Mark Gurman, reporting for Bloomberg:

Apple Inc. executives have held internal discussions about potentially bidding for artificial intelligence startup Perplexity AI, seeking to address the need for more AI talent and technology.

Adrian Perica, the company’s head of mergers and acquisitions, has weighed the idea with services chief Eddy Cue and top AI decision-makers, according to people with knowledge of the matter. The discussions are at an early stage and may not lead to an offer, said the people, who asked not to be identified because the matter is private.

I initially wasn’t going to write about this until I realized my positive take on this news was considered “spicy.” I’m on the record as saying Perplexity is a sleazy company by grifters who don’t understand how the internet works, but I also think Apple is perhaps the only company that can transform that reputation into something positive. After this year’s Worldwide Developers Conference, I had it set in my mind that Apple will never have the caliber of models OpenAI and Google offer via ChatGPT and Gemini. Apple delivers experiences, not the technologies behind them. Gmail today is infinitely better than iCloud’s mail service, and Apple realizes this, so it lets users sign into their Gmail account via the Mail app on their iPhones while also signing them into iCloud Mail simultaneously. Most people don’t know or care about iCloud Mail, but it exists.

Apple’s foundation models are akin to iCloud Mail. They exist and they’re decent, but they’re hardly as popular as ChatGPT or Gemini because they’re nowhere near as powerful. They might be more privacy-preserving, but Meta, the sleaziest company in the world, has billions of users worldwide. Nobody cares about privacy on the internet anymore. I don’t think Apple’s foundation models should be discontinued, especially after this year’s WWDC announcements, but they’ll never even get the chance to compete with Gemini and ChatGPT. They’re just so far behind. Even if Siri was powered by them, I don’t know if it would ever do as good a job as its main competition. (I spitballed this theory in my post-event reactions earlier in June, and I still stand by it, but a version of Siri powered by Apple’s foundation models probably won’t meet Apple’s “quality standard.”)

Perplexity, meanwhile, is about as close as one can get to an AI aggregator that actually has the juice. It’s powered by a bunch of models — Gemini, Grok, ChatGPT, Claude, and Perplexity’s own Sonar — and is search-focused. Here’s how I envision this working: The “more personalized Siri” could rely on App Intents to perform “agentic” work inside apps, the standard Siri could work for device features like playing music or modifying settings, and Perplexity’s technology could be used for search. Most Siri features fall into these three categories: work with apps, work with the system, or search the web. The current Siri is only really good at changing settings, which is why it’s frowned upon by so many people. When most people try to quantify how good a virtual assistant is, they’re mostly measuring how good it is at searching.

The agentic App Intents-powered Siri, if it ever exists, really is revolutionary. It’s akin to Google’s Project Mariner, but I feel like it’ll be more successful because it relies on native frameworks rather than web scraping. It piggybacks off a personal context that any app developer can contribute to with only a few lines of code, and that makes it instantly more interoperable than Project Mariner, which really only has access to a user’s Google data. Granted, that’s a lot of knowledge, but most iPhone users use Apple Notes, Apple Mail, Apple Calendar, and iMessage — four domains Apple controls. They might not use the iCloud backends, but they still use the Apple apps on their phones. If last year’s WWDC demonstration wasn’t embellished, Apple would have been ahead of Google. That’s how remarkable the App Intents-powered Siri is — it truly was a futuristic voice assistant.

But even if Apple ships the App Intents-powered Siri, presumably relying heavily on a user’s personal context, it still wouldn’t be as good as Gemini for search. A Perplexity acquisition would remedy that and bring Apple up to snuff with Google and OpenAI because iOS and macOS would be using their technology under the hood. Apple is great at building user-centric experiences, like App Intents or the personal context, but it struggles with the technology behind the scenes. Even if the Google Search deal falls apart, I don’t think Apple will ever make a search engine, not because it’s uninterested, but because it can’t. Spotlight’s search apparatus is nice — about as good as Apple’s foundation models versus ChatGPT — but it isn’t Google Search. Perplexity would bridge this gap by adding the best models Apple could never make into iOS.

A merger is very different from a partnership, and the ChatGPT integration in iOS today is proof. It’s not very good by virtue of being a partnership. If Siri was ChatGPT, by contrast, there would be no handoff between platforms. It would be like asking ChatGPT’s voice mode a question, except built into the iPhone’s Side Button. Because Apple can’t buy OpenAI, I think it’s best that it tries to work something out with Perplexity, integrating its search apparatus into Siri. Again, in this idealistic world, Siri has three modalities — search, app actions, and system actions — and acquiring Perplexity would address the most significant of those areas. Would I bet Apple will actually go through with buying Perplexity? No chance, not because I don’t find the idea interesting, but because I don’t like losing money. The last major Apple merger was Beats back in 2014, and I don’t think the company will ever try something like that again. I want it to, though.