Llama 4 Metaverse中的新教育應用新時代

隨著薈萃分析的不斷發展，一份新報告表明，教育應用可能是這項新興技術的重要增長領域。

The metaverse is quickly expanding into new applications, and a new report suggests that educational applications could be a significant growth area for this emerging technology.

元媒體正在迅速擴展到新的應用程序，一份新報告表明，教育應用可能是這項新興技術的重要增長領域。

Metaverse in education is poised to attain a market valuation of $69.4 billion by 2032, commencing from its market capitalization of $3.85 billion in 2024. This exponential surge during the forecast period is fueled by a substantial compound annual growth rate (CAGR) of 37.90%.

到2032年，教育中的元教育計劃將獲得694億美元的市場估值，從2024年的市值開始為38.5億美元。在預測期內，這種指數激增源於37.90％的實質性複合年度增長率（CAGR）。

Several key factors are driving the growth of metaverse in education over the next seven years. At the forefront is the burgeoning demand for innovative learning experiences. As educators and students seek to move beyond the confines of traditional classrooms, the pursuit of fresh and engaging educational formats has become increasingly critical.

在未來七年中，有幾個關鍵因素是推動教育中元元的增長。最前沿的是對創新學習經驗的新興需求。隨著教育者和學生尋求超越傳統教室的範圍，對新鮮而引人入勝的教育格式的追求變得越來越重要。

This trend has been significantly amplified by the COVID-19 pandemic, which has catalyzed a widespread shift towards remote and hybrid learning models. In response to this evolving landscape, institutions are now actively seeking to create a virtual campus environment.

Covid-19大流行極大地擴大了這一趨勢，該大流行極大地促進了向遠程和混合學習模型的廣泛轉變。為了回應這種不斷發展的景觀，機構現在正在積極尋求創造虛擬校園環境。

The goal is to foster novel immersive experiences for teachers, lecturers, and students, complete with interactive lectures, seminars, and tutorials that will revolutionize the way knowledge is imparted and absorbed.

目的是為教師，講師和學生培養新穎的沉浸式體驗，並配有互動講座，研討會和教程，這些講座將徹底改變知識的傳授和吸收方式。

Hardware will record an impressive double-digit CAGR, with AR, VR, and MR devices expected to become mainstream. On the flip side, the software will record astronomical growth levels, with extended reality software and gaming engines taking up the lion’s share.

硬件將記錄令人印象深刻的雙位複合年增長率，而AR，VR和MR設備預計將成為主流。另一方面，該軟件將記錄天文學的增長水平，擴展的現實軟件和遊戲發動機佔據了獅子的份額。

In terms of adoption of applications, learning and skills development will be the biggest drivers, with educational apps, cultural understanding, and self-regulation contributing their fair share. Apart from schools, corporations will contribute a chunk to the market share of the metaverse in education.

就應用的採用而言，學習和技能發展將是最大的驅動力，具有教育應用程序，文化理解和自我調節，從而貢獻了公平的份額。除了學校外，公司還將為教育元元的市場份額做出貢獻。

North America, led by the United States and Canada, will lead markets in education while the Asia Pacific will record the fastest CAGR in the forecast period. Industry leaders include Adobe, Devden, Fotonvr, Hatchxr, Lenovo (NASDAQ:LNVGF), Meta (NASDAQ:META), and Samsung Electronics, among others.

由美國和加拿大領導的北美將領導教育市場，而亞太地區將記錄預測時期最快的複合年增長率。行業領導者包括Adobe，Devden，Fotonvr，Hatchxr，Lenovo（NASDAQ：LNVGF），Meta（NASDAQ：META）和Samsung Electronics等。

The report notes that achieving a market capitalization of nearly $70 billion in seven years will be an uphill climb. Analysts point out that the technology used in the metaverse is still nascent and will require significant investment for large-scale implementation.

該報告指出，在七年中實現近700億美元的市值將是艱難的攀登。分析師指出，元元中使用的技術仍然很新生，將需要大量投資才能進行大規模實施。

Furthermore, the report highlights the steep implementation costs on the infrastructure and content creation aspects of deploying metaverse in education.

此外，該報告強調了在教育中部署元元的基礎設施和內容創建方面的巨大實施成本。

Several governments, including South Korea, have invested large sums to develop metaverse use cases. The goal of these institutions is to generate solutions for the problems of VR/AR hardware and software that are currently hindering the mainstream adoption of metaverse in the education sector.

