard of comfort and accuracy expected by human communicators – we face several obstacles. On the one hand, the recent promotional discourse has confirmed the vast conceptual gulf between the tech and ‘humanities’ communities about what real-life ‘language interpreting’ entails. On the other, proprietary interests prevent developers from releasing technical information needed to assess how these neural networks might develop the necessary ‘human-like’ abilities (which even developers admit is mysterious).
In this presentation we explore some of the challenges to human-compatible machine interpreting (MI), as acknowledged by its designers, through a view of linguistic communication from a different, more humanities-oriented branch of cognitive science, rooted in the observation of ordinary language: i.e., pragmatics, and in particular, relevance theory. If time allows, we will try to answer three questions: “Can we?” (is MI feasible?); “Should we?” (is it worthwhile?); and “Will we?”: What would be the likely impact on intercultural communication if beta-versions are imposed regardless, either to augment or replace human interpreters?
About the Speaker:
Robin Setton is a conference interpreter (AIIC 1983-2020, freelance and staff), researcher, trainer, translator and author of dual French and British nationality, working mainly between these languages and from German and Chinese. He holds a PhD in Applied Linguistics (1997), and has been active since 1990 as a trainer and course designer for interpreting schools and institutions in Europe and Asia (Paris, Geneva, Shanghai, Taipei), and more recently, subtitling Chinese independent films. Publications reflect an interest in the interaction between pragmatics, cognition and culture in cross-lingual communication, and include a 1999 monograph on simultaneous interpreting and the co-authored Conference Interpreting: A Complete Course and Trainer’s Guide (2016).
Background Reading (Publications):
Relevance Theory and Interpreting:
Setton, Robin and Andrew Dawrant. 2016. Conference Interpreting: A Complete Course. (Benjamins Translation Library: BTL 120). Amsterdam and Philadelphia: John Benjamins.
(Pages 10-16 (2.1.4-2.1.8) How verbal communication works)
Setton, Robin and Andrew Dawrant. 2016. Conference Interpreting: A Trainer’s Guide. (Benjamins Translation Library: BTL 121). Amsterdam and Philadelphia: John Benjamins.
(Pages 473-485 (12.2.2) Overview of Relevance Theory (and interpreting))
Technology and Interpreting:
Fantinuoli, Claudio. From Assistive tools to Full Automation. Third HKBU International Conference on Interpreting. Translation Centre, HKBU, December 2022.
Relevance Theory (more in-depth)
Sperber, D. and D. Wilson (1986) Relevance: Communication and Cognition. Oxford: Blackwell (2nd edition, 1995). Available here (Google books).
Sperber, D. and D. Wilson (1987) “Précis of Relevance: Communication and Cognition.” Behavioral and Brain Sciences 10: 697-754. Available here.
Carston, R. (2011a) “Relevance theory.” In: Routledge Companion to the Philosophy of Language. Eds. G. Russell and D. Graff Fara. London: Routledge. Available here.