I was about to accept a friend’s restaurant suggestion for lunch (an Indian restaurant on the 2nd) when Apple Intelligence came up with another idea. “How about The Ritz?” appears above the keyboard as a suggested response, highlighted with an AI iridescent glow. Another suggested response is “Sounds great!” It was much more reasonable. But ignoring both, I typed in the affirmative, hopped on my bike, and headed to downtown Seattle. As far as I know, there were zero Ritz’s there.
Suggested replies aren’t a new feature in iOS 18.2, but they are part of the Apple Intelligence feature set and are being introduced in this week’s general release of 18.2. These suggestions, which I got while planning my lunch, sum up my entire experience with Apple’s AI so far. Sometimes it’s helpful, sometimes it’s off-topic, and sometimes it’s funny. But once the novelty wears off, it’s easy to ignore, much like the AI feature set of other so-called AI smartphones I’ve used this year.
Apple needed to do something for an iPhone that was ‘built for Apple Intelligence’
Apple took its time getting to this point. The first set of AI features were removed in iOS 18.1 at the end of October. This includes notification and email summaries, generative writing tools, and cleanup tools to remove distractions from your photos. It felt like a very bare-bones viable product, but Apple needed to deliver something for an iPhone that was “built for Apple Intelligence.”
After months of beta testing, iOS 18.2 has been officially released with richer updates, including the Image Playground app for AI image generation, Genmoji, and the ChatGPT extension for Siri. Visual Intelligence is also available, but for unknown reasons only on iPhone 16 or 16 Pro. There’s more to come, of course, but Apple has finally shipped a set of AI features similar to Samsung and Google. The problem is that all these phone manufacturers still have a long way to go to deliver the AI smartphones we’ve been promised.
The big update to Siri in 18.2 is the addition of ChatGPT. Set timers and answer basic questions as before, but now you can send more complex queries to ChatGPT. It’s opt-in, so you don’t need an OpenAI account to use it. This is useful. It has the same tendency to make things up, but it serves as a useful starting point if you need help with complex topics.
On Android, Google’s AI-powered Gemini is now the default voice assistant. Although it lacked many basic features when it first launched, it has since moved closer to feature parity. Now you can set timers, play Spotify playlists, and brainstorm dinner ideas together. That’s one thing if you don’t know what to do with your wilting produce, but the real test comes when these voice assistants can perform actions on your phone. Both Apple and Google are working towards that, but for now the new AI-powered virtual assistants are just chatty versions of their predecessors.
Of all the updates iOS 18.2 brings, Image Playground is probably the flashiest. It’s a standalone app with a waiting list, but once you register it also unlocks image creation tools elsewhere throughout the OS. Image Playground is very similar to Google’s Pixel Studio, but with much stricter guardrails. This is almost always a good thing. A request to create an image of Pikachu with a paperclip stuck in an electrical outlet was denied, but this is good news for Pikachu.
It’s cute, and you can also make your spouse look like a chef or an astronaut. They’re kind of interesting if you’re into that. However, they are not immune to the pitfalls that many AI imaging tools fall into, often resulting in results that are completely incorrect or look downright weird. Like most AI tools, Image Playground is a bit of a guessing machine, and often produces pretty good images. But sometimes the guess is wrong, like when that version of avocado toast includes seeds and steam rising from tomatoes in the background of an image of a hot bowl of soup. And don’t ask them to make a move just because you don’t like what you see.
Genmoji is on an even tougher rail, and from my experience things work out just fine. However, it has some pretty obvious limitations, such as being so small that it’s hard to see details. It’s supposed to be small, but you might forget about it and get carried away and add so much stuff that you realize it’s barely visible in the final product. The depiction of me drinking a red coffee cup in front of a Christmas tree is well rendered and looks great in the preview, but it’s impossible to parse at normal emoji size. It also doesn’t work well with group texts using RCS, so you can’t reply to family texts with offensive emojis, which is the main use case for this feature.
New in 18.2 is the ability to tell the AI to pronounce sentences a certain way. No longer are you limited to the words “professional” or “friendly.” You can also make it sound like Mr. T wrote your email. That way, you get a lot of “poor fools” thrown in. That’s kind of interesting, so I’ll give you a little compliment.
But like the image generation feature, this feels like a gamble on the AI table at the moment. When ChatGPT first debuted, we all had a good laugh by having them write a sea shack about farting. If not, we are very sorry to report that ChatGPT is very good in this regard. Having this feature built into your phone doesn’t make it any more convenient. It’s actually more of a pain to try to use this feature with iOS’ text selection and formatting tools. Pity the idiot who bought the iPhone 16 with the promise of Apple Intelligence.
That’s the biggest problem with cell phone AI right now. Most of the time it does what it’s supposed to do. But it rarely helps and doesn’t feel like it’s solving the real problem I was having. That’s my complaint with this year’s devices from Google and Samsung. Now, Apple is at least joining the conversation. But they’re all in the same position, with the same pressure to deliver something in 2025 that’s more than just a collection of funny tricks. The novelty is wearing off quickly.
