DeepLa company specializing in the field of linguistic artificial intelligence, has unveiled its latest machine translation model based on a highly specialized LLM technologywhich promises to “outperform” tech giants like Google, Microsoft and ChatGPT-4.
The new generation of language models has been developed specifically for translation and proofreading of texts and, according to the tests conducted, language experts preferred DeepL translations 1.3 times more often than Google Translate, 1.7 times more than ChatGPT-4, and 2.3 times more than Microsoft.
DeepL “outclasses” competitors in machine translation
The strength of the new DeepL model lies in its specialization. While many general-purpose AI systems rely on publicly available data on the internet, DeepL leveraged over seven years of proprietary data optimized for content creation and translation. This approach reduces the risk of generating “hallucinations” or misinformation, a common problem in general-purpose AI models. DeepL also uses thousands of language experts to ensure the highest quality of translations produced through human review of the results. Its hybrid approach, which combines the efficiency of AI with the accuracy of human experience, aims to set a new standard in machine translation.
Jarek KutylowskiCEO and founder of DeepL, emphasized the company’s commitment to research and development of high-quality language solutions:
“DeepL’s superior linguistic AI solutions for enterprises are just getting started.
We are a research-driven company and are investing heavily in redefining quality, efficiency and adaptability in translation and copywriting. Our goal is to provide our customers around the world with the cutting-edge technology they need to thrive and grow globally without language barriers getting in their way.”
According to tests conducted by DeepL, the experts involved in the comparison with the competitor’s systems had the need to Make twice the changes with Google technologies, triple with ChatGPTcompared to the new DeepL model. In short, using DeepL’s system can achieve a significant increase in productivity for companies that handle large volumes of translations. The new model also showed significant improvements compared to previous versions of DeepL, with the company able to detect a 1.7x increase in quality for translations between English and Japanese/Simplified Chinese, and a 1.4x increase for English-German.
Currently, translations with the new model are available for DeepL Pro customers working with English, Japanese, German and Simplified Chinesewith plans to expand to other languages in the near future. The company ensures high standards of data security, with GDPR compliance and ISO 27001 and SOC 2 type 2 certifications.
The new announcement comes amid a year of significant growth for DeepL, which recently introduced “DeepL for Enterprise” and now counts over 100,000 enterprises, government agencies and organizations worldwide among its customers, including half of the Fortune 500. DeepL’s success was further confirmed by a recent funding round that raised $300 million led by venture capital firm Index Ventures, valuing the company at $2 billion.
Source: www.hwupgrade.it
