Apple is intensely focused on its global AI efforts

Not so long ago, I can remember how Apple’s “failures” in AI made critics smile. Those smiles now seem to have faded. Instead, Apple is accelerating at speed to make people happy for a while with American AI.

How do we know the company is moving fast? With more than 160,000 direct employees globally and hundreds of thousands more across partner firms, suppliers, and the currently beleaguered App economy, when the ship that is Apple moves in a direction the rumor mill usually indicates the destination. Along those lines, we’ve heard a lot of talk across the last week.

Apple’s top secret AI labs

Apple has created a top secret AI research lab in Zurich, Switzerland. The Financial Times also claims the company has hired hundreds of leading AI researchers during the last couple of years, many of them from Google. 

These teams are focused on developing highly advanced AI models. What kind of models? In essence, these seem to be super-lightweight, highly focused neural networks capable of delivering really useful tools that function on the device.

To get a sense of what these might do, Apple researchers recently released a wave of eight new AI models capable of running on the device. The company calls them “Open-source Efficient Language Models”. 

Model behavior

These are small models trained on public data that work on the device to solve focused tasks. The aim is to make it possible to run generative AI (genAI) tools on the device itself, rather than using servers, which preserves privacy, improves efficiency, and safeguards information. These solutions promise truly mobile AI devices that will work offline, with the code is being made available to researchers on GitHub. 

These aren’t the first AI models to slip out of Apple’s research labs. Earlier this year, the company published AI models that can edit photos through written prompts, and another to help people optimize their use of an iPhone. Interestingly, six of the researchers named as authors of a paper describing the latter technology were former Google employees hired in the last two years. 

Making friends

Apple also seems to be exploring potential partnerships. In recent months, we’ve heard it has spoken with both Google and Baidu to make their AI models available to iPhones; last week, we heard it has recommenced discussions with OpenAI. 

This has led both to speculation of an AI-dedicated App Store from which users can access bespoke selections of third-party AI solutions and rumors Apple seeks to license third-party models to enable its devices.

Apple also seems focused on augmenting its existing apps with AI. AppleInsider claims the company is testing a version of Safari with a built in AI-powered intelligent search agent capable of providing summaries of websites.

Think ethically

Throughout all of this, Apple has maintained a tight silence about the totality of its AI strategy. Critically, however, it’s important to understand that the company is not interested in building solutions that provide incorrect or inappropriate responses and would rather be cautious than to introduce an AI that is flawed. It seeks to develop ethical, useful AI that provides real benefits to users while retaining privacy. 

This also extends to how it trains its AI models; if you look at its published research papers, you’ll find many of those it has revealed have been trained using publicly available data, rather than breaching copyright.

Apple is also investing in AI infrastructure

Apple will announce its financial results on Thursday. These aren’t expected to impress, but it seems likely much of the disappointment is already baked in. But for those of us curious about the extent to which Apple is preparing the ground for AI, it will be interesting to track how much the company is investing in capital expenditure. 

We know such spending is taking place:

• Just over a week ago, the company announced an expansion to its Singapore campus to provide space for “new roles in AI and other key functions”, and is making similar investments in Indonesia.
• Apple also recently acquired French AI firm Datakalab. That company specializes in on-device processing, algorithm compression, and embedded AI.
• Hints that Apple will have some reliance on AI in the cloud are also visible on news the company has appointed former Google Sumit Gupta as director of products, Apple Cloud. Gupta has years of experience in AI infrastructure, including a previous six-year stint as chief AI Strategy officer and CTO of AI at IBM.

All of these suggest that sizable investments in the infrastructure required to power AI on two billion actively used Apple devices is already taking place.

Securing the servers with Apple Silicon

Investment extends to R&D for infrastructure. After all, it means something that Apple is allegedly considering building servers powered by Apple Silicon chips. Those servers could go some way toward providing the kind of computational power required to drive AI services in the cloud, while also mitigating the enormous energy consumption such services require.

These data points should provide some color as we accelerate toward introduction of new M4(?)-powered, AI-capable iPads at an online Apple keynote next week, followed by a little more insight at WWDC 2024 in June — and culminating with the big AI iPhone 16 reveal in fall.

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.