Topic of the Month
Apple Takes a Bite of AI
In 2021, at the height of work from home and Zoom meetings, Facebook rebranded itself Meta. Virtual worlds, or the metaverse, were the future of work and play, the company said. Soon after, other tech companies jumped aboard with their own metaverse plans. But just two years on, the metaverse was passé, and AI was in. Now, no earnings call was complete without mentioning it.
In both cases, you’d find one obvious player absent early on: Apple.
The company, so the story goes, is often late to big tech trends but redefines categories when it finally arrives. Last year, Apple weighed in on the metaverse with the Vision Pro, and this month, after a notable absence from the AI maelstrom, the company gave generative artificial intelligence the Apple treatment too: The new features across its devices will be called Apple Intelligence. (Presenters at its flagship software event, WWDC, failed to mention AI or artificial intelligence once. But make no mistake, it’s the same stuff.)
According to Apple, various AI models will soon stitch together personal context from across the apps on your device to get useful things done. In an example given at WWDC, the company said this might take the form of reconciling an incoming meeting request with your daughter’s play. Will you be late? The AI will check your calendar, texts, email, and traffic to let you know. It can also fetch photos of specific people from a quick description, generate images or custom emojis, and proofread and edit text. Some of this will be accomplished with Siri, thanks to a much-needed update.
To preserve privacy, simpler tasks will be handled by small language models on your device. For more complex tasks, the device will interact with larger models living on what Apple claims is uniquely secure AI cloud infrastructure.
But for the most complex tasks, the device will, with your permission, kick requests over to OpenAI’s GPT-4o. The new partnership was a big headline-maker. It involves no cash for now: OpenAI will get exposure to Apple customers in exchange for granting access to its top AI model. Meanwhile, Apple has made it clear it’s open to other partnerships in the future. This could mean a similar arrangement with Google or Anthropic. (How secure and private these third-party arrangements will be remains a point of contention.)
So, how’d Apple do? The reviews were decidedly mixed: Some people thought the presentation knocked it out of the park. Others said the new tools were boring, a security and privacy nightmare, ethically flawed, or proof Apple isn’t capable of fielding its own advanced AI.
Much of what was announced is still off in the future, so we should be wary of judging a splashy tech demo before it’s in the hands of everyday people. Still, the strategy itself is worth reviewing: It’s a first look at where one of the most valuable companies in the world, with 2.2 billion devices in the wild, is going with its own AI offerings.
As a Fast Company headline summed it up: “OpenAI Promised to Give Us Her. Apple Is Giving Us Gary From 'Veep'.” The former is a vision of AI from science fiction, an endpoint we can imitate but not yet match. The latter is less striking but focuses on making today’s capabilities work for average people, regardless of their mileage with AI. Writing for Wired, Will Knight said Apple is selling near-term AI as a “feature not a product.”
“Rather than a stand-alone device or experience, Apple has focused on how generative AI can improve apps and OS features in small yet meaningful ways,” he wrote. The more the tech disappears into the background, the easier it will be for more people to use and trust it.
And although Apple really does look to be lagging AI’s cutting edge, it’s come up with some clever workarounds to make the most of things as they stand.
By offloading as much work as possible to local models—which may be a decent amount—they’re saving on sky-high computing costs in the cloud. For the most complex and therefore pricey tasks, they have a free option. OpenAI will foot the bill in the hopes they’ll pick up paying subscribers. In addition, the riskiest types of requests—think Google’s AI Overviews and other chatbot fails of recent history—are going to a third party.
Lastly, these AI features will only be available on Apple’s newest phones due to legitimate performance requirements. (Even small language models require some nifty hardware to run.) If the features work smoothly and prove popular, they could push people to upgrade their phones to gain access, driving a cycle of device sales.
To be fair, Apple isn’t the first to sketch out an AI-enhanced operating system instead of a chatbot-only approach. Microsoft wants to head in a somewhat similar direction with its Copilot+ PCs, announced last month. (Though it’s already in hot water there.) And Google’s AI Overviews experiment in search—which has notoriously been giving some rather poor advice—shows the models’ weaknesses are magnified when tested at the scale of billions. Finally, Apple’s last-but-not-least strategy doesn’t always pan out right away. After lower-than-expected sales, the company is reportedly moving on from the Vision Pro in its current form in favor of a redesigned device at a lower price point.
Still, the announcement rounds out the AI strategies of the world's biggest tech companies. Generative AI will be coming to pretty much everyone very soon.
Know someone who might enjoy the Singularity Monthly?
Share this newsletter with them.
