AI and New Language Technologies

All Andrea's points are valid, but the people developing AI interpreting and buying it as a service have other priorities, I fear. Those processes are something we have very little influence on.Computer-assisted interpreting means live transcription and boothmate tools that use the transcription to highlight terminology and numbers. It has potential IRL but its use is limited by the confidentiality issue. CymoNote is Chinese, Microsoft's ASR sends data to the US... either way it's going out of the building, which many clients would not allow.

InterpretBank and TErpMate claim to have solved this know with an off-line version of their boothmate tools. Now we should test them to see if they work.

When you have a tool prompting with terminology it should probably be from a specific glossary prepared in advance for that session. Allowing the computer to look for terminology whereever it likes will lead to mistakes. On the other hand, if you have already prepared a glossary the terms will arguably already be activated and you won't need a CAI prompter!

I have tested CymoNote with mixed results. Anyone interested can see the short tests here.

https://www.youtube.com/watch?v=ZIQLi-fZNy4

https://www.youtube.com/watch?v=oS30FrHVSY0

TRAINING: I don't think we should be teaching something that is NOT widely used in the profession yet. Students need to be able to interpret WITHOUT CAI (because CAI will not always be available) and then add CAI to their process after that. Checking a CAI screen is an extra cognitive load which will lead students (and indeed professional interpreters) to interpret less well before it helps them interpret better.

I think it's valid to promote AI literacy amongst professionals, but far less important to do it with students. 99% of an interpreting course needs still to be about how to actually interpret, because without that no AI-tool will help a student.

I would also say that tech-literacy could begin "before" getting as far as AI. Many interpreters do not know how to get the most out of pdf readers (annotation, bookmarks, navigation between documents, exporting terminology) nor about the principles underlying glossary-building and the tools available for keeping glossaries. These topics are less sexy and fashionable than AI but they are easily as useful for interpreters.

The real (and possibly only!) expert amongst interpreters in this field is Josh Goldsmith. He teaches AI and other tech-related stuff at TechforWord.

To answer your questions... "AI-related modules into interpreting courses?" Yes. A brief look at how to make summaries without AI and a comparison of the results with AI generated summaries... and a couple of other functionalities of NotebookLM and Readwise. My conclusions were 1) there's not that much to "teach". AI is very simple to use. And 2) that it is still essential that students be able to make summaries themselves, even if it takes longer. Knowing how to distill the essence out of something is an important skill.The biggest challenge is for the moment confidentiality. We are simply not allowed to put many of our documents through an AI and efficiency means having a similar workflow for all meetings, rather than having a different one (involving AI) only for public meetings/documents.

Another big issue with AI is that if we let it do the analysis, thinking, summarising, terminology work for us, then we are not using the parts of our brains that we need most to be interpreters. I've addressed this in this lecture "AI or DIY" https://www.youtube.com/watch?v=o7VkMC0lGWA&t=1s

I've written about AI and glossary-building here https://interpretersoapbox.com/reasons-not-to-automate-glossary-building/

I ran a survey of AI use by interpreters and confidentiality issues here https://interpretersoapbox.com/ai-tools-confidentiality-conference-interpreters-survey/

I think it's valid to promote AI literacy amongst professionals, but far less important to do it with students. 99% of an interpreting course needs still to be about how to actually interpret, because without that no AI-tool will help a student.

I would also say that tech-literacy could begin "before" getting as far as AI. Many interpreters do not know how to get the most out of pdf readers (annotation, bookmarks, navigation between documents, exporting terminology) nor about the principles underlying glossary-building and the tools available for keeping glossaries. These topics are less sexy and fashionable than AI but they are easily as useful for interpreters.

The real (and possibly only!) expert amongst interpreters in this field is Josh Goldsmith. He teaches AI and other tech-related stuff at TechforWord.

To answer your questions... "AI-related modules into interpreting courses?" Yes. A brief look at how to make summaries without AI and a comparison of the results with AI generated summaries... and a couple of other functionalities of NotebookLM and Readwise. My conclusions were 1) there's not that much to "teach". AI is very simple to use. And 2) that it is still essential that students be able to make summaries themselves, even if it takes longer. Knowing how to distill the essence out of something is an important skill.The biggest challenge is for the moment confidentiality. We are simply not allowed to put many of our documents through an AI and efficiency means having a similar workflow for all meetings, rather than having a different one (involving AI) only for public meetings/documents.

Another big issue with AI is that if we let it do the analysis, thinking, summarising, terminology work for us, then we are not using the parts of our brains that we need most to be interpreters. I've addressed this in this lecture "AI or DIY" https://www.youtube.com/watch?v=o7VkMC0lGWA&t=1s

I've written about AI and glossary-building here https://interpretersoapbox.com/reasons-not-to-automate-glossary-building/

I ran a survey of AI use by interpreters and confidentiality issues here https://interpretersoapbox.com/ai-tools-confidentiality-conference-interpreters-survey/

I think it's valid to promote AI literacy amongst professionals, but far less important to do it with students. 99% of an interpreting course needs still to be about how to actually interpret, because without that no AI-tool will help a student.

