«

微软用户在Copilot中可选用更多AI模型。

qimuai 发布于 阅读:4 一手编译


微软用户在Copilot中可选用更多AI模型。

内容来源:https://aibusiness.com/foundation-models/microsoft-users-get-more-ai-model-choice-in-copilot

内容总结:

微软生成式AI战略转向:弃“独宠”OpenAI,拥抱多模型生态

近日,科技巨头微软对其企业级生成式AI系统Microsoft 365 Copilot进行了一项重大调整,标志着其AI合作战略发生显著转变。

核心转变:从单一合作到多模型布局

9月24日,微软宣布在Copilot中集成来自其原合作伙伴OpenAI的竞争对手——Anthropic公司的Claude大语言模型。这是微软首次在其核心AI产品中引入非OpenAI开发的模型,尽管微软曾在2019年向OpenAI投资超过150亿美元,并助力其成为行业翘楚。

行业观察家将此举解读为微软在企业AI市场的一次关键布局。Constellation Research首席执行官R“Ray”王表示:“微软先与领导者合作,充分学习后整合所需资源,然后继续前进。当前客户普遍需求是多模型选择。”

行业格局生变,合作阵营重组

这一调整正值生成式AI领域出现阵营重组迹象。几乎与此同时,OpenAI也积极拓展新合作伙伴:先是与甲骨文达成为期五年、价值3000亿美元的计算能力供应合约,随后又从AI芯片巨头英伟达获得了价值1000亿美元的芯片投资。分析师认为,这表明OpenAI在展示其独立发展的能力。

战略意图:巩固企业市场领导地位

分析指出,微软推动Copilot支持多模型,核心目的在于强化其企业AI市场的优势。尽管Copilot仍将继续使用OpenAI模型,但企业客户现在可在特定应用(如研究员推理代理和Copilot Studio平台)中选择使用Anthropic的Claude Sonnet 4和Claude Opus 4模型。

Enterprise Strategy Group分析师马克·贝库指出,微软与谷歌、AWS等其他科技巨头一样,早已在其Azure AI平台提供多模型选择。但此次不同之处在于,这是将多模型能力直接集成到全球领先的企业软件平台——Microsoft 365中,使其易于使用的Copilot系统功能更加强大。

挑战与展望

有分析认为,微软转向以“负责任AI”著称的Anthropic,部分可能是对OpenAI近期动向的反应。尽管Anthropic融资成功,估值达1830亿美元,但仍远低于OpenAI约5000亿美元的估值。

Moor Insights & Strategy分析师安谢尔·萨格表示,此举合乎逻辑,但合作的具体规模与深度仍有待观察。无论如何,微软正通过避免“将所有鸡蛋放在一个篮子里”的策略,为其企业AI业务构建更稳健的未来。

中文翻译:

由谷歌云赞助
如何选择首个生成式AI应用场景
要着手应用生成式AI,首先应聚焦于能够优化人类信息交互体验的领域。

科技巨头迈向多模型战略,引入OpenAI竞争对手的生成式AI模型。
微软将Anthropic模型纳入其广泛应用的Microsoft 365 Copilot系统,这一战略决策标志着其正明显疏远最初的生成式AI合作伙伴OpenAI。在9月24日推出搭载Anthropic Claude大语言模型的新版企业生成式AI系统时,微软首次引入了非OpenAI开发的LLM模型——而微软曾在2019年向OpenAI投资逾150亿美元,助力其成为最知名的独立生成式AI供应商。

"这一趋势已酝酿多时。微软先与行业领导者合作,充分汲取经验并整合所需技术,继而转向新方向,"星座研究公司首席执行官R"雷"王表示,"实质上是大多数客户正在寻求多元化模型方案。"

与此同时,OpenAI与AI芯片巨头英伟达建立了充满活力的新合作。本周初,英伟达以芯片供应形式向ChatGPT制造商投入了1000亿美元。这项惊人举措发生前几日,OpenAI刚与甲骨文签订了为期五年、价值3000亿美元的计算力供应合约,由甲骨文为OpenAI开发和运行新模型提供算力支持。

这些进展共同标志着生成式AI格局开始重组,微软与OpenAI各自拓展了新合作伙伴。
"OpenAI实质上是在向微软宣告:'我们现在独立发展得非常好,感谢先前的支持',"王指出,"当时的合作对微软至关重要,因为其在AI领域落后较多,需要OpenAI的技术加持。而Anthropic将为微软的特定目标提供更优化的模型方案。"

Copilot系统仍将保留OpenAI模型的驱动能力,但现在企业客户还可以在微软研究员推理助手和Copilot Studio企业代理创建平台中选用Claude Sonnet 4及Claude Opus 4.1模型。

