This is the free version of the Spyglass column, if this was forwarded to you, you can sign up here. If you'd like to read the rest of these posts and get them sent to your inbox once a day (on weekdays), you can upgrade to the paid plan. Thank you for reading.
Back in January, I wrote a post with the purposefully provocative headline: "Should Apple Switch Off Siri?" It was on the heels of the reports that not only was Apple's AI continuing to embarrass herself in real-world usage, by not knowing something as simple (for such a system) as previous winners of the Super Bowl, but also the notion that the real, core upgrade to Siri – not the mild updates due soon – were likely not coming until 2026. Basically forever in our era of AI. Well, 2026 is now looking more like 2027, reports Mark Gurman for Bloomberg this week: For iOS 19, Apple’s plan is to merge both systems together and roll out a new Siri architecture. I expect this to be introduced as early as Apple’s Worldwide Developers Conference in June of this year — with a launch by spring 2026 as part of iOS 19.4. The new system, dubbed “LLM Siri” internally, was supposed to also introduce a more conversational approach in the same release. But that is now running behind as well and won’t be unveiled in June.
Before Apple can go full-throttle on development of that Siri, which is supposed to finally work more like ChatGPT and the new Alexa, Apple will need to get the underlying system fixed. And that won’t be easy. That’s why people within Apple’s AI division now believe that a true modernized, conversational version of Siri won’t reach consumers until iOS 20 at best in 2027.
If this is true, it's not just a joke, it's downright worrying. (And if it's not true, Apple should probably do something to refute it because it's quite damaging on a number of fronts.) And it couldn't come at a worse time, with Amazon having just finally unveiled 'Alexa+' at an event, their own LLM-based upgrade to their OG digital assistant. Apple and Amazon seemed to be in similar boats with regard to delays hampering their assistant upgrades – in no small part because they had some levels of success with the legacy products. But if Amazon was sort of a tortoise in this race to the hares of Google and OpenAI – as always, there's hope that despite being slow to start, the tortoise can steadily win in the end! – Apple is more like a dead duck now in the same race. A corpse. 2027? How about never – is never good for you? Now, I'm skeptical that Alexa+ will be as revolutionary as Amazon is projecting because I'm skeptical in terms of just how well their agentic elements will work. It's early days and this stuff is hard, even with all the great advancements in AI, and Amazon is saddled a bit by Alexa being a voice-first system (which is undoubtedly why they're rolling Alexa+ out first to devices with screens). But regardless, a more conversational Alexa sounds like it will work well. And actually, voice is the perfect interface for that type of interaction, obviously. OpenAI and Google got there in 2024. So did Meta. Even Microsoft! Amazon is here now in 2025. And Apple... 2027?! It's just beyond embarrassing. Again, it's worrying! And so I no longer find my idea provocative. In fact, I think it's what needs to happen. That is: Given the aforementioned OpenAI deal, why not fully outsource Siri to ChatGPT? You'd still want to keep the task-oriented elements – setting timers, etc – with her, but everything that requires anything resembling a search query should be outsourced. Right now, you can force this by asking Siri to ask ChatGPT something, but it's cumbersome, I'm suggesting this be made the default action.
Apple has a long and illustrious history of teaming up with partners to get a service out the door while they work on their own solution behind the scenes. This situation would be slightly different in that their own solution actually pre-dates that of the partner's, but Apple could rather easily navigate this awkward oddity. The ChatGPT layer of Siri has been live for a few months now, but it's buried fairly deep in the background of most Siri queries. All it would take is some algorithm tweaks to bring ChatGPT to the forefront. And really, it should be as simple as: if the query is about setting a timer or turning off the lights – a system task – go with legacy Siri. If it's a "world knowledge" question or the like, go with ChatGPT. The biggest issue here would be the trepidation Apple clearly has had around sending these queries to OpenAI. That's why we get explicit prompts to ensure a user wants to ask ChatGPT the question (unless you explicitly ask Siri to ask ChatGPT – which is too cumbersome of a way to query). In my rip-and-replace scenario, Apple would need to overhaul such prompts. Perhaps there's a one-time disclaimer when you set up Siri about using ChatGPT. Apple won't love it. But it's what needs to be done. It's 2025, Siri sucks, with no sign of the sucking stopping, and we can't wait until 2027 for what is now already table-stakes technology. They could also, of course, outsource such abilities to Google. And perhaps Apple would prefer that since Google is a trillion-dollar company at scale and not, technically, a startup like OpenAI. And maybe it could help them augment or replace the Google search deal which will clearly be going away – or at least be heavily altered – at some point in the coming years. Still, Apple's reason for going with ChatGPT in the first place is the same reason they go with Google in search over Bing: ChatGPT is the market leader in the space. And again, they already have an agreement in place and ChatGPT already is baked into iOS. While there are signs that other AI systems could be coming (and certainly they need to in China), there's no word or sign that Gemini is any closer to being one of these partners. OpenAI