包括韓國在內的幾個政府已經投資了大筆款項來開發元用例。這些機構的目的是為VR/AR硬件和軟件的問題生成解決方案，這些解決方案目前正在阻礙教育領域的主流採用。

Meta ups AI challenge with Llama 4 models

Meta UPS AI挑戰與Llama 4模型

Meta (NASDAQ:META) has launched four new large language models (LLMs), including Llama 4 Scout, Maverick, and two versions of its flagship LLM, named Behemoth. The models, which are available commercially, are powered by 17 billion active parameters.

Meta（NASDAQ：META）推出了四種新的大型語言模型（LLMS），包括Llama 4 Scout，Maverick和兩個版本的旗艦LLM，名為Bememoth。這些模型在商業上可用，由170億個活動參數提供動力。

The models have key differences. Scout has 16 experts and fits in an Nvidia (NASDAQ:NVDA) H100 GPU, which Meta says is the “best multimodal model in its class.”

這些模型具有關鍵差異。偵察兵擁有16位專家，並適合NVIDIA（NASDAQ：NVDA）H100 GPU，Meta說，這是“同類產品中最好的多模式”。

However, the Llama 4 Maverick has 128 experts but surpasses GPT-4o and Gemini 2.0 across several benchmarks. Despite being compact, Meta’s statement notes that Maverick has the capabilities of DeepSeek’s coding abilities and scores highly on performance-to-cost ratio.

但是，Llama 4 Maverick擁有128位專家，但在幾個基準中超過了GPT-4O和Gemini 2.0。儘管非常緊湊，但梅塔的聲明指出，小牛具有DeepSeek編碼能力的能力，並且在性能與成本比率上得分很高。

Llama 4 Behemoth, Meta’s most powerful LLM, is still in development. Upon launch, the company says it will outclass OpenAI’s latest LLM. Given its design as a teacher model, both Maverick and Scout lean on the incoming Llama 4 Behemoth for distillation.

Llama 4 Bememoth是Meta最強大的LLM，仍在開發中。該公司表示，它將超出OpenAI的最新LLM。鑑於其作為教師模特的設計，小牛和偵察兵都依靠即將到來的駱駝4龐然大物進行蒸餾。

“These models are our best yet thanks to distillation from Llama 4 Behemoth, a 288 billion active parameter model with 16 experts that is our most powerful yet and among the world’s smartest LLMs,” said Meta.

“這些模型是我們最好的，這是我們最好的，這要歸功於Llama 4 Behemoth的蒸餾，這是一個2880億個活躍參數模型，擁有16位專家，這是我們迄今為止最聰明的LLM中最強大的專家，” Meta說。

Despite the positive statement, early users of Meta’s Llama 4 models have criticized the LLMs for falling short on key metrics. EQ-Bench maintainer Sam Paech noted that “Llama-4 is performing not so well” in long-form writing benchmarks.

儘管有積極的陳述，但Meta的Llama 4模型的早期用戶還是批評了LLMS缺乏關鍵指標。 EQ-Bench維護者Sam Paech指出，長期編寫基準的“ Llama-4表現不佳”。

Furthermore, independent evaluations of the models reveal an inability to retrieve information at 300K token context length. However, at 85K tokens, Llama-4 showed glimmers of promises but failed to impress in standard logical puzzles.

此外，對模型的獨立評估揭示了無法在300K令牌上下文長度上檢索信息。但是，在85K令牌時，Llama-4表現出諾言的一線，但在標準邏輯難題中未能給人留下深刻印象。

Despite the shortcoming, Llama-4 was retrofitted with censorship guardrails, but its open-source nature may see users tweak it in the coming days.

儘管存在缺點，但Llama-4還是通過審查護欄進行了改造，但其開源性質可能會在未來幾天內調整用戶。

While blockchain is not a priority for Meta, the company is not slowing down on other emerging technologies. AI is the center of its digitalization pivot, with the technology giant racing to integrate its LLMs into its social media platforms.

儘管區塊鏈不是元鏈的優先事項，但該公司並沒有放慢其他新興技術。 AI是其數字化樞紐的中心，技術巨頭賽車將其LLMS整合到社交媒體平台中。

Outside of AI, Meta has its sights on enterprise

在AI之外，Meta對企業有目光