subscribe and listen
It seems practical to outsource the heaviest AI lifting to Google. This buys Apple time to see how the generative AI market changes. Photo/Getty Images
At first glance, Apple’s decision to pay for Google’s Gemini artificial intelligence model to power its new Apple Intelligence features on iPhones and other devices is a concession that its own generative AI efforts have failed. For companies that have built their brand by owning
This looks like a rare climb down when you target the entire stack, from custom silicon to meticulously crafted software.
This is embarrassing, since Apple has spent years insisting that it tell its own AI story. The M-series chips and Neural Engine in recent iPhones were touted as evidence that Apple was building serious machine learning capabilities into its hardware. But when generative AI exploded into the public consciousness in late 2022 through OpenAI’s ChatGPT and Google’s Gemini, Apple was largely on the sidelines.
Last year, serious problems in the company’s AI division were revealed. The company had a deal with OpenAI to integrate ChatGPT into its AI capabilities, but the results were lackluster at best. Now, the company is helping Google develop an upgraded AI-powered Siri assistant, effectively admitting that its internal OpenAI-backed efforts have not yielded results.
But embarrassment and strategic error are not the same thing. Apple’s choice to rely on Google is one of the smartest decisions the company has made in recent years. To compete head-on with Google and OpenAI in the AI space, Apple will need to divert vast resources away from what actually makes iPhones, Macs, and iPads attractive: integrating hardware and software, and designing interfaces that billions of people can use without a manual.
Apple’s great strength isn’t that it’s the first to develop new technology, or that it’s the best at making it easier to use. The iPod wasn’t the first MP3 player, the iPhone wasn’t the first smartphone, and the Apple Watch wasn’t the first smartwatch. In each case, Apple has taken messy, fragmented technology and wrapped it into an experience that feels coherent, intuitive, and polished.
GGenerative AI is now the opposite: while spectacular, it is often chaotic and inconsistent. The value Apple can add is in turning a raw model like Gemini into a calm and useful feature.
Apple’s brand also has aspects of privacy and trust. Consumers are rightly wary of AI systems that fabricate personal data and occasionally confidently hallucinate nonsense.
If Apple can put guardrails and thoughtful interface design around third-party models, limiting the data shown, making it clear when the AI generates an answer, and providing easy ways to modify or disable functionality, it can aspire to be a responsible front-end to other companies’ experimental backends.
Google is in a good position in the AI race, as Google’s Gemini is now included as the default AI assistant on most Android smartphones. Critics will argue that handing over the core of AI to Google risks undermining Apple’s long-term independence. That’s really concerning. If the future iPhone experience is defined as much by Gemini’s behavior as iOS itself, Apple will rely on its rivals more than it has since the good old days of relying on Microsoft Office to keep the Mac alive.
The challenge will be to use Gemini as a bridge rather than a permanent crutch to avoid repeating the same mistakes while Apple works to develop foldable iPhones, smart glasses, and new Macs that can run efficiently on its own chips.
It seems practical to outsource the heaviest AI lifting to Google. This buys Apple time to see how the generative AI market changes: which models prove reliable, which business models are sustainable, and which applications people actually use beyond novelty chatbots.