AI on smartphones is in a battle with time, and it’s losing

It has been two years plus change since ChatGPT was unleashed upon the world, and the tidal wave of AI buzzwords has basically flooded the smartphone market. And, just like clockwork, every new feature we get on a new smartphone is “AI”, and every popular app in the Play Store and App Store was updated to have “AI” in it.This is not the first time this has happened. Back in 2017, smartphone manufacturers tried to talk up AI for the first time. Huawei boldly claimed that the Mate 10 is an “AI smartphone” because it has real-time translation, scene detection for the camera, and other such tidbits. However, tech pundits online were very quick to criticize such marketing, saying that it’s less “intellect” and more “automation.”Other manufacturers were more careful with their marketing, even though that was also the same time when Apple and Qualcomm started including dedicated neural engine parts in their chips. However, the course-correction led to the use of the more proper term Deep Learning – a subset of Machine Learning.
But make no mistake, there was about a year or so when “Deep Learning” was everywhere. Every event and conference, every popular image editing app or weather app suddenly had Deep Learning in its title. Neural Processing somehow became important when it came to marketing smartphones and apps.

The public AI-n’t impressed

Back to the present day. It has now been a whole year since we saw the first smartphone with a full AI suite of its own launched. That’d be the Galaxy S24 series and Galaxy AI. Since then, we’ve had Google enhance Gemini’s capabilities and Apple launch a three-stage assault on the smartphone world with timed releases of AI feature packs.

And a lot of the features they throw at us… Well, nobody is impressed. Yes, a couple of them are useful, and I myself have praised… Eh, one — the AI Summarize option on Galaxy phones that lets you boil huge articles (like this one) down to a bullet list. Apple’s AI can also do this if you open the page in Reader Mode, and Gemini will also probably do it if you ask it nicely.

There’s also the new feature on iPhones and Pixels where an AI will listen to your phone conversation and automatically take notes for you. If you set a date, if someone gives you a location or phone number, you don’t need to grab a pen or anything; you can just check the annotated call log with bullet points, and find that info.

Admittedly, these are impressive and genuinely useful features. Two of them.

Then, we have Magic Erasers, which will work fine if you are trying to delete a photobomber at the far back of a photo, but will often give you a freakshow of a regenerated background.

We have AI text editor, to both inform us of how our “tone” sounds in a text or email, and also rephrase and change the tone to sound differently, if you choose to do so. Nothing like adding some human touch with the help of your AI assistant.

There are image generation apps on board — Pixel Studio, Apple Image Playground, Google Sketch to Image… they are all bad or ridiculous enough to be a neat party trick, which you forget about a month into owning the phone.

And then, there’s the overall perception of AI in the tech community, artist community, creator community. It quickly became clear that a lot of the AI engines out there — text, image, video, music — have been trained on the work of people. Surprise, surprise. Scraping hours of video, tons of pictures and drawings, more hours of music, and all of that for years and years — that’s how AI “learned” to create. Well, it took a couple more years to get hands right.

So, there’s a controversy brewing online. Regulators are still trying to figure out what to do with and how to treat any of these AI creations — is it original, copy, or plagiarism? And people online are still using their personal moral compass to determine how they perceive AI and its usefulness to humanity. And there are a lot, a lot of people that are not happy with it.

In the meantime, smartphone companies are promising us features that solve problems that don’t exist

It was incredibly awkward to see Samsung pitch the new set of AI features with the Galaxy S25 series. For example, you can now tell the Galaxy AI to change something in settings, instead of looking for it for yourself and tapping away at the screen to get to it.

Admittedly, a very minor but good quality-of-life improvement. But no! It’s AI, so we need to market it as a big deal.

Here’s how Samsung chose to present it — “Imagine you’ve had a long day, and your eyes are getting strained. You can now tell Galaxy AI that your eyes are strained, and it will intelligently suggest to activate blue light filter, and increase font size”. What an incredible solution — and here I am, simply looking away from the phone screen when my eyes hurt.

Nicely produced ad. Here’s the question — how did the “AI features” presented here expand the user experience or improve the hero’s life?

Apple wasn’t doing much better. It launched a string of Apple Intelligence commercials, which were widely hated. Because they depicted lazy, undependable, and not very smart people using AI to get some sort of job done in the very last minute of a work day. Or, even worse, to compose an email asking a co-worker to do their work for them (somehow, the AI is able to compose it in a very convincing way, see…).