The Future of Smartphone AI: Beyond Gimmicks to Practical Solutions

Don't judge Apple Intelligence by today's summaries, smartphone AI is going to be amazing... eventually

It wasn't a great year for smartphone AI. While AI tech advanced in leaps behind the scenes, the actual AI products we got to use seem less useful than ever. Even though AI computations make the best camera phones fantastic, we feel like we're being exposed to more bad AI than good AI. I'm here to offer a ray of sunshine through the cloud! AI is going to be very good for mobile devices. It's going to make everything so much easier that you won't ever think about having AI on your phone. Before then, AI will have to overcome its bad rep. As I'll explain, AI is going to be everything, but we're not there yet. I don't think today's situation is due an AI failure. I blame people, not AI. I'm hopeful about AI - I predict it's going to be an inseparable aspect of mobile computing in a couple of years. You won't consider buying a phone without it. Today's AI isn't even version 1.0. The features we have today - generative AI for creating images and text, live chat agents - aren't the AI, but rather the tools that will build AI. These aren't finished products. They aren't even products at all, they're tricks that the nascent AI systems can perform on the way to being functional. I may be cynical, but look at the AI features that we're being force fed. We're getting image generators rife with stereotypes, copyright violations, and potential for fraud and violence. We're getting language models that offer inaccurate facts and incorrect summaries. Minor features like handwriting recognition and photo lighting effects now rely on cloud computing, draining untold power and adding steps to the process for minimal usability gains. Everywhere you look, pushy AI features try too hard. I can't write an email or text message without my AI assistant offering to help screw it up with gobbledygook. In Google Photos, new AI tools offer to search for the pics I want, then return results that are useless compared to the older, AI-less search capability. New AI wallpaper tools are ubiquitous, but they create wallpapers that are impersonal at best, and they may be a bit racist at worst. Were we running low on wallpaper creators? Why do we even need generative AI for wallpapers? If these tools were all we could expect from AI, I'd say it's time to shut down the project and go home. It's time to look for what's next. I think that will probably be smart glasses, but it's clear that smart glasses, and any future mobile technology, will rely on AI and machine learning to deliver a robust experience. That's why I'm still optimistic about smartphone AI. I have to be optimistic. It's going to get better, a lot better. Today's AI doesn't solve any big problems for users. What's the biggest problem you have with your phone? The battery dies too quickly. Okay, AI might help with that, and I can attest that Qualcomm's newest AI-forward processors are also the most efficient high-end smartphone processors I've ever benchmarked. So, what else? Smartphones are still too difficult to use. I really want to use all the cool features, but I don't know how. I know that my phone can do a million and one things, including a million things I never use because I don't know how. Do you know how to share a photo by tapping your phone to another phone? Do you know how to charge your earbuds using your phone's battery? Have you ever created a safety Check In on your way home at night? I know that my phone can do all of that, but I'd need a refresher to figure it out. Honestly, I don't think I could change my home screen from Light Mode to Dark Mode without searching the Settings menu, and I'm a bona fide smartphone expert. AI will solve this problem. I won't need to know whether it's 'Reverse wireless charging' on my OnePlus versus 'PowerShare' on my Galaxy versus USB-C Power Delivery with my iPhone. I'll just tell my phone's AI what I want to do and it will turn on wireless power to charge my Galaxy Buds. I won't have to know that the iPhone hides the safety Check-In feature under iMessage, while Google hides the same feature in the little-known Safety app. I'll just tell my phone to 'let my Dad know when I've gotten home safely' and it will do the rest. Today's slate of AI features are parlor tricks. Watch as I pour beer out of this bottle without picking it up! Bet me that I can't insert Mickey Mouse into our group photo! See me compose a message using AI and no thought whatsoever! Okay, cool, but that's not something I want to do every day, or maybe even more than once. Thankfully, that's not the future for AI. In the near future, we're going to interact with what Qualcomm and other tech companies call AI agents. Think Siri, Gemini, and Bixby, but more capable and understanding. The AI is going to be the interface. You'll be able to swipe, or open apps, but you won't have to. Just tell the phone what you want, and don't worry about the AI. When AI finally makes our smartphone lives easier, instead of more complicated, it's going to be amazing.

Forget chatbots and magic erasers - my big AI wish for 2025 is much more boring

Please, Apple, give me more contextually aware notifications I can't help but feel like phone manufacturers are missing the point of AI. I've never felt compelled to superimpose my face into a group photo I wasn't a part of using Google's Best Take feature, or send my flatmate a weightlifting squirrel emoji using Genmoji on iPhone, or create infinite new wallpapers using Samsung's Generative wallpaper tool. Granted, all of these gimmicks come as part of broader AI toolsets that do contain genuinely life-improving features - live transcription tools, for instance, are a game-changer for me as a journalist - but Apple, Google, and Samsung seem intent on marketing the heck out of their most superfluous AI features, rather than showing me how their latest products can actually make my everyday life (not my social media life) easier. Ultimately, that's what AI should be used for, right? To improve (or simply speed up) the most mundane aspects of our everyday lives and grant us more time to enjoy the things we enjoy doing. That's why, in 2025, I'm hoping phone manufacturers focus less on generative silliness and more on making my daily commute a slightly less grueling experience. Yes, that's right - my big AI wish for 2025 concerns trains. Or more specifically, train times. Why is it that my AI-packed iPhone 16 Pro can accurately identify random objects in my immediate vicinity using Camera Control alone (that's Visual Intelligence FYI), but it can't automatically tell me when my train to Farringdon station is (or more likely isn't) due to arrive on a rainy Tuesday morning? In the same way that my Tesco Clubcard pops up every time my iPhone knows I'm near a Tesco (that's the equivalent of Walmart in the UK, US readers), wouldn't it be cool if it also memorized my frequent train journeys and provided me with real-time train times in a helpful pop-up? Contextually aware notifications like these would save me from having to open up the Trainline app, input my station origin and destination information, and scroll through the list of inevitably delayed trains twice a day. And yes, I appreciate that this is more of a plea to Trainline than to Apple - if my iPhone's dynamic Tesco pop-ups are anything to go by, Apple's software is already capable of delivering contextually aware notifications - but this seemingly simple request serves as a microcosm for the type of help I want AI toolsets like Apple Intelligence to give me. Any software that reduces the number of steps needed to move between apps greatly enhances the mobile experience. Chinese mobile brand Honor, for its part, seems to understand this desire more than most, with its Magic Portal interface - a shortcut feature that lets users switch between apps and services with a single swipe - proving both innovative and, crucially, useful in our testing of its latest phones (see our Honor Magic 6 Pro review and Honor Magic V3 review for more details). Magic Portal comes as part of Honor's broader 'intent-based UI', which is great, but it still requires an initial input from the user (be it a click, swipe or long-press) before delivering the desired information. I'm asking mobile manufacturers (and app developers) to go one step further in 2025: predict what information I want to receive on my phone, and serve it up to me *before* I go searching for it in an app or on a website. I want to spend less time on my phone, not more - Genmoji is just another time sink, but mobile software that helps to get me home from work quicker? Now that's what I call intelligence.