Apple’s plan to focus on the personal side of AI takes centre stage after years of development

The AI race is picking up speed.

Microsoft earlier this year made the first change to its keyboards since the introduction of the Windows key, when it said it would include a dedicated button to activate its CoPilot AI assistants. Google has made its Gemini generative AI available through its Pixel phones and made a deal with Samsung to integrate its technology into the South Korean group’s AI offerings.

And now Apple has said it will build AI tools into its operating systems for iPhones, iPads and Macs, creating Apple Intelligence and transforming Siri from a voice assistant to a more useful digital assistant that will eventually be able to reach into apps on your phone to give you a helping hand. AI is a core part of the new software going forward – for those who can access it.

One of the last to announce its intentions on generative AI, Apple unveiled its plans at its Worldwide Developer Conference (WWDC) last week, quietening some of the speculation around its ability to innovate. “This is a moment we’ve been working toward for a long time,” Apple’s senior vice-president Craig Federighi said.

Does the ‘Taylor Swift lift’ really add up to €150m for Dublin?

---

Karlin Lillington: GDPR isn’t the problem, Big Tech and governments are the issue

---

Sadhbh O'Neill: Are the tools for measuring greenhouse gas emissions fit for purpose?

---

But the company largely avoided using the term “artificial intelligence”, preferring instead to refer to it as “personal intelligence”, and “AI for the rest of us”, making the point that this is not simply technology for technology’s sake.

Global Data’s Anisha Bhatia described it as a “consumer-friendly approach” to explaining the benefits of AI to its core market, and it appears to have hit the mark. Although shares in Apple fell almost 2 per cent in the immediate aftermath, they quickly recovered ground and Apple regained the title of most valuable company in the world, snatching it back from Microsoft.

Apple may not have been talking much about generative AI until recently but the company has certainly been working on it in the background. While it may not have fully embraced generative AI publicly, there is more to AI than chatbots. Apple first introduced the Neural Engine AI processor into its devices in 2017. And the company has been quietly building its expertise, with a number of acquisitions over the past couple of years.

It has also been training its own large language models, which underpin generative artificial intelligence, using a variety of data – from public web data, licensed data from news archives, stock photography and in-house data created by Apple.

“Apple doesn’t need to be the first; it just needs to do it better,” said PP Foresight analyst Paolo Pescatore. “While it typically takes a more cautious ‘wait and see’ approach, it is interesting to see Apple make such a bold statement with AI still in its infancy.”

The anticipation around this year’s Worldwide Developer Conference, or “Dub Dub” as it is affectionately known, was at fever pitch by the time Tim Cook took to the stage to begin his keynote. Many of Apple’s announcements had largely leaked in advance of the event but there was still an element of the unknown about the event. And while generative emojis, image editing and automatic transcriptions were attention-grabbing, there were other questions to answer.

Among them were how Apple, a company that was vocal about user privacy and data control, could stick to its principles while also using a technology that not only requires large amounts of data to train models, but will be dependent on using highly personal user data to be at its most useful.

Apple Intelligence is hoping to bridge that gap. It uses a combination of its own models for on-device processing, occasionally extending into a locked-down cloud environment with specially built servers when needed.

Federighi told WWDC attendees that private cloud computing solved a necessary, profound challenge around the range of capabilities of AI. While many of the queries dealt with by Apple Intelligence could be dealt with on-device, there are times when more power would be needed, and Apple has created a solution that allows users to throw server-class hardware at the problem, extending the device itself into the cloud without the compromises that other cloud options may bring.

The device figures out what data is needed to get the question answered, and only sends that to the cloud. There is no permanent storage of data, and there are no logs; all of this will be subject to checks and audits. In other words, Apple cannot see the data even if it wanted to, and no storage means the company will almost certainly be unable to comply with requests that may come from law enforcement around the data.

“This is a fascinating tension that sees Apple arriving at the same conclusion as rivals: that it is not possible to fully run today’s AI features on-device and that elements must be outsourced to the cloud,” said CCS Insight’s Ben Wood.

“Apple will try to play up its security credentials, but this marks a shift in approach, nonetheless. The move also reintroduces other inescapable problems with server-powered AI, such as latency and cost, which Apple will have to manage. Despite this initial approach, CCS Insight predicts that Apple will seek to move more processing to the device over time.”

There is one catch, though. Apple Intelligence needs the neural engine in the iPhone 15 Pro line’s A17 chip, or that contained in the M-series chips designed by Apple. That excludes a large part of the current iPhone user base, and while it may spur some users to upgrade, it means there will be a large group of current iPhone users who will be waiting a year or more to access the new features.

Apple’s approach may be surprising to some. Creating large language models is an intensive process that does not always make sense for companies. Amazon, for example, ditched the idea of its own large language models and has opted instead to support other companies, investing in Anthropic.

In a nod to this, Apple has chosen to partner with OpenAI for queries that Apple Intelligence cannot handle. Federighi described it as “the best choice for our users today”, although it comes with restrictions for OpenAI. It cannot save user data and requests, for example, and accounts are not required to use the service. Those who choose to sign in with an OpenAI account, though, will have the ChatGPT creator’s data use policies applied.

“This can be framed as evidence that Apple is behind the curve and has been pushed into a position of dependency on competitors. However, we would highlight that AI has yet to deliver any game-changing applications, and that no rival has yet offered an AI feature or device that poses a genuine threat to Apple,” said Wood. “Apple Intelligence should do enough to calm investors and show that the company is on a journey to offering more attractive AI applications and experiences, with progress set to continue in the coming years.”

The ChatGPT integration may not be as crucial as it first seemed. It will only deal with wider queries that cannot be answered by Apple’s own AI, and it may soon be joined by other third-party arrangements.

Federighi said that although the company was starting with OpenAI, Apple was looking to integrate additional models to give its users options in the future.

Apple’s former director of marketing Dave Edwards described it as “arm’s length”.

“In some ways, this is parallel to the way Apple has approached search, relying on Google as a partner,” he said. “At the same time, this feels incomplete as a solution and it will be interesting to see how this changes over time.”

Time is certainly something that Apple is willing to give the new system. The features look set to be rolled out slowly, even to those devices that can run them. Apple is reported to be releasing some next year, perhaps in an effort to avoid the missteps of some of its rivals. But, regardless of the timescale, they are coming.

“Seeking to take AI to the masses is a pivotal moment for Apple; providing developers with the tools that will create most sought-after features among users,” said PP Foresight’s Pescatore. “And this is just the beginning. The next wave of development will be mind-boggling as everyone seeks to outpace each other.”

• Sign up for push alerts and have the best news, analysis and comment delivered directly to your phone
• Find The Irish Times on WhatsApp and stay up to date
• Our In The News podcast is now published daily – Find the latest episode here