Twelve years on from its launch, Siri now feels like a modern-day Apple Newton. Both started life as bold new personal assistants, only to stagnate and fall by the wayside. And both have been the butt of jokes on big comedy shows – the Newton was famously skewered on The Simpsons, while Siri has more recently ‘starred’ in the latest season of Curb your Enthusiasm (warning: the scene in question contains a tirade of expletives).
Yet while the Newton was put out of its misery and canceled in 1998, Apple has kept Siri ticking over in our iPhones. Well, barely – anyone who’s used Siri will have their own tale of frustration about its seemingly diminishing IQ. Apple can’t let this continue. And fortunately, the evidence is mounting that we’ll finally get a Siri reboot (or at least, a makeover) at WWDC 2024.
Apple is strongly hinting that AI and, to a lesser extent, Siri will be at the forefront of its annual developer’s conference. Its SVP of Marketing Greg Joswiak posted on X (formerly Twitter) that the event will be “Absolutely Incredible”, with the capitalization unsubtly spelling AI. Plus, the promo’s typography (below) also has echoes of Siri’s color scheme.
But a wave of credible rumors and Apple’s own machine learning research also support the theory that Siri could soon become, if not great, then at least not an anachronistic embarrassment.
How exactly might Apple reboot Siri in June? It’s a long road back. Every time I’ve spoken to ChatGPT‘s baked-in Voice function I’ve marveled at how natural it feels in comparison. The trouble is, ChatGPT doesn’t have direct access to iOS for controlling my phone (unless you use Shortcuts). And Apple seemingly doesn’t have the generative AI chops – or the willingness to compromise on privacy – to do proper cloud-based AI.
The solution is likely to be a compromise, combining Apple’s latest on-device machine learning with third-party AI models like Google Gemini. That might result in a full Siri reinvention, but if it rescues the voice assistant from its current malaise, that’d be good enough for me…
A private chat
A Siri reboot isn’t certain at WWDC 2024, with the latest rumors a little confusing. This week, Bloomberg’s Mark Gurman claimed that Apple “isn’t planning to debut its own generative AI chatbot” at WWDC 2024.
But that doesn’t mean Apple isn’t planning some Siri upgrades. Apple has never seen Siri as a chatbot like ChatGPT – since June 2021, the voice assistant has processed our questions on-device by default. “This addresses one of the biggest privacy concerns for voice assistants, which is unwanted audio recording” Apple said in a press release at the time.
Apple almost certainly won’t have changed its mind since then. So the improvements to Siri’s on-device performance will have to come from Apple – and there’s evidence that the tech giant has been exploring this.
Mark your calendars for #WWDC24, June 10-14. It’s going to be Absolutely Incredible! pic.twitter.com/YIln5972ZDMarch 26, 2024
Last week, Apple researchers published a March 2024 paper that tested whether or not it’s possible for voice assistants to ditch trigger words like ‘Siri’ and instead use on-device AI to tell whether you’re speaking to your phone or someone else. This followed Apple’s quiet publication of a family of multimodal models (AI models that can simultaneously interpret different types of data) called MM1.
For the voice assistant paper, Apple’s researchers trained a large language model (LLM), based partly on OpenAI‘s GPT-2, to look for voice patterns that signify whether or not we’re asking for help from our phone. That’s pretty futuristic stuff. While the results were promising, it’s likely too soon for this kind of tech to find its way into iOS 18 or our iPhones.
Still, Apple is clearly working hard on voice assistant tech and we’ll likely see some of the fruits of this at WWDC 2024. Only six months ago, Apple was rapidly increasing its spending on conversational AI to “millions of dollars per day”, according to a report from The Information. Given the rapid advances of its rivals, some of this investment will surely go towards improving Siri.
New Siri announcements also haven’t been completely extinct in recent months. In December 2023, Apple announced that its new S9 SiP (system in a package) meant the Apple Watch Series 9 and Watch Ultra 2 could “now process Siri requests on device”. This meant you could now ask Siri about over 20 health data markers, because your data wouldn’t leave the smartwatch.
None of this is the work of a company that’s given up on its voice assistant. The question is what Apple will do about the bigger part of the puzzle – answering queries that on-device Siri won’t be able to answer…
I’m not sure I understand…
If Siri can’t answer a question, you can ask it to search the web for you. This is where Apple might be looking to outsource to an external AI model to help with more challenging queries – with Bloomberg predicting that this could be Google Gemini in the US and Europe, with Apple also in talks with Baidu in China according to the Wall Street Journal.
Bloomberg’s report says this generative AI “heavy lifting” could include functions like “creating images and writing essays based on simple prompts”. But these AI features will be delivered via the cloud, with Apple still using its own AI models to process on-device functions and Siri actions.
This mooted Google Gemini deal wouldn’t be unusual for Apple. The tech giant doesn’t have its own search engine, so it’s long had an agreement (reportedly worth $18 billion a year) with Google for it to be the default option on Safari. That deal is now under threat from EU regulators.
Bloomberg says that a similar deal for Google’s Gemini AI models would simply build on this existing partnership between Apple and Google – again, if regulators allow it. It could even be a convenient replacement for that current deal between the tech giants, if it doesn’t crumble from the pressure of regulatory scrutiny.
Whatever happens with these talks, Apple seems unable to solve the whole generative AI puzzle on its own – and unwilling to shift its stance on privacy. That will likely mean some compromises for an upgraded Siri, rather than a giant leap.
Baby steps
The thing is, I’m not demanding (or expecting) a full Siri relaunch at WWDC 2024. All I want is for Apple to acknowledge Siri’s existence, make it less cloth-eared, and lay out how it’s going to make it a more conversational iPhone assistant. Right now, Siri is a brake on the potential of products like the AirPods and Apple Vision Pro, but it can start turning over a new leaf in iOS 18.
For years, Siri has been held back by internal squabbles, a lack of tech breakthroughs, and the wider problem that voice assistants don’t generate much revenue. But if Apple does, as the rumors suggest, outsource some of its generative AI features to the likes of Gemini AI, then it could play catch-up while reducing its liability when the occasional controversy inevitably strikes.
None of this would rocket Siri towards being a voice assistant leader. But it’d be enough to keep Apple in the AI game, while reinvigorating one of the iPhone’s most frustrating features. At this point, Siri is damaging Apple’s reputation, so WWDC 2024 has to be a tipping point one way or the other.
Tim Cook has already promised during an earnings call with Apple’s annual shareholders that the company will “break new ground” this year in generative AI. While that doesn’t necessarily refer to Siri, could Apple do all of that while leaving Siri as it is? I don’t think so. And while Cook said the technology will “unlock transformative opportunities for our users”, I’m just hoping it’ll turn Siri into more than just a cooking timer.