AI研发受挫,苹果或采用OpenAI或Anthropic模型赋能Siri

AI研发受挫,苹果或采用OpenAI或Anthropic模型赋能Siri

2025-07-02Technology
--:--
--:--
David
大家好,我是David,欢迎收听 <Goose Pod>。
Ema
嗨,我是Ema!今天我们来聊个热乎的:苹果的AI好像有点卡住了,现在要找OpenAI来给Siri“补课”。这感觉,是不是有点像“自家孩子学习跟不上,只能请个超级外援”?
David
嗯,你这个比喻挺有意思的。说真的,最近这个消息确实很重磅:苹果在考虑一个大动作,就是用OpenAI或Anthropic的模型,来驱动WWDC上那个全新的Siri。
Ema
没错!当时苹果可是画了个大饼,说新Siri能懂我,还能直接操作App。结果呢?现在说要推迟到2026年。这胃口真是被吊得老高了。
David
是啊,这个延迟是关键。报道说,苹果已经在和OpenAI、Anthropic接触了,想让他们帮忙。
Ema
哦?帮忙做什么?专门为苹果定制一套AI系统吗?
David
对,差不多是这个意思。就是训练出能适配苹果“私有云计算”平台的模型。这个平台,专门用来处理那些在手机上跑不动的复杂AI任务。
Ema
等等,把Siri这个亲儿子级别的业务交给外人,这可真不像苹果的风格。他们不是向来喜欢在“自家后院”里种菜,凡事都自己来吗?
David
没错,你问到点子上了。这绝对是个重大的路线偏离。你想,苹果一直主打自家的“苹果基础模型”,还计划2026年全用自研技术。现在突然要找外援,这不就等于承认,自家的研发…嗯…遇到麻烦了嘛。
Ema
我明白了。所以情况就是,苹果画了个大饼,但发现自家的厨房和面点师傅好像跟不上了,于是开始寻思去隔壁五星级酒店直接订购成品。这对我们这些普通用户来说,真是又期待又有点担心。
David
要理解苹果为什么这么纠结,我们得稍微考古一下Siri的历史。很多人可能不知道,Siri的源头,其实不是苹果。
Ema
哦?还有这回事?我一直以为Siri是乔布斯时代的苹果原创产品呢。
David
它最初是2003年斯坦福一个叫CALO的科研项目,背后还有DARPA的资金支持。
Ema
哇,DARPA?就是那个听起来很神秘,研究各种黑科技的美国军事机构?所以Siri一开始竟然有军方背景?这可真是个冷知识。
David
可以这么说。这个项目的目标是创建一个能理解自然语言、会学习的“认知助理”。2007年,核心成员出来成立了Siri公司,想把它商业化,目标是做个“行动引擎”。
Ema
“行动引擎”,这个名字比“语音助手”听起来酷多了!所以它不只是问天气,而是要能真的帮你做事,比如订餐厅、买电影票。这不就是我们现在最想要的吗?
David
完全正确。苹果2010年收购了它,2011年整合进iPhone 4S,当时确实惊艳了世界。可以说,Siri开创了一个时代。
Ema
想当年,那确实是科幻照进现实。但说实话,这么多年过去,Siri总给我一种“起了个大早,赶了个晚集”的感觉。跟后来者一比,总觉得慢半拍。
David
这正是问题的核心。Siri的底层技术,早期依赖外部公司。虽然苹果后来努力自研,但好像一直没完全摆脱最初的框架,导致理解复杂问题时有点力不从心。
Ema
我懂了,就像Siri的“操作系统”太老旧了,核心处理能力没跟上。特别是最近几年,以ChatGPT为代表的大语言模型一出来,AI的能力简直是指数级增长。
David
没错。大模型的出现,让AI从“执行命令”变成了“理解和创造”。你看,三星整合了谷歌的Gemini,微软深度绑定OpenAI。大家都在AI的快车道上飞驰。
Ema
这么说来,苹果如果再完全依赖自研,就可能真的掉队了。所以这次寻求合作,更像是一次迫不得已但又极其关键的“战略补救”。这棋局,真是越来越有意思了。
David
是的。而且这次战略转向的背后,是苹果内部激烈的冲突。最直接的体现就是,原来负责AI战略的高级副总裁John Giannandrea,在Siri项目上被“边缘化”了。
Ema
被边缘化?这可是从谷歌寄予厚望挖来的大人物啊。他出局,说明苹果对Siri的进展确实非常不满意。那现在谁来接这个“烫手山芋”呢?
David
现在由两位高管共同监督:一位是软件工程的Craig Federighi,我们很熟悉了;另一位是Mike Rockwell,他就是成功领导Apple Vision Pro研发的那位。
Ema
哇,让Vision Pro的负责人来管Siri?这个安排太有意思了。这不就是“明星项目经理前来救火”的经典剧情吗?是希望他能给Siri也带来颠覆性的创新?
David
可以这么理解。据说Rockwell上任后,马上让团队评估一个核心问题:新Siri到底是用自家的模型好,还是用第三方的技术更好?这就直接引发了路线之争。
Ema
这个冲突点太关键了。一边是苹果引以为傲的“自主研发”文化,另一边是“时不我待”的AI发展现实。坚持自研可能继续落后,采用外部技术又像在打自己的脸。真是两难。
David
是的。而且,人才流失也是一个因素。有报道说,苹果近年来流失了一些关键的AI工程师,这无疑加剧了研发的困境,也促成了这次变革。
Ema
所以,这不仅仅是技术路线的冲突,更是内部管理、人才战略的全面挑战。就像一艘巨轮要转向,内部的齿轮肯定要承受巨大的压力。苹果这次能否顺利转过来,真的让人捏一把汗。
David
这次延迟和潜在的合作,影响是多方面的。首先,对苹果的品牌形象来说,是个不小的打击。在最热门的AI赛道上,它现在被看作是“落后者”,这很尴尬。
Ema
确实,我作为用户都能感觉到。每次看到朋友手机上那些聪明的AI功能,再看看我的Siri,有时候真的有点“恨铁不成钢”。这会动摇对苹果“科技领袖”地位的信任。
David
对市场竞争的影响就更直接了。AI正在成为手机的核心竞争力。如果苹果在2026年前拿不出像样的Siri,那它在和三星、谷歌的较量中就会非常被动。
Ema
这倒是。但有个很有趣的现象,当苹果考虑和OpenAI合作的消息出来后,它的股价反而上涨了。这是为什么呢?难道投资者不担心苹果的创新能力吗?
David
这是个好问题。这反映了市场的务实。投资者可能认为,一个能用的、强大的、哪怕是第三方的Siri,也比一个不断跳票、性能落后的自研Siri要好。市场更看重结果。
Ema
我明白了。与其守着“自研”的空头支票,不如先给用户一个实实在在的好产品。这就像你承诺要亲手做米其林大餐,但迟迟做不出来,大家宁愿你先点个顶级外卖。对我们用户来说,体验最重要。
David
展望未来,苹果的路径也清晰了一些。今年秋天,我们会先看到一些相对温和的AI功能更新。而真正的重头戏,也就是大模型驱动的新Siri,我们要等到2026年了。
Ema
所以是“小步快跑,逐步到位”。不过我最好奇的是,如果真的用了外部技术,未来的Siri能做到什么地步?是不是能跟它进行真正有深度的对话,让它帮我规划旅行、写邮件?
David
理论上是的。那将是一次质的飞跃。不过,这种做法也并非没有先例。还记得吗?苹果在2012年用自家的地图取代了谷歌地图。
Ema
哦,苹果地图的例子很有启发性!虽然初期问题很多,但最终还是建立了自己的生态。所以这次合作,可能也只是一个过渡方案?先稳住阵脚,再秘密研发,最后“后发制人”?这很符合苹果的风格。
David
嗯,这很可能。总而言之,苹果在AI领域的追赶之路充满了挑战与变数。今天的讨论就到这里。我是David,感谢收听 <Goose Pod>。
Ema
是的,让我们拭目以待,看看Siri能否在2026年给我们带来真正的惊喜。我是Ema,感谢您的收听,我们明天再见!

