It seems clear that generative AI will touch almost everything we do with computers in some way. This observation has brought many to ask what the AI strategy is for one of the biggest and most important players in computing hardware and software: Apple.

Apple’s executive commentary around AI has been cautious. Apple’s management team understands that this generative AI trend is new, and that the technology still has trust and safety issues. Apple is unlikely to be an early integrator of the technology until it believes it to be trustworthy enough to bring to its billion-plus customers.

Apple’s largest computing-platform competitors—Microsoft and Google—have been showing their cards around their bets in generative AI. We can assume that, beyond their current applications, they will likely add AI to their Windows and Android operating systems. How Apple might compete is an open question.

But Apple has some core differentiators when it comes to anything it does in AI, including generative AI.

SILICON

Apple develops some of the best semiconductors on the planet for every category of product it makes. The Mac is the most capable computer on the market to run generative-AI applications. Apple’s M2 processor is capable of an approximate industry-leading 15 trillion operations per second (TOPS). And it has a powerful neural processing unit, which might enable AI applications to run on-device instead of in the cloud.

PRIVACY

Unlike earlier new technologies, users require no new hardware to access generative AI, just some form of modern computing device (PC, smartphone, tablet). That means more than four billion humans on the planet will have instant access to this new technology. For anyone creating applications, services, or any end-customer experience, that’s an exciting reality.

Because so much data would be flowing into and out of large AI models, it’s also cause for concern about user-data privacy. If we want those 4 billion people on the planet to adopt AI rapidly, privacy will have to be front and center. So Apple’s clear commitment to privacy could be a key differentiator. Developing AI responsibly is fundamental to moving the technology forward—and respecting user privacy is a big part of that.

DEVELOPERS

While today’s developers are mostly creating AI applications that run in the cloud, they will inevitably focus on building apps that run AI models locally on client devices such as smartphones and laptops.

No company has benefited more from a vibrant developer ecosystem over the past decade than Apple. And while Apple has no direct play in generative AI, it will support third-party generative-AI apps on iOS and macOS. I use, or have tried, every generative AI tool I can find, and they all work perfectly on my Mac or iPhone. And Apple can continue to offer its developers tools via its API and SDK to realize their AI ambitions on Apple platforms.

USER DATA

Apple might differentiate itself by training its AI models with data that nobody else has. While it’s unlikely Apple will attempt to make a generic large language model AI system like ChatGPT or Google Bard, the company may instead focus on an AI language model built with its intimate knowledge of its customers. Apple has always wanted Siri to be the most helpful AI assistant to an Apple customer over all other AI assistants. Insert generative AI and the compute power of Apple Silicon to train Siri on the unique patterns of behavior and individual nuances of its owner, and add Apple’s ability to keep all that information private and on-device, and I think we get a glimpse of what Siri could become. Let’s call it a personal language model that can adapt and evolve as it learns more about each specific Apple customer.

Many companies have considered mixing third-party large language models with smaller models trained on their own proprietary data. Apple could use this approach, so that its customer could still use the general-purpose large language models but also get the benefit of personal language models that are unique for each user. If this happens, you can imagine that no two Apple customers would have the same generative AI experience. Each personal model would be uniquely customized to the user’s workflows, language preferences and style, personality, and more.

All of this is interesting to ponder, and this is my best hunch at how Apple could approach generative AI in a way that offers something unique compared to its competition.

I doubt that Apple will make any announcements in this space any time soon. As of now, there are still trust and safety issues around all these generative-AI models. I strongly doubt that Apple will expose its customers to any technology that comes with trust and safety risks. So until those issues are resolved, I don’t see Apple integrating any generative AI technology into its software.