Report of the CBTI conference on AI held on October 13
On 13 October 2023 the Belgian Chamber of Translators and Interpreters’ Working Group on Artificial Intelligence organized an afternoon on ‘Artificial intelligence in translating and interpreting’.
Some sixty people – members and non-members of our trade organization – attended the event. The conclusion that ran through all the talks can be summed up in a few words: both the translation and interpreting sectors are definitely in for changes and we have no other choice than to adapt to them! Very luckily, technological advances in the field have not yet reached the point where technology can replace human beings. On the contrary!
Ana Lioara Firea, an interpreter with COPA-COGECA, got the ball rolling with a rundown of the history of AI in interpreting, then went on to tackle CAI (Computer-assisted Interpreting) tools. These tools help interpreters with proper nouns, figures, specialized terms and acronyms, amongst other things. She also reviewed Automatic Speech Recognition (ASR) systems from various angles. Whilst these tools were indeed created by interpreters for interpreters, the matter of refining them with interpreters is a question well worth asking. Ana then quickly ran through a few AI-powered RSI platforms (Kudo, Smarterp and Interpretbank). In light of the knowledge amassed to date, it turns out that using these technologies, albeit promising, will require that a number of conditions be met: training to enable interpreters to master these tools, a guaranteed minimum latency time, appropriate precision and ergonomics and the participation of a large sample of users (from students to experienced interpreters), without of course overlooking the need to include a larger number of languages and to allow for native and non-native speakers’ accents.
Silvia Mandrioli (freelance translator, beta tester for Cotranslator.ai and Custom.MT and annotator for Google BARD), then took the floor to tackle the issue from the more specific viewpoint of the translator. She first examined the historical differences between Google Translate and ChatGPT and then went on to consider their respective advantages and drawbacks. She also gave an overview of the software (CoTranslator AI, Lokalise and GabyT) and plug-ins available for using AI in translating, with a focus on plug-ins for SDL Trados (OpenAI Translator and Custom.MT), and presented several hands-on examples of using ChatGPT in translating and revising texts. In conclusion, these tools are still in the development phase and must be further optimised. Silvia also raised some financial issues – should payment be based on time or the number of words? She tackled the specific needs of translators, proofreaders and annotators and wrapped up with a brief examination of certified translations and ChatGPT.
Alba Carvajval (interpreter) and Jeroen Steel (translator) then took over with a report on the ATA conference ‘Empowering professionals to innovate and strive’, which they both attended. Alba began by underlining the fact that fewer interpreting than translation applications were covered during this conference. What she learned from attending were the urgent need to cope with the changes on the cards and the importance of differentiating oneself from the competition and of developing marketing skills at the same time. Hyper- specialisation and the undeniable IT component of work in these fields are also developments on the horizon; fewer interpreting positions can doubtless be expected, but this will be offset
by markedly more specialised new positions. The matter of the under-representation of women in the AI field was also raised by Alba, who ended her talk with a mini-demo of using Notion to create glossaries, abstracts and definitions.
Jeroen Steel began with an introduction on the changes that AI will engender in the translation sector with a description of the three phases of development of machine translation, namely, rules-based MT (RBMT), then statistical MT (SMT) and finally neural MT (NMT). Certain orders in this new market will be manna from heaven for machines whereas others will be closer to post-editing machine translations. Certain jobs will be assigned to translators with social skills or to specialists. After that, Jeroen took on the task of defining the advantages of human translation over the performance of machines, to wit: providing added value, sensitizing the client and developing a marketing approach. He then turned to the prospects for new jobs, notably in the area of post-editing, be it light or complete. He went through the processes and shed light on the traps that these developments present, before concluding with considerations of the rivalry between or symbiosis of machine translation tools and translation memories.
Rita Roggen, a translator-interpreter and member of CBTI/BKVT’s LinguaJuris Commission, ended the series of presentations with a talk on confidentiality in the realm of artificial intelligence. After dwelling upon the definition of AI, Rita brought the cultural and emotional skills of humans to the fore. In a nutshell, unlike machines, humans can detect hints of humour and irony and have the ability to correct their mistakes. The thread running through her presentation was the matter of confidentiality, with the imperative need to guarantee maximal data protection and to anticipate the dangers that may be inherent in data storage and use. She wrapped up by considering our ethical obligations and complying with the GDPR and touching upon the EU’s preparatory work on the AI Act, which is expected to provide the underpinnings for Belgian legislation to promote reliable AI and to guarantee security.
To wrap up these very rich exchanges and discussions, CBTI/BKVT President Max De Brouwer took the floor to present the work launched by the Chamber in various areas, including the draft contractualization agreement for linguists and the meeting with some fifteen trade association presidents that has led to the creation of AI working groups under the aegis of FIT. He extended a warm invitation to all interested participants to join CBTI/BKVT’s AI working group and expressed the hope that this meeting would be the first of a long series.
If you would like to support the AI WG, please don’t hesitate to contact the CBTI/BKVT secretariat (firstname.lastname@example.org).