Pretty much all the big tech companies jumped onto the genAI bandwagon when ChatGPT made its debut. This gold rush mentality was accompanied by a sizable blooper reel of bugs and missteps on the part of the mega-cap players, showing, perhaps, the wisdom in a slow and steady wins the race attitude. Conspicuously absent from the genAI rush is Apple—the biggest of the big tech companies. Does this absence signal disinterest, or is Apple deliberately pacing itself, quietly monitoring how the market matures, and and carefully plotting its moves? I’d argue the latter. So what is Apple planning for AI?
We won’t have to wait long to find out for sure. Apple intends to unveil its AI strategy at its Developers Conference on June 10th. However the outlines of Apple’s plans can already be read in the moves it has already made over the years, suggesting that Apple has had serious AI ambitions for some time. Everything points to Apple leveraging its privacy policy, hardware chops, and neigh infinite resources to accomplish feats with AI that nobody else can.
Apple’s Five Pillars of AI Strategy
As I see it, in addition to the expected Siri upgrades and obligatory search functionality improvements, there are five main pillars to Apple’s AI strategy: privacy, miniaturization, localism, and hardware. Each pillar is interdependent and mutually supporting and add up together to a complete package.
1. Privacy: Apple is famously unique among the tech titans in that it doesn’t wantonly pillage your data. The higher ups at Apple are smart enough to know that being the odd one out here is a strong value-add and distinguishing mark. Since the overwhelming lion share of Apple’s revenue comes from hardware sales they don’t have to hustle people’s data to turn a profit anyway. This emphasis on user privacy puts Apple in a strange position when it comes to AI model training, however. Unlike a Meta or Goggle, who have access to vast pools of their user’s data to train their models, Apple is subject to self-imposed restrictions on how it can use its customers data. At first glance this decision might seem to have painted Apple into a corner, but things make more sense when you factor in the other pillars. The upshot of privacy focused software is that Apple may be able to do stuff with user’s private data locally that other providers cannot. I would argue privacy is especially important for AI, considering that people may have anthropomorphizing psychological hangups about a bot “stranger” learning certain details about them.
2. Miniaturization: Apple recently released 8 “small language models” to the public in a project called OpenELM, short for Open-source Efficient Language Models. These somewhat euphemistically named “efficient” language models are of interest in and of themselves, as they provide important experimental fodder for investigating just how large bulky, expensive and data-hungry language models absolutely must be to perform certain tasks effectively. They also help us look into how far algorithmic optimization can get us over plain old brute force size increases to models. Some of the models Apple released are really micro, with the smallest being trained on a mere 270 million parameters. Notably, these models are built to run locally, on an iPhone or MacBook. Which gives us the next clue to Apple’s plan. Apple’s interest in ELMs makes sense considering it has to be a bit more modest about where it can draw training data from.
3. Hardware: Apple has long invested in hardware optimized for AI. These chips started showing up in iPhones for the better part of a decade now. Considering Apple’s total end-to-end dominion over its technology stack, it seems better poised than most to design hardware-specific AI functionality. With supposedly AI-focused M4 chips for Macs coming out soon, which are rumored to come with supercharged AI capabilities, Apple’s strengths as a hardware company may give it a crucial edge in the AI race over the long term. Apple’s latest gadget offering, the Vision Pro is, if nothing else, a remarkable demonstration of Apple’s unquestioned hardware design chops. What it brings to the table in terms of both AI centered processing, and robotics and devices will surely set it apart from the pack.
4. Localism: Unifying privacy, ELMs, miniaturization and neural hardware is locally running AI. Perhaps much can be done within the niche of private, locally running AI applets that cannot be done outside of it. Personally, I’m especially excited about the M4 chips for MacBooks, which could be a great boon to developers working on their own models. Apple is also said to be working on AI powered robots and wearables, feats of engineering which would look far less realistic if such machines had to wirelessly ping a remote server to do all their AI related number crunching every time they acted.
Oh, and there’s one last pillar to Apple’s AI strategy that I’ll put down as an honorary mention. Loads of money. Unlike many companies, Apple doesn’t have to answer anyone, including investors (who can be notoriously narrow-minded). This means they can experiment without having to court disaster. Amidst news that production is being cut for the Vision Pro due to low demand, Apple is faced with a mere trifle whereas virtually any reasonable company would be seriously damaged by such a turn of events.
Putting it all together
Having taken Apple’s various strategic pillars into account, a coherent picture of its AI plan emerges. Apple seems to be less interested in putting out just another chatbot. Apple wants to leverage its status as a hardware design company to market privacy-respecting, locally running, efficient AI applications with a physical presence. A major avenue for Apple to differentiate itself from its competitors is its ability to embed neural processors in wearables, robots, phones and computers. Apple could also capture the attention of the AI developer community if it builds computers that are designed to develop AI.
Apple’s slow-to-enter approach to the AI race does not signal to me that it has been asleep at the wheel. Rather, it’s a sign of confidence. They are proceeding deliberately and at their own pace and are waiting for the most opportune time to make a splash. By the looks of it, that time will be June 10th. Stay tuned!