On March 30, just a few weeks after the bombshell of GPT-4 echoed through our world, several language executives and LangOps experts met to discuss the future of multilingual content. The questions were many and the discussions vibrant. Will AI eliminate the translation industry and dissect our brain? Let´s get serious, not happening until… who can really tell? Till then we have a lot of work to do, ensuring our future in a good way.
Jochen Hummel and Britta Aagaard kicked off the meeting with a presentation of what the future likely entails. The latest LLMs are at par with the invention of the letterpress. Gutenberg eliminated the painfully slow work of monks copying books. The church was desperate to stop this sacrilege, since copying was a profitable endeavour as well as interpreting of the bible was left to influencers – priests. In a similar way many negative voices fear now again progress.
In Berlin it was clear that LangOps Pioneers are forward looking and embracing change, as they presented their views and why they were participating. During the afternoon the participants were split into three focus groups working on key use cases selected in the morning, namely Translation and Automation, Data Centric Hub, and Enterprise Applications based on multilingual data.
LangOps Use Cases
The first group focussed on the lack of interoperability as the main bottleneck for automation. Their moderator Arthur Wetzel presented the conclusions of the group. Tool providers seem to pride themselves by providing tools that can’t communicate and thus not work together. Even the same provider doesn’t seem to be able to make their own tools interoperable. This makes it hard for LSPs always having to provide a work around. Semantic interoperability that should be natural in the language industry is hardly heard of.
The second group, Data Centric Hub, worked on centralising a multilingual resource and using it to build innovative solutions. Moderator Mathjis Sonnemanns presented their findings with great enthusiasm. The Hub needs to be secure, well maintained, and will underpin the creation of customised AI solutions for the customer, of course using their own customer data.
The third group, was very innovative and stretched the limits of the potential for multilingual content. Their moderator Krzysztof Zdanowski presented the group findings and who’s to say what happens when multilingual content is truly activated.
In the final discussions it was decided to continue the work via a Slack group. Workshops participants will further develop and describe their use cases to be integrated later into the LangOps Manifesto.
There is reason for thinking the future is bright and promising. After all LangOps principle #4 give us a key answer for the future: Try AI first!