Translation tools are improving drastically thanks to artificial intelligence (AI), and this new entrant on the market could still be a game-changer. His name is Lara.
Lara justifies her choices
This solution was designed by Translated, a Rome-based company specializing in translation. Introduced to great fanfare in early November, Lara was trained with 1.2 million hours of NVIDIA GPU, and articulates its translation choices using a collection of 25 million real-world translations, produced by leading professional translators.
According to its creators, the tool outperforms the most popular machine translation systems, namely Google Translate, DeepL and even GPT-4o, one of the OpenAI models that power ChatGPT. “ Lara is an important step in our mission to enable everyone to understand and be understood in their language “, explains Marco Trombetti, co-founder and CEO of Translated.
Because Lara’s strength lies in her ability to not only translate text, but also to adapt to the context. The translator also explains why one word is used rather than another.
Solid experience
The company already has significant experience in the field. Founded in 1999, it has established itself as a pioneer in the use of AI to help professional translators. In 2017, she created commercial, context-specific machine translation (MT) with ModernMT. Its solutions are now used by more than 300,000 customers around the world, including giants like Airbnb, Uber and Glovo.
“ Language has been the driving force of human evolution, allowing us to understand each other, collaborate for a better future, and develop faster than any other species. From the experience we have gained from our clients, we believe that once technology approaches human standards, the demand for AI-assisted translation (100x) and human translation (10x) will increase . By fostering global understanding, Lara helps us progress toward the next phase of human evolution », explains Marco Trombetti.
Soon the singularity of the language
Translated claims that Lara consistently demonstrates superior accuracy, making fewer errors than professional translators. For their part, its leaders believe that technology will achieve language singularity next year, that is to say the ability to generate language that is perfectly natural and authentic, as if it had been produced by a native speaker.
Lara is already available in beta, for both personal and professional use, only in English, French, German, Italian, Japanese and simplified and traditional Chinese for the moment. Ultimately, 200 languages should be supported. Translated offers several prices depending on the user’s needs, with a free version however limited.
