Siri’s AI Era Arriving Soon

There has been much debate about what Apple’s upcoming AI-based features could look like. With WWDC’s keynote just eight days away, Mark Gurman has some information about how Siri could be transformed by new technology:

The big news at Apple’s Worldwide Developers Conference next week will be the company’s renewed push into artificial intelligence. As I’ve outlined previously, Apple’s approach will integrate AI into many of its core apps and features, including Photos, Messages, Notes and the Safari browser. It will handle voice memo transcriptions and provide recaps of meetings, text messages, emails and articles. There also will be new bells and whistles like AI-created emojis.

But Apple’s original AI product — Siri — will get some love as well. The company is overhauling the digital assistant with its own large language models, an underlying technology behind generative AI. The new system will allow Siri to control individual features within applications for the first time. This won’t require any setup by the user or from developers, differentiating from existing features like Siri Shortcuts and App Intents. Instead, the iPhone’s AI will analyze what a person is doing and auto-enable Siri to help.

For instance, users could ask Siri to delete or forward an email. Or they could have Siri edit a photo, summarize a meeting or move a note to a different folder — all within the apps themselves. Today, Siri mostly lives outside of the app universe, controlling more general items like smart home appliances, music and system settings. Over time, this new feature will expand to allow multiple commands at once. For example, you could tell your iPad to write an email and send it to your spouse.

The day after OpenAI demoed ChatGPT-4o, Microsoft announced a new class of Windows laptops, powered by Qualcomm Arm chips and boasting a slew of AI-powered features. Some of these features run on device, while some rely on the cloud, but at the heart of almost of them is Microsoft’s multibillion dollar investment in OpenAI.

Google, on the other hand, is going it alone, building LLM-powered features internally. As is its usual style, Google has already announced and cancelled several different products under the AI umbrella, but in recent weeks, it has launched the single most important AI-powered service the world has yet seen in the form of Google Search with Generative AI.

Unless you’ve been living under a rock — or ate one — you know how that started. Google has taken the most important property on the web and turned it into something that will confidently lie to its users. It’s been astonishing to watch it burn so much brand capital on a new feature.

That brings us back to Apple. There’s no doubt that Apple is next in line to infuse its products with AI-powered features, but much of the baggage that comes along with AI seems antithetical with Apple’s brand.

Let’s talk about hallucinations for a moment. I think IBM’s definition of the term is pretty good:

AI hallucination is a phenomenon wherein a large language model (LLM) — often a generative AI chatbot or computer vision tool — perceives patterns or objects that are nonexistent or imperceptible to human observers, creating outputs that are nonsensical or altogether inaccurate.

Preventing a LLM from giving a user incorrect data has become a devastatingly difficult problem to solve. On a recent episode of Decoder with Nilay Patel, Google CEO Sundar Pichai said this on the topic:

Are we making progress? Yes, we are. We have definitely made progress when we look at metrics on factuality year on year. We are all making it better, but it’s not solved. Are there interesting ideas and approaches that they’re working on? Yes, but time will tell. I would view it as LLMs are an aspect of AI. We are working on AI in a much broader way, but it’s an area where we are all definitely working to drive more progress.

“Year on year” sounds like a long time scale to me, and one that won’t be met by the time iOS 18 and its siblings launch this fall. Maybe Apple has made a breakthrough on this front, but that’s yet to be seen. If Apple is going to rely on OpenAI for some things, don’t expect Siri to be much more reliable than ChatGPT.

It’s not surprising that the Siri brand is going to be home to many of these features. It’s been around since the iPhone 4S. Here’s a bit from that 2011 press release:

iPhone 4S also introduces Siri, an intelligent assistant that helps you get things done just by asking. Siri understands context allowing you to speak naturally when you ask it questions, for example, if you ask “Will I need an umbrella this weekend?” it understands you are looking for a weather forecast. Siri is also smart about using the personal information you allow it to access, for example, if you tell Siri “Remind me to call Mom when I get home” it can find “Mom” in your address book, or ask Siri “What’s the traffic like around here?” and it can figure out where “here” is based on your current location. Siri helps you make calls, send text messages or email, schedule meetings and reminders, make notes, search the Internet, find local businesses, get directions and more. You can also get answers, find facts and even perform complex calculations just by asking.

A lot of that sounds like the things people expect from the likes of ChatGPT or Gemini today, doesn’t it? Maybe’s Siri’s habit of whiffing on basic requests will buy Apple some leeway when it comes to hallucinations.

Then there’s the environmental impact of all the hardware required to power these features. Undoubtedly, Apple is working hard to run as many of these models as possible locally on users’ devices, but even Apple is rumored to be looking to the cloud to power some things. Then there’s the company’s stance on user privacy.

There a lot of circles to square, but Apple probably doesn’t have a choice here. Users — and the market — expect these features from platform owners now, regardless of their actual utility.

Don’t hear what I’m not saying: I think generative AI has some amazing uses and is here to stay. This movement is more than a flash in the pan, and Apple should have a plan for incorporating it across its operating systems. However, I think Apple has to move more carefully here than the likes of Microsoft and Google. The way Apple sees — and talks about — itself may hang in the balance.¹