is also in a different spot than it was last year when Apple signed the pact. Yes, there's been more turnover and turmoil – which probably played a big role in Apple not investing in OpenAI – but there's also been one big change which may actually help Apple outsource Siri to ChatGPT: Microsoft is no longer the sole compute partner. It was also an awkward situation that Apple was partnering with OpenAI to send millions of Apple device users to ChatGPT, when Microsoft was the main one footing the bill for such usage. An investment in OpenAI would have given Apple a way to pay for such usage, even if indirectly, but again, that fell apart, so it's a weird situation. But with SoftBank now stepping in as OpenAI's new main capital partner going forward, and with Oracle and others stepping in to be compute partners for 'Project Stargate', there's a new, less awkward path here. And Stargate itself, assuming they can actually scale it, matters here as well since if Apple truly did replace Siri's brain with ChatGPT, it would mean billions of queries hitting their servers. It's hard to fathom how big of a moment this would be for OpenAI. The first Apple deal was a huge endorsement, but again, usage is probably muted because the functionality is pretty buried. If ChatGPT becomes the key component of most Siri queries... Let's just say their lead in the consumer AI race will be further extended. The other thing Apple may have to reconsider if they were to do this new type of deal is how they upsell premium ChatGPT tiers. Right now, you can upgrade through iOS, which is great for OpenAI. But with such deep integration, Apple would probably want most, if not all, interactions to be free for their userbase. And this might mean actually paying OpenAI for the usage. And that probably means billions of dollars over some period of time. Again, that would be a watershed moment for OpenAI, but this time as a business. And/or perhaps such a deal would make Apple revisit the idea of an investment in OpenAI. Luckily for them, a new $40B round seems to be coming together... Anyway, I just feel like while this solution may sound extreme on paper, it's not so extreme when you think about it. And especially when you consider Apple's long history of partnering on technology before their own is ready to go. This is what needs to happen here! Would OpenAI be okay signing up for perhaps just a couple years of deep integration before being replaced by a new Siri? I mean, given the tradeoffs – millions, if not billions, of users exposed to ChatGPT and perhaps billions in revenue coming in the door – yes, they obviously would take that deal. And actually, nearly a year ago I wondered if this wouldn't be how any partnership between the two sides wouldn't be executed, after reports surfaced that Apple had tried hooking up Siri to ChatGPT in an internal test: The hooking up ChatGPT to Siri is obviously the big element here. While Apple seems to be working on other upgrades to their assistant, those may be a bit farther out. The question is: in order to improve Siri now, is Apple willing to do a brain transplant, as it were? Replace whatever Siri is now with GPT-4o or some variation? Or could that model be made to simply augment Siri?
They went down the augmentation approach, clearly. And it's honestly barely that. And overall, Apple opted for their more hybrid-approach, with again, ChatGPT pretty buried in the background – after many explicit opt-ins, of course. Another option here may be for Apple to take a page from Amazon and replace Siri with many different models from various third-parties, all routed via their system. Of course, Apple lacks the deep partnership (and investment) that Amazon has with Anthropic. And so perhaps OpenAI and Google and others may not be okay with simply being a faceless API call on Apple's backend. And OpenAI has quite a bit more leverage here now, it would seem, given how slow Apple has been to build and roll out their own AI systems. Again, they are far behind right now and can't seem to be able to muster the will to become more competitive for at least two more years! Sure, bringing Kim Vorrath, a long-time get-shit-done task master at Apple, over to the AI team will help. But all of this points to a deeper cultural and perhaps systemic issue. The reality may be that the cadence at which Apple operates is simply different than the cadence at which a company must operate in the era of AI. Perhaps that pace slows as the underlying technological advances level-off. Or perhaps it doesn't! That's why I still believe Apple may need to make a big acquisition in the space in order to better compete, fast. But for now, this ChatGPT solution is just sitting right there for the taking. And while it may require eating some humble pie, Apple can help swallow it down by reframing it as bringing their customers a deeper partnership with the leader in AI. Can't you just hear them say something like that on stage at WWDC in a few months? "The response to ChatGPT integration inside of Apple Intelligence has been amazing. So amazing that we're extending the partnership to bring ChatGPT integration even deeper into Siri with [pause for dramatic effect] SiriGPT." At first, this could be mainly powered by ChatGPT – maybe Apple would couch it a bit, noting that it was a hybrid of their own models and OpenAI's, again, much like Amazon is doing with Alexa+ – but over time, this could slowly morph. Again, perhaps they start to offer other models behind the scenes and seamlessly route queries (which would also ensure they're future-proofed and not putting all their eggs in one basket). And over time, Apple's own models – which they clearly need at this point, even if distilled from others – take over most of the Apple Intelligence workload. But it all starts with ChatGPT giving Siri that lobotomy. Boom. Ship it.
|
