I used to be about to okay my good friend’s restaurant suggestion for lunch — an Indian place on 2nd — when Apple Intelligence swooped in with one other thought. “How about The Ritz?” appeared above the keyboard as a prompt response, highlighted in that telltale AI rainbow glow. The opposite prompt response, “Sounds good!” was rather more cheap. However ignoring each, I typed out my affirmative reply, hopped on my bike, and headed to downtown Seattle, the place there are, to my data, zero Ritzes.
Instructed replies aren’t new in iOS 18.2, however they’re a bit of the Apple Intelligence characteristic set that’s falling into place with this week’s public launch of 18.2. These options I received whereas planning lunch form of sum up my complete expertise with Apple’s AI up ’til now: sometimes useful, typically approach off base, and sometimes good for fun. However as soon as the novelty wears off, it’s simply ignored — similar to the AI characteristic units on each different so-called AI smartphone I’ve used this yr.
Apple needed to get one thing out the door for its “constructed for Apple Intelligence” iPhones
Apple took its time getting right here. The primary set of AI options dropped with iOS 18.1 on the finish of October, together with notification and e mail summaries, generative writing instruments, and a cleanup instrument to take distractions out of images. It felt like a deeply minimal viable product, however Apple needed to get one thing out the door for its “constructed for Apple Intelligence” iPhones.
Now, iOS 18.2 has formally arrived after months of beta testing with a meatier set of updates: the Picture Playground app for AI picture era, Genmoji, and a ChatGPT extension for Siri. You additionally get Visible Intelligence, however solely with an iPhone 16 or 16 Professional, for causes which are unclear. There’s extra to come back, in fact, however Apple has lastly shipped a set of AI options that resembles Samsung’s and Google’s. The issue is that each one of these telephone makers are nonetheless a great distance from delivering the AI smartphones we’ve been promised.
Siri’s massive replace in 18.2 is the addition of ChatGPT. It’ll nonetheless set timers and reply your primary questions the way in which it at all times has, however now it could ship extra complicated queries to ChatGPT. It’s decide in and doesn’t require an OpenAI account to make use of, which is good. It’s nonetheless simply as inclined to creating stuff up as ever, however it could act as a useful start line in order for you some help with a posh subject.
Over on Android, Google’s AI-powered Gemini has turn into the default voice assistant, and whereas it lacked loads of primary capabilities at launch, it’s been creeping towards characteristic parity ever since. Now, it could set timers, play your Spotify playlists, and brainstorm dinner concepts with you. It’s all effectively and good in the event you’re not sure what to do with a bunch of wilting produce, however the true take a look at will come when these voice assistants achieve the flexibility to take motion on our telephones. That’s one thing that each Apple and Google are working towards, however to date, the brand new AI-powered digital assistants are simply chattier variations of their former selves.
Of all of the updates iOS 18.2 affords, Picture Playground might be the flashiest. It’s a standalone app with a waitlist, however when you’re in, it unlocks picture creation instruments somewhere else all through the OS, too. Picture Playground is rather a lot like Google’s Pixel Studio, however with approach stricter guardrails — that’s factor, principally. Requests to create a picture of Pikachu sticking a paper clip in {an electrical} outlet had been denied, which is nice information for Pikachu.
It’s cute, and you’ll generate pictures of your partner as a chef or an astronaut or no matter. They’re form of humorous in the event you’re into that. But it surely’s not proof against the pitfalls that numerous AI picture instruments endure, and outcomes will usually look not fairly proper or simply plain bizarre. Like most AI instruments, Picture Playground is somewhat guess-making machine and sometimes makes moderately good pictures. But it surely additionally guesses fallacious, like when its model of avocado toast features a pit or steam rising from a tomato within the background of a picture of a scorching bowl of soup. And don’t ask it to make palms since you received’t like what you see.
Genmoji is on even stricter rails, and in my expertise, it will get issues proper rather a lot. However there are some fairly apparent limitations, together with the truth that they’re so tiny it’s onerous to see a lot element in them. They’re purported to be tiny, however you’ll be able to overlook that and get carried away including a bunch of stuff after which discover all of it barely seen within the remaining product. It made an honest depiction of me in entrance of a Christmas tree ingesting from a pink espresso cup that appears good in preview however is unattainable to parse at typical emoji dimension. In addition they don’t work effectively in group texts with RCS, so I can’t reply to household texts with obnoxious emoji, which is my main use case for this characteristic.
New in 18.2 is the flexibility to direct AI to make an editorial sound a sure approach. You’re not certain by simply “skilled” or “pleasant” descriptions; you may make it sound like Mr. T wrote your e mail, and it’ll inject a bunch of “I pity the idiot,” which I’ll give it somewhat credit score for, as a result of that’s form of hilarious.
However just like the picture era options, this seems like AI desk stakes at this level. All of us received chuckle getting ChatGPT to write down sea shanties about farting when it first debuted; in the event you didn’t, then I’m very sorry to report that ChatGPT is extraordinarily good at this. Having this functionality baked into my telephone doesn’t make it any extra helpful — if something, it’s extra annoying making an attempt to make use of it with iOS’s textual content choice and formatting instruments. I pity the idiot who purchased an iPhone 16 on the promise of Apple Intelligence.
That’s my greatest downside with AI on telephones proper now. Typically, it does what it’s purported to do. But it surely’s not often useful and doesn’t really feel prefer it’s fixing any actual downside I used to be having. That’s been my criticism about this yr’s units from Google and Samsung; now, Apple is at the very least within the dialog. However they’re all in the identical place, with equal stress to ship one thing in 2025 that isn’t only a assortment of humorous methods — the novelty is carrying off quick.
