Just a month after SlatorCon London 2024 wrapped in May 2024, the Slator team and key players in the language industry reconvened for a “vibe check” 18 months into the AI hype, disruption, and opportunity.

Slator Head of Advisory Esther Bond welcomed attendees to the event, providing an overview of the day’s presentations and panels.

Florian Faes, Slator’s Managing Director, shared highlights of Slator’s 2024 Language Industry Market Report. While the findings showed an overall market decline over FY2023, there are some “early signs that there is some stabilization now,” with stock prices and other factors hinting at “cautious optimism” as companies begin to make some money with AI-related services developed since late 2022. However, experienced insiders know that long-term profitability and staying power depend on more than marketing buzz. 

“Anecdotally, I’m hearing people are getting a little tired of just all AI,” Faes said, noting that “maybe less mature buyers” might think AI should have solved all localization challenges by now. “It is the big theme, yes, but we may have to look a little bit beyond it.”

Next up was Christopher Kurz, Head of Translation Management at Enercon GmbH and a project leader for the recently published ISO 5060 for translation output. Developed between January 2020 and February 2024 by a group of participants from 30 countries, the goal of the new standard is an easy-to-implement framework for objective translation evaluation. 

Kurz noted that quality evaluation is not the same as quality estimation (i.e., what portion of text needs to be post-edited), or quality management, which has its own standard, ISO 9001. The analytic standard, which compares segment-per-segment outcomes, allows users to define their own error types and severity levels, based on clients’ specifications.

“Here’s your product. To what degree does it fulfill the requirements the client gave you?” Kurz said. “Because that makes it measurable and that makes it objective.” 

Slator Senior Research Analyst Alex Edwards moderated a panel on how AI is impacting talent in the language industry. Joining him were Jorgelina Venturi, Business Development and Account Manager at tbotalent, and Scott H. Jackson, tbo.’s Head of Innovation. 

AI has already prompted the creation of new roles and necessitated the adaptation of existing jobs now using AI tools. While the expanding application of AI has drawn talent from outside sectors, Venturi and Jackson agreed the noticeable “global talent shortage” in the language industry can be shored up with continuing education for both technical and soft skills.

“The biggest low-hanging fruit in terms of upskilling is just education about how [large language models] work,” Jackson said. “It’s a good idea for people to understand what’s behind the scenes, what’s happening when we’re using LLMs.”

Balancing the Buzz

Faes returned to the stage for a discussion with George Bara, Founder and Chief Strategist of Zetta Cloud, an AI software company that specializes in enterprise and government-grade NLP. 

Bara, who is also contributing to Romania’s national AI strategy as an expert in regulation, expanded on the idea of language technology beyond localization and MT, touching on speech-to-text, OCR, text analytics, data triage, and sentiment analysis, among other tasks. 

The disciplines and tasks that brought ChatGPT to the fore have been in development for more than a decade, Bara said, but GenAI is still in its infancy in terms of market adoption for real-world use cases. 

Perhaps surprisingly, the public sector is a “heavy adopter” of AI due to its massive amounts of data, which can be in various formats, spread across unconnected legacy databases. 

Before GenAI can even be proposed, users want to know and connect their data. And whether public or private, Bara said, “security is still the number one most important aspect when you want to sell something […]. The first question they’re going to ask you is where the data resides, ‘What happens to my data?’”

LanguageWire’s Chief Product and Technology Officer Roeland Hofkens gave attendees a primer on retrieval augmented generation, also known as RAG, a technique that uses prompting to add context and coax more relevant results from LLMs.

Cintia Kanowitz, Senior Lead Localization PM at BOLD, continued the theme of getting back to basics. 

“If there’s a project that can showcase to your company how much we need humans, it’s the first-time right-to-left [RTL] project,” Kanowitz told attendees.

BOLD, which specializes in online resumes and cover letters, is currently localized in 15 languages. Arabic, localized for Saudi Arabia, was BOLD’s first foray into RTL localization and was an experience that has laid the foundation for the company’s future RTL projects.

Expectations vs. Reality

Rocío Txabarriaga, Slator’s Senior Research Analyst and Editor, eased AI back into the spotlight, joined by Phrase CEO Georg Ell and Chief Product Officer Simone Bohnenberger-Rich for a panel on how GenAI can improve LQA. 

Ell reported that Phrase, which has “rapid cycles” between research and productizing, has already begun to integrate AI into the Phrase platform, with a focus on post-editing, review, and personalization. 

“There’s still a little bit of a gap, I think, between expectations and reality,” Ell said of the C-suite’s interest in GenAI. “Sometimes we see that localization teams get reduced budgets in anticipation that GenAI will just lead to cost reductions,” Bohnenberger-Rich added. “That’s a little bit [of] what I call blind enthusiasm.”

Closing out the conference, Faes moderated a panel with Konstantin Savenkov, Co-Founder and CEO of Intento, to discuss the results of Intento’s eighth annual “State of Machine Translation” report. Julie Molino de Alvarenga, e2f Program Manager of Data Services, shared how e2f helped manage data analysis for the report, and together they talked about how LLMs have reshaped the MT landscape.

Bond delivered the closing remarks, thanking panel partners tbotalent and Phrase and inviting the audience to join the Slator team in person again on September 5, 2024, at SlatorCon Silicon Valley.