When we talk about translation tools, Google Translate is the first that comes to mind. It has been used by Internet users for years, and has made a place for itself almost everywhere on the web. Many sites use the Google Translate API to quickly translate text, like Elon Musk's social network, X. Thanks to this integration, users around the world can exchange and debate, even if they don't do not speak the same language.
A startup called DeepL, however, joined the translation market in 2017, with the ambition of surpassing all technology companies in this field, including Google. For this, it relies on artificial intelligence,like Alphabet.DeepL has just presented a new LLMspecially designedfor the translation and revision of content.
An LLM adapted to content creation
The particularity of DeepL's new LLM intended for translation is that it relies on DeepL's proprietary data. Since this data is suitable for content creation and translation, DeepL's LLM would be able to provide better performance.The model was developed with the help of linguistics experts,particularly with regard to the quality of the translation.
2 times more effective than Google Translate?
According to DeepL, its new large language model requires half as much revision as Google Translate, and 3 times less than GPT-4. In other words, if you ask DeepL's new LLM to translate text for you, it will manage to finish in2 times fewer steps than Google Translate, and 3 times fewer steps than OpenAI's most advanced model: GPT-4.
DeepL's claims require verification once the model becomes available to the general public. Given DeepL's already impressive performance, however, we imagine that the startup has not exaggerated. The current version of DeepL is, according to some, already more efficient than Google Translate.Continue reading here.
i-nfo.fr - Official iPhon.fr app
By : Keleops AG
