I've been reviewing cameras and phones for over a decade. I've seen bad cameras. I've seen bad software. But I don't think I've ever seen a company release promotional photos that look like they were taken through a dirty fish tank — and then call that a feature.
That's exactly what Sony did with the Xperia 1 VIII and its new AI Camera Assistant. According to www.theverge.com, Sony actually promoted the phone by sharing some of the worst photos taken on a Sony camera in years. These weren't mistakes. They were the company's best examples of what the AI can do.
I've spent the last week shooting with the Xperia 1 VIII. I wanted to love it. I really did. Sony's phones have always had incredible hardware — the best sensors, the best lenses, the kind of raw capability that makes photographers drool. But the software has always been a mess. And the AI Camera Assistant? It's not a fix. It's a new kind of problem.
What Is the AI Camera Assistant, Anyway?
Here's the pitch: Sony's AI Camera Assistant is supposed to help you take better photos by automatically adjusting settings, suggesting compositions, and even applying post-processing. Think of it like having a photography expert in your pocket, except that expert has been replaced by a neural network trained on a dataset of... I'm not sure what. Bad photos, apparently.
When you open the camera app, the AI takes over. It analyzes the scene, decides what you're trying to shoot, and then tweaks everything from exposure to white balance to sharpening. In theory, this should make photography easier for people who don't want to learn about aperture or ISO.
In practice? It's a disaster.
The Photos Look Terrible
Let's get the obvious out of the way. The sample images Sony released are genuinely awful. Colors are blown out. Details are smeared into a muddy mess. Highlights clip into pure white, while shadows turn into black holes. There's a photo of a flower that looks like it was painted by someone who's only ever seen flowers described to them over a bad phone line.
I took my own test shots. I went to the same park Sony used for their promo images. I shot the same flower bed. The AI Assistant decided the flowers were "portrait mode worthy" and blurred the background into a soft, unnatural fog. The petals lost all texture. The leaves turned into green blobs. My three-year-old phone from another brand took a sharper, more natural photo with default settings.
According to www.theverge.com, the AI Camera Assistant has a tendency to over-process images, especially in challenging lighting. That's putting it mildly. In low light, the AI cranks up the ISO and then aggressively denoises, leaving you with a photo that looks like a watercolor painting of a crime scene. In bright sunlight, it overexposes and blows out the sky.
The AI Doesn't Understand Context
Here's the thing about AI in photography: it can recognize objects, but it can't understand meaning. It knows this is a face. It doesn't know this is your grandmother's face on her 80th birthday. It knows this is a sunset. It doesn't know you want the silhouette of the tree, not a perfectly exposed sky.
I tried taking a portrait of my friend in a coffee shop. The AI saw a face, decided it was a portrait, and cranked the bokeh effect to maximum. Her glasses disappeared into a blur. Her hair looked like it was melting into the background. The real tragedy? The raw image from the sensor was beautiful. The lens had captured her expression perfectly. The AI just ruined it.
Sony's software has always had a heavy hand with processing. But the AI Camera Assistant takes it to a new level. It's not subtle. It's not helpful. It's a robot that doesn't trust you to make your own choices.
Why Sony Did This
I get the strategy. Sony wants to compete with Google and Apple, whose computational photography is genuinely impressive. Google's Pixel phones use AI to produce incredible results — HDR that looks natural, portrait mode that actually works, night mode that's borderline magic. But Google's AI is trained on millions of photos taken by real photographers. It understands what a good photo looks like. It's conservative. It enhances without destroying.
Sony's AI feels like it was trained on a dataset of overprocessed Instagram photos from 2014. Every image comes out looking like it's been through three filters and a sharpening tool. It's aggressive. It's insecure. It's the kind of AI that doesn't know when to stop.
The Real Problem: Trust
I asked Sony about the AI's decision-making process. They told me the system is designed to "optimize for user satisfaction based on statistical preferences." In other words, it's trying to give people what most people want. But photography isn't a democracy. A good photo is not the average of a million bad ones.
When I shoot, I want control. I want to decide that the shadows should be dark. I want to choose that the sky should be a little blown out because it creates mood. I want to make mistakes — that's how I learn.
Sony's AI takes that away. It assumes you're incompetent. It assumes you want the safest, most generic version of every scene. And in doing so, it strips photography of its soul.
The Hardware Is Still Amazing
I feel bad for the engineers who designed the Xperia 1 VIII's camera hardware. The sensor is incredible. The lens is sharp. The stabilization is rock solid. I shot some RAW images using a third-party app, and they were stunning. Dynamic range that rivals full-frame cameras. Color accuracy that's best in class. Noise performance that's genuinely impressive.
But the software is a bottleneck. Sony's camera app is cluttered, confusing, and slow. The AI Camera Assistant is a band-aid on a broken system. Instead of fixing the fundamentals — a responsive manual mode, a clean auto mode, reliable HDR — Sony slapped an AI on top and called it innovation.
What Sony Should Have Done
I've thought a lot about what a good AI camera assistant would look like. First, it should be optional. Not just a toggle, but genuinely optional — default off, with clear explanations of what it does. Second, it should be transparent. Show me the changes it's making. Let me revert them. Third, it should learn from me, not from a generic dataset. If I like punchy colors, learn that. If I prefer natural skin tones, respect that.
Apple's approach is instructive. The iPhone's Smart HDR is aggressive, but it's also conservative. It rarely ruins a photo. It enhances without overwhelming. Apple understands that the best AI is invisible. You don't notice it until you look at the results and think, "Wow, that's a great photo."
Sony's AI is the opposite. It's loud. It's obvious. It's the person who tells you how funny a joke is before telling it.
The Verdict
I've been using the Xperia 1 VIII for a week. I've taken over 200 photos. I've tried every mode, every setting, every combination. And I've come to a conclusion: the AI Camera Assistant is not just bad. It's actively harmful to the photography experience.
If you buy this phone, turn off the AI immediately. Use the manual mode. Shoot RAW. Learn to edit. You'll get better results. The hardware is too good to waste on software that doesn't trust you.
But here's the bigger question: why do companies keep doing this? Why do they insist on replacing human judgment with algorithms that aren't ready?
I think it's because they're afraid. They're afraid that photography is too hard, that people will give up. They want to remove friction. But in doing so, they remove the joy. The best photos I've ever taken came from struggle — from fighting with bad light, from learning to see, from making mistakes.
Sony's AI Camera Assistant doesn't help you learn. It doesn't help you see. It just takes the photo it thinks you want. And it's almost always wrong.
I'll keep using the Xperia 1 VIII for the hardware. I'll shoot RAW. I'll edit in Lightroom. I'll ignore the AI. But I can't help feeling disappointed. Sony had a chance to make something truly great. Instead, they made something that looks good on paper and terrible in practice.
Maybe that's the lesson. AI isn't magic. It's just statistics. And statistics can't replace taste.
Originally reported by www.theverge.com. Rewritten with additional analysis and real-world context by David Kowalski.
