The year 2023 started with a nervous economy and optimistic AI investors, and at Slator we found ourselves saying more than once how hard it was to keep up with so many AI-related language technology breakthroughs and startups, not to mention hundreds of scientific papers dealing with large language models (LLMs).
It has been a most interesting year, and we assembled a few of the most notable quotes of the year, from business leaders who were right on point, to the memorable and even the brow-raising. Quotes are selected from Slator’s 2023 news coverage, SlatorCon Remote and In-person presentations, and SlatorPod interviews.
The Direction of the Language Industry
June 2023. During SlatorCon Remote, keynote speaker and Propio Language Services CEO Marco Assis said that very few clients had asked his company about ChatGPT or LLMs. “When you talk about a regulatory environment, you have to be careful about what you automate… we’re talking about outcomes that cannot be reversed, and coming back to similar analogies to the driverless cars, to where a concept seems doable, reality is quite deeper.”
June 2023. Translated CEO Marco Trombetti believes there will still be growth in the language services industry, with higher demand in multiple verticals, including trade, migration, culture, and entertainment. In his view, which he laid out in Translated’s Imminent Report, “it will be possible to translate more, faster, and much better … An industry that manages to improve productivity and enjoys growing demand is a healthy and prosperous industry because it is always innovating and developing new markets.”
August 2023. Phrase CEO Georg Ell covered many different topics during a SlatorPod episode, and his take on the immediate future and how people experience localized content stood out: “I was imagining a world in which the entire internet is in real time, real-time internet streamed content. And the implications of that for volume are literally astronomical because it’s not ten, a hundred thousand X increase in volume of content produced. It’s a billion X or a billion billion x.”
October 2023. RWS Group CEO Ian El-Mokadem delivered the keynote speech at SlatorCon Zurich, addressing several key points about the current transformation the language industry is undergoing. Referring to the way the linguist’s role is changing, El Mokadem said there is a need for smart technologists, creative consultants, content optimizers, and industry experts. He asked “What’s our message to translators right now? … Well, it’s goodbye, translators. And hello, language specialists.”
October 2023. Sheila Castilho, Assistant Professor at Dublin City University, spoke at SlatorCon Zurich about how she sees things developing as LLMs enter and transform translation workflows: “I hope that there’s going to be more evaluation being done. We’re going to be able to really tell what is realistic in translation with LLMs and then, after we’ve figured that out, we’re going to reach a plateau of productivity until the next breakthrough comes in. That’s basically what has been happening in the past 50 years.”
November 2023. Lars Thorsell co-founded Sweden’s multimedia LSP Svensk Medietext in 2006 and has since served as head of production, CTO, VP and, as of November 2023, CEO. Thorsell takes on his new role at this major provider of subtitles for Nordic languages in the era of automatic speech recognition and AI applications, but as the CEO put it to Slator, “We’re still developing and defending traditional human-based subtitling. We’re not giving up on that. We mainly work with premium content and we cannot compromise on quality.”
November 2023. ASAP-Translation CEO Jakub Absolon was a guest on SlatorPod and shared some insights on machine translation post-editing. To him, “full post-editing is simply human translation and should be priced and timed as such.” He added that “now some translation agencies are also using the process and they call it ‘MTPE plus revision.’ Very strange for me because it’s similar to when you cook instant soup and ask the chef to season it to the level of a five-star restaurant.”
November 2023. Co-founder and CEO of Intento, Konstantin Savenkov addressed the transition from the known use cases of machine translation to the advantages of generative AI automation during SlatorCon Remote: “You can go from 70% effort-saving, which you can get just by using good, custom neural machine translation, to 95% effort-saving by automating every step of a workflow, to a certain extent.”
January 2023. LXT Chief Growth Officer Phil Hall was a guest on SlatorPod, where he referred to the comparison of the AI startup tsunami to a gold rush. Hall said that “When my friends have asked me what I do, and I explain to them and they still don’t get it, I’ve actually used the same metaphor to describe it … I read a report last year that said there’s probably 150 players in this space now. So, much like a gold rush, there’s a lot of people coming and setting up shop close to the mines.”
