EasyTranslate is using generative AI to drill down to client-specific language models

Earlier this year Copenhagen-based content creation and translation platform EasyTranslate raised €2.75 million in a funding round that was aimed at adding new functionalities to its marketplace for freelance translators, including AI-generated content capabilities. 

This announcement was only a few months after Open AI unleashed ChatGPT to the world, and now EasyTranslate is going all-in on generative content, but perhaps not in the way you might expect.

While the powers of AI-generated content certainly have their advantages, one of the keys to getting things “right” right now is knowing precisely how to prompt the machine in order to get the best results. In an overly laden world of content, getting the mix right, as well as retaining brand speak this prompting process can be the stuff of headaches and heartaches.

Where EasyTranslate is heading is effectively sunsetting its translation services and instead angling to offer customers SEO-optimised AI-generated content in any language, with ‘humans in the loop’, that is, copywriters, doing edits atop what the AI churns out.

What this effectively means for EasyTranslate customers, which include Monday, Dyson, Wix.com, wagamama, and Acer, is that not only is their content creation process massively accelerated, but thanks to the human edits, over time these LLMs, or rather perhaps better thought of as SLMs will be trained to offer better and more localised content while at the same time matching specific customers’ tone of voice.

Now if you’re thinking along the lines as I did in that this approach will be putting copywriters, i.e. those training the LLM, out of a job, EasyTranslate founder CEO Frederik R. Pedersen responded, “I don’t think that copywriters will at any point be out of a job, actually quite the opposite. This will only represent a huge growth of content, and creating engaging content, so I think they will be very busy to meet this supply and demand.”

And I’m inclined to believe him. For now. When pressed further, Pedersen did admit, “Yes, I think eventually at some point the profession might change. This is of course a question of security levels for example.”

Right now EasyTranslate is using GPT engines, but a portion of that $3+ million funding round has been earmarked to build out their own infrastructure of LLMs trained on customer specifics, a project that according to company founder and CEO Frederik R. Pedersen will reach better levels of accuracy and less hallucination than using OpenAI’s GPT4 out of the box.

While Frederik does admit that there is plenty of competition out there, he firmly believes where EasyTranslate can stand out from the crowd is through its ‘human in the loop’ process and offering services at a price point that enables clients to build up an AI solution specifically tailored for their needs, ultimately becoming a defacto part of the business process.