Of course. Here is a comprehensive summary of the news article, formatted as requested. *** ### **Summary of News Report: Apple's AI Strategy for Siri** This summary details a report from Engadget, based on information from Bloomberg, regarding significant potential changes in Apple's strategy for developing its voice assistant, Siri. * **News Title:** Apple may power Siri with Anthropic or OpenAI models amid AI struggles * **Report Provider:** Engadget (citing Bloomberg) * **Author:** Ian Carlos Campbell * **Publication Date:** June 30, 2025 * **Topic:** Technology / Artificial Intelligence --- ### **Executive Summary** Apple is reportedly considering a major strategic shift by using third-party AI models from **OpenAI** or **Anthropic** to power the next-generation, more capable version of Siri. This move is being explored due to internal development challenges that have led to a delay of the updated Siri until **2026**. The potential partnership and a recent leadership reshuffle within Apple's AI and Siri teams signal that the company is struggling to keep pace with its in-house AI development goals. --- ### **Key Findings and Developments** #### **1. Potential Partnership for a "New Siri"** * Apple is in discussions with OpenAI and Anthropic to use their large language models for a significantly improved version of Siri, which was first previewed at WWDC 2024. * The plan involves having these third-party companies train specialized versions of their models to run on **Apple's Private Cloud Compute**—a network of secure servers powered by Apple's own chips. * This represents a "major departure" from Apple's traditional reliance on in-house technology. The article quotes Bloomberg, stating: *"The company currently powers most of its AI features with homegrown technology that it calls Apple Foundation Models... and had been planning a new version of its voice assistant that runs on that technology for 2026."* #### **2. Development Delays and Timeline Changes** * The advanced version of Siri, capable of understanding personal context and performing in-app actions, has been **officially delayed** from its original timeline. * **New Expected Launch:** The updated Siri is now not expected to launch until **2026**. * **Interim Release:** In the fall, Apple will ship a "more modest collection of AI-adjacent features" with the launch of **iOS 26, iPadOS 26, and macOS 26**. #### **3. Internal Leadership Reorganization** The report highlights significant changes in the leadership of Apple's AI efforts, suggesting a response to internal struggles: * **AI Teams:** Leadership has reportedly shifted from **John Giannandrea** (SVP of Machine Learning and AI Strategy) to **Craig Federighi** (SVP of Software Engineering). * **Siri Team:** The team is now reportedly led by **Mike Rockwell**, who most recently managed the development of the Apple Vision Pro. #### **4. Market Context and Strategic Parallels** * **Competitor Strategy:** An OpenAI or Anthropic-powered Siri would mirror **Samsung's approach with Galaxy AI**, which primarily utilizes Google's Gemini models alongside some custom software. * **Historical Precedent:** The article compares this potential move to Apple's transition away from Google Maps to its own Apple Maps service in 2012. This suggests that using a third-party AI model could be a temporary measure until Apple's in-house solution is fully ready. #### **5. Existing and Future AI Integrations** * Apple already integrates OpenAI's ChatGPT for certain functions within **Apple Intelligence**. However, relying on a third party for the core functionality of Siri would be a much deeper and more significant partnership. * At WWDC 2025, Apple announced it would make its own **Apple Foundation Models** available to third-party developers, indicating a continued, albeit slower, commitment to its proprietary AI technology. *** *Note: The dates mentioned in the article (e.g., publication date of June 2025, WWDC 2025, iOS 26) are futuristic and should be interpreted within the context of the report as written.*