June 2023. Aki Hayashi, a Senior Manager of Dell’s Globalization Program Management team, discussed the topic of AI in her company’s business during SlatorCon Remote, saying that “We started this journey in response to the changing business landscape with AI … But we realized that simply relying on ‘inside-out’ thinking could be self-imposing and limit our progress.”
October 2023. Smartling CEO Bryan Murphy was a guest on SlatorPod and commented on perceptions of how AI will impact the bottom line across the language industry. “A lot of CEOs are saying, ‘this is great, let’s use it to create content and to translate content … that should be it, it should take our cost to zero, right?’ And I think that this is where the education process comes and says, ‘well, no, not so fast’ … Yes, this is a huge opportunity for you to reduce your cost of translation and improve your digital footprint and conversion rates, etc., but it’s not a standalone solution.”
November 2023. Bertrand Gstalder, CEO of Acolad, gave the keynote speech at SlatorCon Remote November 2023, and described to attendees how using an AI workflow for a multimedia localization project had shortened turnaround times, and integrating synthetic voices had reduced costs. “We have been down this path before,” Gstalder said of the advent of AI. “No reason to fear, just motivation to act and take the lead before others do.”
November 2023. Gabriel Fairman, founder and CEO of Bureau Works spoke on SlatorPod about how the paradigm shift with AI in translation production is that “all of a sudden there’s a much, much simpler way to get things done, a much simpler way to think about translations. You don’t need any more translation-editing-proofreading to get to professional-level translations: you’re getting all these layers of added verification within a single workflow step.”
All Things GPT and Other Large Language Models
February 2023. Angelo Passalacqua, founder and CEO of Zin Global (a CRM company) asked a memorable rhetorical question on a LinkedIn discussion that followed a press release posted by Lilt. In the press release, Lilt compared its machine translation model performance to that of OpenAI’s GPT-4 and Google Translate. Asked Passalacqua “I wonder why knowledge of the entire internet is not adequate for machine translation.”
March 2023. During a panel presentation at SlatorCon Remote, Jon Ritzdorf, Procore’s Senior Manager of Global Content Solutions spoke about the similarities of discussions around GPT to those about neural machine translation (MT) back in 2016, including the need to address privacy, speed, accuracy, and reliability. Ritzdorf said “Where MT is probably still very good at structured stuff, GPT might be a little better at the free-form craziness that we all really use in real life.”
August 2023. Referring to GPT-4 document-level translation capabilities, Longyue Wang, a Research Scientist at Tencent AI Lab who has extensively tested LLMs, said of OpenAI’s model during a SlatorPod episode that “One of the standout features of GPT4 is its ability to grasp the broader context. When we’re translating entire documents … we need the flow, the consistency, from start to finish .. GPT-4 can handle it well.”
September 2023. Graham Neubig, Associate Professor of Computer Science at Carnegie Mellon University, summed up the potential use of large language models as part of translation system development during a SlatorPod episode: “Companies that are serious about building translation systems will probably move in the direction of taking a general-purpose language model and then fine-tuning it to be a really good translation model.”
‘Wait, What?’ Moments
June 2023. Eric Mamer, the European Commission’s Chief Spokesperson called the addition of machine-translated press releases without human review on the EC site’s Press Corner “a little housekeeping announcement.”
November 2023. A speech by Joana Cotar, a member of Germany’s lower house of parliament, was translated and lip-synced into English using AI, and posted online in an unedited format. Cotar qualified the availability of the AI-dubbed video as “wonderful” … that her speeches are now accessible in “English, Portuguese, Spanish, French, Indonesian, and even Chinese”.
November 2023. The European Council of Literary Translators’ Associations (CEATL, in French) urged authors to insist on opt-in clauses for the use of copyrighted material in AI training and declared that “machines are not translators but ‘translatoids’.”
Machine Says …
Since we are now having chats with robots about, well, everything, we had to include a couple of quotes from bots.
Google Bard: “I am an LLM trained to respond in a subset of languages at this time, so I can’t assist you with that.”
ChatGPT: “please note that translation is as much an art as it is a science.”