AI News AI资讯 7d ago Updated 7d ago 更新于 7天前 50

Apple Intelligence gets a second shot with help from Google and Nvidia Apple Intelligence 在 Google 和 Nvidia 的帮助下获得第二次机会

The ghost of Steve Jobs must be haunting the halls of Cupertino. Apple just stood on stage at WWDC and announced that the future of Siri—the digital assistant it has owned, neglected, and bungled for over a decade—now runs on foundation models built with Google and crunches complex queries on Nvidia GPUs. This isn't just a partnership. It's a capitulation. A full-throated admission that for all its talk of vertical integration and owning the entire stack, Apple cannot, in fact, build a world-cla 史蒂夫·乔布斯的魂灵想必正萦绕在库比蒂诺的走廊里。苹果刚刚在WWDC舞台上宣布,Siri的未来——这个被其拥有、忽视并搞砸了十多年的数字助手——如今将运行在与谷歌合作构建的基础模型上,并依靠英伟达的GPU处理复杂查询。这不仅仅是合作,这是一份投降书。苹果全盘托出:尽管常年高谈垂直整合与技术栈全掌控,事实上它无法独自构建一个世界级的AI大脑。

80
Hot 热度
65
Quality 质量
70
Impact 影响力

Analysis 深度分析

The ghost of Steve Jobs must be haunting the halls of Cupertino. Apple just stood on stage at WWDC and announced that the future of Siri—the digital assistant it has owned, neglected, and bungled for over a decade—now runs on foundation models built with Google and crunches complex queries on Nvidia GPUs. This isn't just a partnership. It's a capitulation. A full-throated admission that for all its talk of vertical integration and owning the entire stack, Apple cannot, in fact, build a world-class AI brain on its own.

Let's be clear: this is a stunning reversal of Apple's core ethos. For years, the company's moat was built on the premise of a tightly controlled ecosystem where silicon, software, and services were designed in concert, offering a seamless, private, and superior experience. "What happens on your iPhone stays on your iPhone" was more than a slogan; it was a structural promise. Now, for the most consequential new feature in its OS in years, Apple is piping in intelligence from Mountain View and compute from Santa Clara. The "Apple Intelligence" branding starts to feel less like a promise and more like a fig leaf.

Why? Because Siri has always been the embarrassing cousin at the smart assistant family reunion. While Amazon's Alexa and Google Assistant were learning skills and holding conversations, Siri was still struggling to set a timer reliably. Apple's approach to machine learning, while potent in specific, useful applications like photo recognition and predictive text, clearly hit a wall when it came to the broad, generative capabilities now demanded by users. Training a frontier model is a trillion-dollar game of scale, data, and talent. Apple, for all its riches, apparently decided it was cheaper, faster, and more effective to outsource the core intelligence rather than play catch-up from three laps behind.

The choice of partners is telling. Google is, obviously, the king of the castle when it comes to large language models. Gemini is formidable. By licensing Google's models, Apple gets instant state-of-the-art capability without the years of R&D and the colossal training bills. It's a pragmatic, desperate move. And leaning on Nvidia's H100s or whatever Blackwell-class silicon is now in play for heavy lifting? That's an open admission that Apple's own M-series chips, for all their power efficiency, aren't yet the workhorses for this new class of AI inference at the scale needed for a billion devices. It’s a tacit acknowledgment that the AI revolution is, for now, running on Nvidia's playbook.

This partnership reveals the new, uncomfortable power dynamics of tech. Apple, the world's most valuable company, needs Google's brain and Nvidia's brawn. Google, in turn, gets its AI models embedded at the OS level on hundreds of millions of premium devices—a distribution win it could only dream of. Nvidia solidifies its position as the indispensable picks-and-shovels provider for the entire industry. It's a realignment that makes the old Android vs. iOS rivalry feel quaint. The real battle is now over who supplies the foundational intelligence layer, and Apple has just declared itself a customer.

The privacy implications are a minefield. Apple will undoubtedly wrap this in its privacy rhetoric, claiming on-device processing and secure enclaves. But the fundamental architecture has changed. Queries that are "complex" will leave your device. Where do they go? What are Google's and Nvidia's logging policies? Can Apple credibly promise "privacy" when the core cognitive engine is a black box from a company whose entire business model is built on data aggregation? The clean, simple story of on-device processing is over. Welcome to the messy, federated reality of cloud-dependent AI.

Is this a good thing for users? Probably, in the short term. A suddenly competent Siri, powered by top-tier models, could be transformative. The dream of a truly helpful, context-aware assistant might finally arrive. But it comes at the cost of Apple's soul. The company that sold us on independence and integrated excellence is now an integrator of other people's excellence. It's a sign that the AI race has become so ferocious, so capital-intensive, that even the world's richest company feels it can't build everything itself.

This move feels less like a confident step forward and more like a forced adaptation to a new world order. Apple didn't lead the generative AI charge. It was caught flat-footed, and now it's paying the "tax" of partnership to get back in the game. The golden handcuffs are now on. Siri may get better, but Apple's path is no longer fully its own. The future of AI on the iPhone will be written in part by its fiercest competitors. That's not innovation. That's survival.

史蒂夫·乔布斯的魂灵想必正萦绕在库比蒂诺的走廊里。苹果刚刚在WWDC舞台上宣布,Siri的未来——这个被其拥有、忽视并搞砸了十多年的数字助手——如今将运行在与谷歌合作构建的基础模型上,并依靠英伟达的GPU处理复杂查询。这不仅仅是合作,这是一份投降书。苹果全盘托出:尽管常年高谈垂直整合与技术栈全掌控,事实上它无法独自构建一个世界级的AI大脑。

史蒂夫·乔布斯的魂灵想必正萦绕在库比蒂诺的走廊里。苹果刚刚在WWDC舞台上宣布,Siri的未来——这个被其拥有、忽视并搞砸了十多年的数字助手——如今将运行在与谷歌合作构建的基础模型上,并依靠英伟达的GPU处理复杂查询。这不仅仅是合作,这是一份投降书。苹果全盘托出:尽管常年高谈垂直整合与技术栈全掌控,事实上它无法独自构建一个世界级的AI大脑。

必须明确:这是对苹果核心理念的一次惊人逆转。多年来,该公司的护城河建立在一个严密封装的生态之上——芯片、软件与服务协同设计,提供无缝、私密且卓越的体验。“iPhone上发生的事就留在iPhone上”不仅是广告语,更是一种结构性承诺。如今,在其操作系统数年来最具颠覆性的新功能上,苹果却从山景城引入智能,从圣克拉拉调用算力。“苹果智能”这个品牌听起来越来越不像承诺,更像一块遮羞布。

为何如此?因为Siri历来是智能助手家族聚会里令人尴尬的表亲。当亚马逊Alexa和谷歌助手在学习技能、流畅对话时,Siri甚至仍难可靠地设置计时器。苹果在机器学习领域的路径——尽管在图像识别、预测文本等特定实用场景中效果显著——在应对用户当前渴求的广泛生成式能力时显然撞墙了。训练前沿模型是关乎规模、数据与人才的万亿级游戏。苹果尽管富可敌国,最终显然判定:与其落后三圈奋力追赶,不如外包核心智能来得更经济、快捷、有效。

合作伙伴的选择颇具深意。谷歌显然是这座城堡的王...

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

产品发布 产品发布 大模型 大模型 GPU GPU
Share: 分享到: