Inflation, layoffs, rising interest rates: Rather than sugarcoat economic realities, Slator Co-Founder and Commercial Director Andrew Smart opened SlatorCon Remote March 2023 with some advice for attendees facing down a potential recession.
“I recommend you ski the path and not the tree,” Smart advised on March 22, 2023. “If you look for the path, you will see the path, and you will make your way through.”
And the 400 registered participants from more than 40 countries, split evenly between vendors and buyers, responded in kind, with highly engaging discussions at the 10th SlatorCon Remote since May 2020.
Slator’s Co-Founder and Managing Director Florian Faes furthered the case for optimism as he discussed the rapid developments taking place in the world of large language models (LLMs) and their potential impact across the language industry, in areas as disparate as voice acting, dubbing, coding, marketing, and content production.
The common link? The input is text; users can interact with these models in natural language — no code needed.
“It’s a much broader user base across the enterprise that wants to get access to these types of tools,” Faes noted, adding that 40% of professionals reportedly already use this technology.
Faes suggested that society may be climbing into the next peak of inflated expectations for LLMs. Case in point: a March 2023 paper by OpenAI and the University of Pennsylvania attempting to measure the “exposure” of more than 1,000 jobs, including interpretation and translation, to LLMs.
“80-90% of the language industry already runs on post-editing, meaning we’re already way advanced in the labor augmenting stage here, and these large language models aren’t superior at translation than other, narrower models,” Faes explained.
But time will tell how, and to what extent, LLMs can be put to use. Transcription, for example, has already gotten a major boost from models such as Whisper, and a new feature of GPT-4 performs translation using both text and images as input.
Other potential use cases for the language industry include pre-editing (making text more suitable for translation); automatic post-editing (to address doubts about style); and machine translation (MT) quality estimation (though more focused models would likely still outperform broader LLMs).
All of these spell opportunity for language service providers (LSPs) and buyers, many of which have, or can create, collect, and curate, data otherwise unavailable on the internet.
“You want to make sure that you understand how you can potentially leverage this data for niche use cases going forward, because I think it’s going to get a lot more valuable than it was in the past,” Faes said. “This is going to be an absolutely pivotal year for a lot of businesses in this space.”
The People, and Players, Behind the Scenes
While tech — especially the newfound abilities of ChatGPT and its ilk — featured heavily throughout the conference, keynote speaker and Argos Multilingual CEO Véronique Özkaya returned the focus to the human element.
“I think that if you have the right people, you can achieve anything,” Özkaya said, explaining that when she joined the company in May 2020, she and her colleagues identified “movers and shakers” who could be entrusted with more responsibilities and challenges.
Nurturing entrepreneurial staff with stretch goals, she added, has helped cultivate a culture that supports strategic M&A in a global organization — leading, in turn, to Argos Multilingual’s impressive 64% revenue growth in 2022 to USD 72m in 2022.
One vertical that has seen growth on this scale, particularly since Covid, is gaming, with approximately three billion gamers worldwide.
In a panel on game localization trends, Slator’s Senior Research Analyst Anna Wyndham asked Lionbridge Games Managing Director Tugdual Delisle and Altagram Founder and CEO Marie Amigues whether MT can help LSPs meet tighter turnaround times and address the growing lists of target languages.
Delisle pointed out that as the technology matures, MT may be used more frequently in game localization. On the other hand, MT for gaming must be accompanied by human verification — a task made that much more difficult when English is not used as a pivot language, as human translators are that much harder to find.
Automatic Dubbing and ChatGPT
Automatic dubbing has already made significant strides, AppTek’s Managing Director, Volker Steinbiss, explained in a presentation on the technology’s advancements.
Most recently, automatic dubbing has moved beyond traditional neutral-emotion speech delivery to more emotionally engaging speech, through the post-editing of synthetic voices.
Picking up the thread of LLMs and AI, Konstantinos Chatzitheodorou, Head of Machine Learning at Pangeanic, led a group discussion on the future of user-friendly tools for the language industry, moderated by Smart. To avoid ChatGPT’s US-centric biases, Chatzitheodorou recommended that European users opt for “competitive” alternatives that will incorporate cultural diversity and preserve European languages.
Booking.com’s Senior Director of Content, Chris Dell, walked attendees through the global travel platform’s localization practices, which recently culminated in the launch of its 46th language, Hindi. The company draws on the large, high-quality corpuses of content translated for years by freelancers to develop and improve its own MT system.
In addition to the investment in machine learning and an internal localization team, the company works with a mix of outside contractors and LSPs to meet the needs of more than 100 groups of regular stakeholders for browser- and, increasingly, app-based content.
Smart returned to (Chat)GPT once again as he moderated a panel featuring Intento Co-Founder and CEO Konstantin Savenkov; VMware Senior Technical Localization Program Manager Martian Xiao; and Jon Ritzdorf, Senior Manager of Global Content Solutions at Procore. The discussion touched on GPT’s use in enterprise localization and language education, among other instances, and triggered dozens of questions in the event chat. Clearly, ChatGPT and its implications are on the mind of everyone in the language industry.
Coupa’s Director of Localization, Jose Palomares, pivoted the discussion in his presentation on how to drive, nurture, and grow a localization practice. He reviewed some of the unique challenges of B2B localization and shared strategies to navigate them better. His parting advice is, “Maintain a plan, never be surprised, be prepared to do what you need to grow your team.”
Participants closed out the online conference in breakout sessions for further networking and discussion with language buyers and vendors.
For those who missed SlatorCon Remote March 2023 in real-time, recordings will be available in due course via our Pro and Enterprise plans.