I would also say that tech-literacy could begin "before" getting as far as AI. Many interpreters do not know how to get the most out of pdf readers (annotation, bookmarks, navigation between documents, exporting terminology) nor about the principles underlying glossary-building and the tools available for keeping glossaries. These topics are less sexy and fashionable than AI but they are easily as useful for interpreters.

The real (and possibly only!) expert amongst interpreters in this field is Josh Goldsmith. He teaches AI and other tech-related stuff at TechforWord.

To answer your questions... "AI-related modules into interpreting courses?" Yes. A brief look at how to make summaries without AI and a comparison of the results with AI generated summaries... and a couple of other functionalities of NotebookLM and Readwise. My conclusions were 1) there's not that much to "teach". AI is very simple to use. And 2) that it is still essential that students be able to make summaries themselves, even if it takes longer. Knowing how to distill the essence out of something is an important skill.The biggest challenge is for the moment confidentiality. We are simply not allowed to put many of our documents through an AI and efficiency means having a similar workflow for all meetings, rather than having a different one (involving AI) only for public meetings/documents.

Another big issue with AI is that if we let it do the analysis, thinking, summarising, terminology work for us, then we are not using the parts of our brains that we need most to be interpreters. I've addressed this in this lecture "AI or DIY" https://www.youtube.com/watch?v=o7VkMC0lGWA&t=1s

I've written about AI and glossary-building here https://interpretersoapbox.com/reasons-not-to-automate-glossary-building/

I ran a survey of AI use by interpreters and confidentiality issues here https://interpretersoapbox.com/ai-tools-confidentiality-conference-interpreters-survey/

Hi Andrea,

thanks for launching the discussion. As a translator I do not have any relevant experience with AI tools for interpreters, but I just came across an interesting event AI ThoughtCon 2026 by the AI Localization Think Tank. On day 2 there is a session on The New Role of Linguists in the Age of AI Speech that might be interesting for the interpreter community.

Hello Andrea,

I would recommend watching the presentation called 'Code and prejudice' by Marina Pantcheva from Translating Europe Forum 2025. Here's the link: #2025TEF - TEF–TALK: CODE AND PREJUDICE

The presentation touches upon how AI bias emerges in larger language models, and illustrates the presence of AI bias through examples. Marina Pantcheva also explains why we should avoid bias in AI, which could be interesting to see in light of how AI shapes or affects our language right now.

Thank you! I would like to add another conference on this topic:Convergence 2026: Human-AI Integration for Multilingual and Accessible Communication - Guildford, UK, 17 - 19 June 2026

The second edition of the Convergence conference will create an opportunity to bring together innovative research on the evolving landscape of AI in the context of multilingual and accessible communication, reflecting on the complexity and effects of using AI-driven technologies in these fields. The conference will foster a multidisciplinary dialogue that will generate new theoretical perspectives and practical research, focusing on themes such as the ethical aspects of AI in translation and interpreting, AI-enabled digital accessibility and societal inclusion, and the impact of Generative AI on language mediation. We will also examine the evolving role of language professionals, the power of Large Language Models (LLMs) in supporting multilingual communication, and the crucial need for responsible use of language AI in the public sector. The conference will publish full papers in open access proceedings with assigned ISBN and DOI.

Hi Susan!

I've experienced similar problems with our AV-technicians. According to them the problem lies in how cables can feed into computers. But I think Barbara Ahrens in Cologne found a soulution to that when they installed their system couple of years ago.

In Stockholm we have resorted to having students plug in mp3 players to their own console and thereby recording themselves on two channels, only works with live speaker though. As soon as we want to do it with a recording from e.g. Speech Repository we have the sound mixed on the two channels.

That's all I have,

Elisabet

Hi Andy,

I have used GoReact before https://get.goreact.com/ and it works very well as a on online platform to upload activities, listen to students' recordings and give feedback. It is not perfect, but it is a great platform to create different types of assignments (upload webinars, video or audio interpreting assignments, recordings uploaded by yourself, students' presentations, etc.) and the best feature is that it is very interactive. You can provide feedback in different ways (written comments, tags created by yourself to flag up different aspects/criteria, audio feedback, links to videos, etc.). Feedback comments are time coded: students can see when they made that "minor omission" or came up with a beautiful solution that you tagged as "great solution"; they get statistics of the different types of "errors" or good solutions for each recording and course and you can create templates to enter marks. You can set up activities for self assessment, peer assessment and tests.

From a more technical point of you, students need to have a webcam in order to be able to upload recordings, even if they want to upload audio recordings only, but most laptops have one and, alternatively, students can use the app on their phones to access the platform and record themselves.

Each student has to pay around $30, if I remember correctly, unless you have an institutional license. That is a downside, but cheaper than most text books in other disciplines!

I tested it and led its implementation with my former colleague Aída Martínez-Gómez at John Jay a couple of years ago, so happy to answer any questions.

Eloísa

Hi Andrew. If a platform can offer a number of languages, i.e. channels, then I am sure it can do the same for a number of interpreting students. So rather than switch between FR, DE, ES, EN, PL etc you'd switch between student 1, 2, 3, 4 etc. I'm not an expert, but that would be my guess anyway. The best thing to do would be to contact providers with that query.