Core Insights - DeepL is deploying Nvidia's latest systems to significantly reduce the time required to translate the entire internet from 194 days to just 18 days [1] - The company is utilizing Nvidia's DGX SuperPOD, which consists of 36 B200 Grace Blackwell Superchips per server rack, to enhance its AI capabilities [2] - The upgraded infrastructure aims to provide more computational power for developing advanced AI models and improving existing products like Clarify, which focuses on context in translations [3][4] Company Developments - DeepL has developed its own AI models and competes directly with Google Translate [1] - The implementation of Nvidia's technology is part of a broader trend where startups leverage high-end products to create advanced AI applications [2] - The advancements in technology have made it feasible for DeepL to pursue more ambitious projects that were not possible before [4] Industry Trends - Nvidia is expanding its customer base beyond major hyperscalers like Microsoft and Amazon, indicating a growing demand for AI applications across various sectors [1][2] - The focus on computational power and advanced AI models reflects the industry's shift towards more sophisticated AI solutions following foundational models [2]

在瑞士东南部达沃斯小镇举行的首届欧亚经济峰会（EAES）5月28日落下帷幕。此次峰会以"人工智能 （AI）与可持续发展"为主题，中国AI技术成为峰会焦点。与会欧洲企业代表纷纷表示看好与中国企业 的合作前景，认为中国在AI领域的创新突破将为欧中经济合作注入新动能。 在主题为"融合人类智慧与机器智能"的人工智能主旨演讲环节，来自中国的人形机器人甫一亮相便吸引 了全场与会者的目光。随着指令下达，机器人以铿锵有力的步伐在场内自如行走，并流畅完成蹲起动 作，其精准的运动控制和拟人化表现引起阵阵惊叹，观众纷纷举起手机记录。 世界贸易组织副总干事安杰拉·埃拉德在视频讲话中强调，当今世界面临诸多紧迫挑战，这些挑战呼唤 对话、协同行动、重视全球公域的价值。 "我们应该共同强化这一理念：贸易的本质是共赢而非零和。"埃拉德说，"在这个充满不确定性的时 代，通过世贸组织加强合作是明智的商业选择。" 瑞士圣加仑大学经济学讲师斯特凡·莱格接受记者采访时表示，如今还有很多人以零和思维看世界，这 很危险。世界各国应强化共赢思维，寻找各方共赢的机会。 为期两天的峰会上，包括华为云和阿里云在内的中国企业代表与来自欧洲、亚洲的跨国公司、中小企业 ...

Core Insights - DeepL has announced significant updates to its API, introducing a self-developed AI language model and a writing assistant called DeepL Write, aimed at enhancing translation and writing support capabilities [1][2]. Group 1: DeepL Write - DeepL Write is positioned as a professional writing assistance tool, similar to Grammarly, focusing on improving text quality rather than generating new content [2]. - The tool provides real-time language optimization suggestions, including grammar correction, wording adjustments, and style adaptation, helping users produce more standardized and fluent text [3]. Group 2: AI Language Model - The newly developed language model primarily serves translation scenarios, correcting and optimizing translation results [3]. - It demonstrates deep semantic understanding, allowing for accurate recognition of proper nouns and culturally specific expressions, significantly enhancing reliability in complex translation tasks, especially for specialized terminology and sensitive content [3]. Group 3: API and Language Support - The DeepL API now supports 33 languages, including English, Dutch, Swedish, Japanese, German, Korean, and Simplified Chinese, covering major global language systems [3]. - DeepL has committed to not using user-submitted content for model training, ensuring data security and privacy protection from both technical and operational perspectives [3].