Gemini iPhone a win-win for cautious Apple and aggressive Google

IPhones are likely to get some zeitgeisty new capabilities this year,powered by Google’s Gemini AI models,which could bring Apple’s smartphones up to par with the latest Androids as far as chatbots and generative editing are concerned. But what exactly would that look like? And why hasn’t Apple developed an AI of its own?

Reports this week cite Apple insiders as saying the company has been exploring deals with a number of companies to bring large language models and other generative tech to iPhone,and that it will likely settle on Google as the provider.

The iPhone is full of AI,but not the kind that consumers and investors are asking about.

The iPhone is full of AI,but not the kind that consumers and investors are asking about.AP

The web giant is of course already a major partner of Apple’s,supplying some of the most popular software on the App Store and providing Google Search as the default method for iPhones. That deal — in which Applegets paid $28 billion to use a far better search engine than it could build itself,and Google gets exposure to millions of eyeballs for serving ads — is a win-win,and the AI deal likely will be too.

Google’s Gemini app lets you type or talk to an AI agent about anything. You can ask for help drafting a note,get a summary of complex topics,or even use your camera to show it something and ask about it. On Android you can alsoreplace Google Assistant with Gemini,so in theory you get slightly more human human-like responses when you say “Hey Google”.

Google says Gemini can help with everything from creating learning plans to writing text messages.

Google says Gemini can help with everything from creating learning plans to writing text messages.Supplied

Google’s AI models are seen in other parts of its own Pixel phones too,such as a voice agent that can screen and translate calls for you,or a photo editor that can move or remove objects and fill in the blanks. Samsung phones have somesimilar features.

A potential iPhone adaptation could mean Siri is imbued with new Gemini capabilities,or could expand the chatbot’s powers beyond the iOS Google app. There could be a gesture or Control Centre button that lets you ask Gemini about whatever you’re seeing on your iPhone screen at a particular moment,or a new icon on the keyboard that generates and edits text for you in Notes or Pages. Safari,like Chrome,could get the ability to summarise any web page into a dot points of text.

The benefit to Google of such a deal is clear:it wants as many people as possible using its AI features,because of the tech’s insatiable appetite for user data. Getting access to the queries and behaviours of millions of iPhone users,which rivals like Microsoft will largely not have,would be worth whatever it has to pay Apple for the privilege. For its part,Apple gets a feature the market is demanding,but which it can’t (or doesn’t want to) make for itself.

Apple has along history with AI,but the generative models that power the likes of Gemini and DALL-E are a bit messier than the iPhone-maker tends to like. Developing one in the first place requires collecting entire libraries worth of data to train on (which makes it no surprise that the likes of Google and Meta have been building the tech for years),and they’re infamously difficult to control or predict.

ChatGPT,Microsoft’s Bing and Google’s Gemini are primarily designed to sound cool and confident,but it’s unclear thus far if it will ever be possible to ensure their responses are appropriate and factually accurate. Most recently Googlesuspended Gemini’s ability to generate images,after its attempts to correct the model’s inherent biases meant it was consistently creating women of colour when asked for white,male historical figures.

Apple has purposefully distinguished itself in the market by not collecting or selling user information,and generally shying away from data brokering. (Though it doesn’t mind serving its users up for profit less directly,see the Google Search example.) So soaking up a mountain of data to create a large language model might simply not be its style,but it still considers (and will have heard from investors) that it needs one on its smartphones to keep up with the times.

We’ve seen a similar thing occurring very recently with Mac computers. Apple has always referred to any AI capabilities as “machine learning” or “neural processing”,but in promoting its newestMacBook Air with the M3 chip it has explicitly used “AI”,both in the sense of developing generative applications and in using chatbots as part of a creative workflow.

The reason is clear:customers and enterprises are demanding a so-called “AI PC”,and Apple is aware that AI capabilities are something buyers are now looking for when comparing laptops. Dedicated AI hardware has been in Macs for years,but now Apple has to point it out clearly,and since it doesn’t have a chatbot of its own it’s been highlighting the Co-Pilot function of Microsoft Office in presentations and demos.

Phones are a similar story. There’s a lot of buzz about the AI features in Google and Samsung phones,to the point that someone could conceivably expect an equivalent in iPhone. Apple’s smartphone is already filled with so-called neural processing,but it’s all in the background. And AI from Google,Microsoft,OpenAI and others of course works on the iPhone,but you don’t see it as a dot-point on Apple’s website. By highlighting Gemini,it could tick that box while also keeping a certain amount of distance from the more problematic side of developing and deploying an AI chatbot,at least until the company figures out how to do it in an Apple way.

Get news and reviews on technology,gadgets and gaming in our Technology newsletter every Friday.Sign up here.

Tim Biggs is a writer covering consumer technology,gadgets and video games.

Most Viewed in Technology