Apple may power Siri with Anthropic or OpenAI models amid AI struggles

Read original at Engadget

Apple is considering using AI models from OpenAI or Anthropic to deliver the more capable version of Siri it debuted at WWDC 2024, Bloomberg reports. The company has promised it could deliver a new version of its voice assistant that understands personal context and takes action inside of apps since last year, but officially delayed the updated Siri in March 2025.

As part of this proposed new plan, Apple has asked Anthropic and OpenAI to train versions of its models that can run on Apple's Private Cloud Compute, secure servers running on Apple chips. The company already relies on its servers for certain AI features that can't be run locally.Apple uses OpenAI's ChatGPT for some parts of Apple Intelligence, but completely relying on a third-party company for Siri would be a major departure.

"The company currently powers most of its AI features with homegrown technology that it calls Apple Foundation Models," Bloomberg writes,"and had been planning a new version of its voice assistant that runs on that technology for 2026." One of the few AI announcements Apple made at WWDC 2025 was to make those foundation models available to third-party developers.

Even considering using third-party AI models reflects internal changes at Apple. Leadership of the company's AI teams has reportedly changed hands from John Giannandrea, Apple's senior vice president of machine learning and AI strategy, to Craig Federighi, the senior vice president of software engineering.

Separately, Bloomberg reports Apple's Siri team is now being led by Mike Rockwell, who most recently oversaw the development of the Apple Vision Pro.As Bloomberg notes, an Anthropic or OpenAI-powered Siri would actually mirror Samsung's current approach to AI. Galaxy AI relies on some custom Samsung software, but primarily uses Google's Gemini.

Using third-party models wouldn't preclude Apple from switching back to something in-house in the future. The company made a similar transition — albeit, perhaps too early — when it went from a Maps app that relied on Google Maps to its custom Apple Maps service in 2012.Wherever Apple lands, the updated version of Siri isn't expected to launch now until 2026.

The company will ship a more modest collection of AI-adjacent features this fall with the launch of iOS 26, iPadOS 26 and macOS 26.

Analysis

Phenomenon+
Conflict+
Background+
Impact+
Future+

Related Podcasts

AI研发受挫,苹果或采用OpenAI或Anthropic模型赋能Siri | Goose Pod | Goose Pod