Google推出革命性 Trillium 晶片和超大 Gemini 1.5 Pro，顛覆人工智慧發展

隨著高效的第六代張量處理單元 Trillium 加入資料中心晶片系列，Google的人工智慧不斷取得進步，效能提升了近五倍。此外，Google 還推出了 Gemini 1.5 Pro，該版本具有包含 200 萬個令牌的擴展上下文窗口，可供開發人員增強檔案分析和複雜模型訓練。

Google Unveils Cutting-Edge Trillium Chip and Supersized Gemini 1.5 Pro, Empowering Developers with Unprecedented AI Capabilities

Google推出尖端Trillium晶片和超大尺寸Gemini 1.5 Pro，為開發者提供前所未有的AI能力

Mountain View, California, May 15, 2024 - In a resounding testament to its relentless pursuit of artificial intelligence innovation, tech behemoth Google today unveiled two groundbreaking advancements that promise to revolutionize the development and deployment of AI applications: the Trillium chip and Gemini 1.5 Pro.

加州山景城，2024 年5 月15 日- 科技巨頭谷歌今天推出了兩項突破性的進步，有望徹底改變人工智慧應用的開發和部署：Trillium 晶片和Gemini 1.5 Pro 充分證明了其對人工智慧創新的不懈追求。

Trillium: A Quantum Leap in AI Processing Power

Trillium：人工智慧處理能力的巨大飛躍

With the relentless march of technology, the demand for machine learning computing has skyrocketed by a staggering million-fold over the past six years, and the trajectory continues to soar exponentially. Recognizing this insatiable appetite for computational prowess, Google has poured its considerable resources into developing the sixth generation of its Tensor Processing Units (TPUs), culminating in the groundbreaking Trillium chip.

隨著技術的不斷進步，機器學習計算的需求在過去六年中激增了驚人的數百萬倍，而這一趨勢仍在指數級飆升。認識到人們對運算能力的永不滿足的需求，Google投入了大量資源來開發第六代張量處理單元 (TPU)，最終開發出了突破性的 Trillium 晶片。

Trillium, the latest iteration in Google's TPU family, boasts a remarkable 4.7-fold improvement in compute performance per chip compared to its predecessor. This extraordinary leap in efficiency will empower developers to tackle even more complex AI challenges, paving the way for the next generation of transformative applications.

Trillium 是 Google TPU 系列的最新版本，與前代產品相比，每個晶片的運算效能顯著提高了 4.7 倍。這種效率的非凡飛躍將使開發人員能夠應對更複雜的人工智慧挑戰，為下一代變革性應用程式鋪平道路。

Gemini 1.5 Pro: Pushing the Boundaries of Language Comprehension

Gemini 1.5 Pro：突破語言理解的界限

Google's AI prowess extends beyond hardware, as evidenced by the unveiling of Gemini 1.5 Pro. This latest version of the Gemini multimodal foundation model pushes the boundaries of language comprehension with an expanded context window of two million tokens, double the capacity of its previous incarnation.

谷歌的人工智慧實力超越了硬件，Gemini 1.5 Pro 的發布證明了這一點。 Gemini 多模態基礎模型的最新版本透過擴展了 200 萬個標記的上下文窗口，突破了語言理解的界限，是先前版本容量的兩倍。

This massive context window allows Gemini 1.5 Pro to ingest and process vast amounts of information, enabling it to make more accurate predictions and generate more coherent and insightful responses. Developers now have the unparalleled ability to analyze dense files like videos and code, opening up new possibilities for AI-driven applications in fields such as content creation, data analysis, and software development.

這個巨大的上下文視窗使 Gemini 1.5 Pro 能夠攝取和處理大量訊息，使其能夠做出更準確的預測並產生更連貫和富有洞察力的回應。開發人員現在擁有無與倫比的能力來分析影片和程式碼等密集文件，為內容創建、數據分析和軟體開發等領域的人工智慧驅動應用程式開闢了新的可能性。

AI Hypercomputer: Supercharging AI Innovation

AI超級電腦：加速AI創新

Google's AI Hypercomputer, a massively scalable supercomputing architecture, harnesses the combined power of Trillium chips, CPUs, and GPUs. This heterogeneous approach provides developers with the flexibility to tackle any workload, from training massive language models to solving complex scientific problems.

谷歌的人工智慧超級電腦是一種大規模可擴展的超級運算架構，利用了 Trillium 晶片、CPU 和 GPU 的綜合能力。這種異質方法為開發人員提供了處理任何工作負載的靈活性，從訓練大規模語言模型到解決複雜的科學問題。

The AI Hypercomputer delivers remarkable efficiency gains of over twice that of simply purchasing raw hardware and chips, empowering businesses and developers to achieve their AI aspirations with unparalleled cost-effectiveness.

AI超級電腦的效率提升是簡單購買原始硬體和晶片的兩倍以上，使企業和開發人員能夠以無與倫比的成本效益實現他們的AI願望。

Project Astra: A Vision of the Future

阿斯特拉計畫：未來願景

Google is not content to rest on its laurels, as evidenced by the unveiling of Project Astra, an AI assistant that possesses the remarkable ability to remember what it sees, process that information, and understand contextual details.

谷歌並不滿足於固步自封，Project Astra 的推出證明了這一點。

Project Astra's capabilities were showcased in a live demonstration, where a prototype AI agent seamlessly interacted with a human tester through a Google Pixel phone. The AI agent took in a continuous stream of audio and video input, reasoning about its environment in real time and engaging in a natural conversation with the tester.

Project Astra 的功能在現場演示中得到了展示，原型 AI 代理透過 Google Pixel 手機與人類測試人員無縫互動。人工智慧代理接收連續的音訊和視訊輸入串流，即時推理其環境並與測試人員進行自然對話。

Gemma: Empowering Indian Developers

Gemma：為印度開發者賦權

Google's commitment to AI innovation extends beyond its own products, as evidenced by its Gemma family of open language models. Gemma has empowered Indian developers to create large language models in Indic languages, fostering the development of AI applications that are tailored to the diverse linguistic landscape of India.

谷歌對人工智慧創新的承諾超越了自身產品的範疇，其 Gemma 系列開放語言模式證明了這一點。 Gemma 讓印度開發人員能夠用印度語言創建大型語言模型，促進針對印度多樣化語言環境的人工智慧應用程式的開發。

A Glimpse into the Future

未來一瞥

These latest advancements from Google represent a giant leap forward in the realm of AI technology. Trillium's unparalleled processing power, Gemini 1.5 Pro's expanded context window, the AI Hypercomputer's supercomputing architecture, Project Astra's groundbreaking capabilities, and Gemma's focus on inclusivity empower developers to create AI applications that can solve complex problems, drive innovation, and improve lives across the globe.

谷歌的這些最新進展代表了人工智慧技術領域的巨大飛躍。 Trillium 無與倫比的處理能力、Gemini 1.5 Pro 擴展的上下文視窗、AI 超級電腦的超級運算架構、Project Astra 的突破性功能以及Gemma 對包容性的關注，使開發人員能夠創建能夠解決複雜問題、推動創新並改善全球生活的AI 應用程式。

As Google continues to push the boundaries of AI, one can only imagine the transformative applications that will emerge from this fertile ground of innovation. The future of AI is bright indeed, and Google is leading the charge.

隨著Google不斷突破人工智慧的界限，人們可以想像從這片創新的沃土中將會出現哪些變革性的應用程式。人工智慧的未來確實是光明的，而谷歌正在引領這一潮流。