观察家视微软此次多模型战略为纯粹的企业级布局,此举将巩固其在企业生成式AI市场的优势地位。反观OpenAI的企业市场抱负似乎陷入停滞——尽管ChatGPT已成为全球使用最广泛的生成式AI聊天机器人,但其近7亿全球用户中多数使用的是免费版本。

"微软早已决定实施多元化策略,避免将AI模型筹码集中于单一篮子,"Informa TechTarget旗下Omdia研究机构企业战略集团分析师马克·贝丘表示。他指出微软与谷歌、AWS等竞争对手一样,长期在Azure AI平台的模型园地中提供多种选择,正如谷歌的Vertex AI和AWS的Amazon Bedrock所做的那样。

但本次举措具有特殊性。"这是在强化嵌入'全球头号企业软件平台'的易用型生成式AI系统Copilot,"贝丘强调微软Office 365的行业地位。

穆尔洞察与战略公司分析师安谢尔·萨格认为,微软转向以"负责任AI"理念闻名的Anthropic,某种程度上可视为对OpenAI近期动作的应对。尽管Anthropic融资成果显著,但其1830亿美元的估值仍远低于OpenAI约5000亿美元的当前价值。

不过,"这一合作具有充分合理性,"萨格补充道,"但合作的具体规模——或者说这种关系变化的实质——仍有待观察。"

您可能还喜欢

英文来源:

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The tech giant goes multi-model with the addition of generative AI models from an OpenAI rival.
Microsoft's strategic decision to add models from Anthropic to its widely used Microsoft 365 Copilot system marks a distinct shift away from its original generative AI partner, OpenAI.
In rolling out the freshly configured enterprise generative AI system with Anthropic Claude large language models on Sept. 24, Microsoft for the first time included LLMs not developed by OpenAI, in which it invested more than $15 billion in 2019 and has helped become the best-known independent generative AI vendor.
"This has been going on for a while. Microsoft partnered with a leader, learned as much as they could and incorporated what they needed to, and then moved on," said R "Ray" Wang, CEO of Constellation Research. "What's really going on is most customers are looking for multiple models."
For its part, OpenAI found a dynamic new partner in AI chip giant Nvidia, which earlier this week invested $100 billion in the ChatGPT maker in the form of chip supplies. That startling move came a few days after OpenAI and Oracle entered into a contract in a five-year, $300 billion deal for Oracle to supply the compute power OpenAI needs to develop and run new models.
Taken together, these developments punctuated the start of what appears to be a realignment of the generative AI landscape, with Microsoft and OpenAI each moving to new partners.
"OpenAI, meanwhile, is saying to Microsoft, 'Look, we're doing really well on our own too, so thank you for your help,'" Wang said. "It was a good partnership at the time because Microsoft was so behind on AI and needed OpenAI's capabilities. What Anthropic will give them is a better model for some of the things they're trying to do."
Copilot will still be powered by OpenAI models, but now Copilot customers can also choose to use Claude Sonnet 4 and Claude Opus 4.1 in Microsoft's Researcher reasoning agent and the Copilot Studio enterprise agent creation platform.
Observers are looking at Microsoft's multi-model move in Copilot as a pure enterprise play, further cementing its strong position in the enterprise generative AI market, while OpenAI's enterprise ambitions appear stalled. Even as ChatGPT has become the most used generative AI chatbot in the world, most of its nearly 700 million or so users worldwide are on the free version.
"Microsoft has for a while been deciding to diversify, so there's not one egg in the basket for them for AI models," said Mark Beccue, an analyst with Enterprise Strategy Group, a division of Informa TechTarget's Omdia research unit.
Beccue noted that Microsoft, like its tech giant competitors Google and AWS, has long offered multiple models in the model garden of its Azure AI platform, just as Google does in Vertex AI and AWS with Amazon Bedrock.
But this move is different. It's buttressing an easy-to-use generative AI system, Copilot, embedded in "the No. 1 enterprise software platform in the world," Beccue said, referring to Microsoft Office 365.
As for downsides for Microsoft, its turn toward Anthropic -- known for its emphasis on responsible AI and AI safety -- could be seen as somewhat reactive to OpenAI's recent moves, said Anshel Sag, an analyst with Moor Insights & Strategy. Anthropic has been hugely successful in fundraising, but its $183 billion valuation is still far short of OpenAI's current value of about $500 billion.
However, "It makes a lot of sense," Sag said. "But I also don't know if we really know the scale of the partnership -- whatever you might call this change of relationship."
You May Also Like

商业视角看AI

文章目录


    扫描二维码,在手机上阅读