ASUS’ ZenFone 5 stretches the limits of the term ‘AI’

Too bad no one can agree on a definition.

As expected, ASUS officially revealed its new ZenFone 5 in Barcelona today, and if you don't mind I'm going to skip my usual meandering intro. The company talked up a handful of AI features in its latest midrange smartphone, but I think ASUS is throwing around the word "AI" haphazardly. More on that later though: Let's get the usual hands-on stuff out of the way first.

Yes, with its notched 6.2-inch screen and its vertically mounted dual camera, the ZenFone 5 looks quite a bit like an iPhone. (Just, you know, bigger and with a striking finish on its rear.) The 19:9 screen is particularly nice (and that's not a typo): Colors appear much brighter and punchier than I had hoped, and it's longer than the standard 18:9 displays so videos played at full resolution aren't obscured by the notch. All told, generally a pleasure to gaze at, as long as you're not looking for a super high-resolution panel.

Inside, you'll find one of Qualcomm's Snapdragon 636 chipsets with either 3GB of RAM and 32GB of storage or 4GB RAM and 64GB storage. While it isn't the fastest machine I've tried at MWC, navigating was mostly painless -- I'm willing to chalk up the handful of performance issues I encountered to our phone's non-final hardware. If the truly complete ZenFone 5 runs as smoothly as this one did most of the time, people searching for a midrange phone would do well to keep this thing in mind. And while I couldn't offload any of the sample images I took, the 12-megapixel main sensor seemed to capture sufficiently detailed images with respectable dynamic range. It has an f/1.8 aperture too, which should help out a lot in low light (which I didn't get to test) -- too bad the 120-degree, secondary wide-angle camera was mostly just OK.

ASUS hasn't told us how much the ZenFone 5 will cost, but everything about it screams above-average midrange. There's nothing wrong with that, but the way ASUS describes some of its features rubs me the wrong way. When the company introduced us to the ZenFone, a spokesperson proudly talked about its "10 AI features." The problem is, ASUS is playing pretty hard and fast with the way it defines artificial intelligence. This is nothing new: While AI has become more accessible and more relevant, it has become clear that there isn't one true definition of artificial intelligence.

ASUS has staked its position inside that gray area. Some of those AI features it spoke about to us didn't seem to rely on AI as we know it at all, and I'm concerned that ASUS is overselling things by banking on the general confusion that comes with talking about artificial intelligence. The company's position is that the definition of AI has shifted in recent years toward one that involves some level of machine learning. In applying the AI label to many of the ZenFone 5's features, ASUS is simply sticking with an older, broader definition and hoping the average consumer won't notice (or care about) the difference.

Consider the ZenFone 5's ability to change its screen's color temperature in different situations. The same feature is called Truetone on Apple's most recent iPhones, and ASUS concedes there's no machine learning going on. Ditto for another feature that keeps the phone's screen on and unlocked while you're looking at it. If that sounds familiar, it's because Samsung Galaxy S phones have been doing that since the GS3. Again, the company was clear: No machine learning involved here either.

Meanwhile, a feature called Power Boost automatically manages the ZenFone 5's performance and provides an extra dose of oomph in certain intense situations. We've seen similar features in action in devices like Huawei's Mate 10 Pro, which leaned on an algorithm that was trained to help the phone understand how to best tweak performance over time. In the ZenFone 5, ASUS conceded that the Power Boost feature didn't rely on machine learning. Actually, ASUS wouldn't confirm how the feature worked at all, aside from saying that it isn't maintaining a white list of apps that are allowed to push the processor harder than others. Well, what is it then?

To be clear, I'm not saying these features aren't helpful. The ones that actually worked in the preproduction sample we tested actually held up well -- the screen never turned off while my eyes were on it, and Power Boost did seem to work for certain applications. What's really odd is that ASUS actually built some features into the ZenFone 5 that rely on machine learning to improve performance over time. Consider the camera. After taking photos for between one and three weeks, the ZenFone will start to offer edited versions you might like.

If you find them pleasing, you can accept those changes, and that feedback will help shape the way the camera processes images in the future. If you don't like them, you can dismiss them and never worry about them again. Also in the camera is an intelligent-scene mode that interprets what's in the frame and fires up the correct scene mode. Since we were inside an office on a chilly Barcelona afternoon, there weren't many sweeping vistas to test this feature with. The phone quickly launched its sunset, food and flower modes when pointed at printed pictures of, well, sunsets, food and flowers. Squeezing this kind of functionality into a midrange smartphone is good news for consumers, and for that, at least, ASUS deserves credit.

Ultimately, whether or not these are actually AI-powered features is a philosophical question. As long as these features work -- and my current sense is that they do -- most people probably won't care about the distinction. Even so, I got the impression that ASUS wanted a nice, round number of buzz-inducing AI features it could tout at its very first smartphone launch at Mobile World Congress and applied the AI label too liberally. I don't think that ASUS is necessarily pulling a fast one, but I do think it's being at least a little disingenuous.

Catch up on the latest news from MWC 2018 right here.