Home / News / These smartphone photos prove that AI has its limits

These smartphone photos prove that AI has its limits

50
Image of sign and street taken in autumn

Quick links

It’s easy to think that AI can generate any type of image, but these tools have their limits. To test this, I took a series of smartphone photos to prove that AI can only go so far – some with a focus on aesthetics and others on practicality.



1 Burst Photo Series

One day, on my morning walk, I captured a series of photos on my iPhone of people cycling across the bridge. I wanted them to look a bit like a magazine or comic and show a specific sequence of events. As you can see, I managed to capture every few seconds between the bikes crossing the bridge.

While AI is good at generating single images, it is not really capable of generating series of images. For example, ChatGPT will often completely change an image when I ask it to make simple edits. Additionally, using the same prompt will often yield completely different results.


2 Photography sign

I’ve talked in depth about how generative AI is bad at creating text in images. While I sometimes get decent results using software like ChatGPT, these tools are often off the mark. I’ve had numerous experiences where the text was unreadable, so I won’t be relying on generative AI for these kinds of images anytime soon.

Trying to create text in AI images made me realize how often I take the simplicity and utility of smartphone cameras for granted. Whenever I want to take a photo of the sign, I simply point the camera at it and capture the shot.

Danny Maiorca/MakeUseOf

However, I also wanted to make the image a bit more interesting. So I made the emblem as part of the foreground. Generative AI tools can be good at creating bokeh and background effects, but because they fall short in generating text, these kinds of images are better if you shoot them on your phone.


3 The front lines

The generative AI is pretty good at creating landscape-style images, but I’ve sometimes found it lacking when I’m trying to create images with more complex guide lines. It’s fine on straight roads, but beyond that the AI ​​tends to struggle. On the other hand, it is much easier to reach the front lines with a smartphone.

Front line on the beach in summer
Danny Maiorca/MakeUseOf

In this example, I took the photo on the beach while standing on the sidewalk. As you can see, I captured the lens flare with my smartphone at the same time. I loved how the light shone on the sidewalks and it’s pretty cool how the sidewalk starts to bend to the left as it reaches the background.


I could potentially create a similar image using generative AI, but it wouldn’t reach the main element of the leader line. However, you may want to try some key AI tips and tricks to increase your chances of achieving the desired result.

4 Night smartphone photography

The generative AI can take some interesting night shots, but it tends to fall short of getting the colors right. In my opinion, generative AI is better for daily shots. In this nighttime smartphone photo, I specifically wanted to capture a building with an interesting shape. This kind of architecture is usually more difficult for a generative AI tool to replicate.

Photo taken early in the morning with light reflecting off the water
Danny Maiorca/MakeUseOf


The reflections of the light in this picture are also unique. While AI can create reflections in water, it can make them look more like watercolor paintings than the actual reflections you’d see in photos. If you want to take even better night photos on your smartphone, consider trying these night photography apps.

5 Silhouettes

Shooting silhouettes can be very difficult, and to be honest, I often find it more of a stroke of luck than an actual attempt. Silhouette photography isn’t just about making something look darker, you also need light to hit your shot just right. For example, in the image below, the light hit the modern building on the left at the right angle.

Silhouette image of cityscape in the morning
Danny Maiorca/MakeUseOf


It’s also difficult for AI to mimic the different hues in the sky in silhouette photos, even with specific challenges. Additionally, I feel that these tools often exaggerate shadows; if I want my pictures to look like this, I just use a photo editing app on my phone.

6 Macro photography

AI can be good at creating close-up shots, but with some limitations. While these tools are great for generating images of people, I’ve found that they struggle with images of plants. To test this, I took various macro shots of flowers and leaves.

Photo of flowers in a city courtyard
Danny Maiorca/MakeUseOf

My smartphone camera was able to capture the different textures of these plants, which was exactly what I wanted. It was also good to capture true colors without oversaturating them; generative AI tools usually do this with flying colours. However, as I discovered when trying macro photography with my smartphone, this style takes a lot of practice.


7 Crouch down

I’m always trying to find new angles when shooting with my smartphone, and crouching is one way to do that. I can ask an AI platform to create an image that looks like I did, but it’s unlikely to succeed. Since I like to have more control over composition, it’s easier with a smartphone.

Image of a building with signs
Danny Maiorca/MakeUseOf

The characters in this picture were my main area of ​​interest and I went under the bridge first before crouching down further. As a result, I also managed to saturate the sky in my photo. Asking an AI tool to do this probably wouldn’t give me what I wanted.


8 Frames within frames

The AI ​​is good at generating images that only have one or two layers, but things get more complicated when I try to add more elements to the image. For example, I like capturing frames within frames; ChatGPT holds its own in this area, but the results are far from perfect. To try and do something the AI ​​couldn’t do, I took the image below.

Bridge and street with car
Danny Maiorca/MakeUseOf

I wanted to include a car at a distance, which is something the AI ​​would struggle with. Since it often struggles with different patterns, I don’t think it would be able to create lines on both sides of the bridge – the AI ​​tends to be better at curves.

Artificial intelligence has its advantages, but it is not reliable for every type of image. Smartphone cameras can often outperform artificial intelligence, and while this may change in the future, I still find smartphone photography to be excellent.


Comments