For decades, improvements in video game graphics have followed a relatively simple promise: greater fidelity to developers’ visions. NVIDIA’s new AI-powered graphics technology, DLSS 5, complicates this promise.

At a basic level, DLSS (Deep Learning Super Sampling) is designed to make games look and run better. It uses AI to upscale lower-resolution images, which allows for higher performance without sacrificing visual quality.

Previous versions, including the very well-received DLSS 4, mostly worked behind the scenes to subtly improve graphics. DLSS 5, however, goes further. Rather than simply filling in gaps, it uses generative AI to actively reinterpret what’s on screen, adjusting lighting, textures, and faces in real time.

The backlash to NVIDIA’s DLSS 5 was immediate, visceral, and pretty hilarious. Across social media, players described it as “AI slop,” an “Instagram filter for games,” and a system that “yassifies” characters into something smoother and shinier.

What was meant to be a leap forward in graphical technology instead felt, to many, like a step sideways into something uncanny.

That reaction isn’t just knee-jerk anti-AI sentiment.

DLSS 5 raises an uncomfortable question: if a game’s look can be rewritten in real time by an AI model, where does that leave artistic vision?

Video game visuals are rarely about realism alone. That’s something I love about them. They are designed. Every shadow, texture, item, and character model is the result of deliberate choices. Nothing is put into a game by accident. Studios hire teams of artists to reflect the mood and identity of their worlds.

And crucially, that doesn’t always mean looking “better.” Games often prioritise style over realism. Colours are pushed up, facial features are exaggerated, and worlds are designed to be distinctive rather than realistic.

Design choices make games recognisable at a glance. Would Fallout feel the same without its retro-futuristic aesthetic? Or Disco Elysium without its painterly, almost oil-brushed look?

Lighting artists, character artists, and environment designers can spend days, sometimes weeks, refining how a single scene is meant to feel. When a system like DLSS 5 steps in and begins adjusting those choices in real time, that work, those painstaking hours of human labour, are instantly overridden.

By introducing a generative AI layer that can “enhance” lighting, textures, and faces, DLSS 5 risks flattening the very things that give games their personality.

Sure, the result may be more conventionally realistic, but it’s also less distinctive, as if every character has been passed through the same Instagram filter.

AI systems are, by design, very good at producing averages. They learn from vast amounts of data and generate outputs that sit somewhere in the middle of it all, polished, coherent, and broadly appealing. But art doesn’t usually live in the average.

It lives in exaggeration and imperfection. And that is being stripped away by this new technology.

The gap between console and PC visuals has narrowed significantly in recent years, and games are increasingly prioritising art direction over raw technical showmanship. There’s only so much more realistic you can get before the improvements stop being noticeable.

This is the problem of diminishing returns – something the games industry has been grappling with for years.

You only have to look at the leap from Pac-Man to GTA 4 to see how far graphics improved from the 80s to the 00s. But then look from GTA 4 to GTA 5, and the difference feels far less dramatic. The improvements are there in textures and animation, but they’re incremental rather than transformative.

On this plateau, progress becomes harder to sell.

DLSS 5 feels like an attempt to solve this problem. Not by rendering games more accurately, but by changing how they’re rendered altogether. If you can’t push fidelity much further, you redefine what fidelity even means.

Normally, that would just be one idea among many. But NVIDIA isn’t just a small company throwing things at the wall. It controls the overwhelming majority of the graphics market, with over 80% of players on Steam using its GPUs.

In recent years, it has grown into one of the most valuable companies in the world. When a company that large decides what “better graphics” looks like, it doesn’t stay a suggestion for long.

That’s what makes this feel less like an experiment and more like a forced direction. Even if players don’t like it, even if developers are hesitant, it can still become standard simply because of how widely NVIDIA hardware is used.

You don’t have to opt in for it to shape the future aesthetic of games.

And that’s the uncomfortable part. This isn’t just about whether DLSS 5 looks good or bad. It’s about how much control one company should have over the visual language of an entire medium.

Something that shouldn’t be ignored is the fact that NVIDIA is one of the primary companies supplying hardware for the current AI boom, with its GPUs powering the vast majority of AI training data centres.

Billions have been poured into AI over the past few years, but it’s still not entirely clear how it’s supposed to make that money back.

The systems are expensive to run, energy-hungry, and often feel like solutions in search of problems.

Thus, NVIDIA has a clear incentive to find new ways to embed AI into everyday technology, whether or not those uses are genuinely needed. Their push to wedge AI into more consumer tech starts to feel obvious.

If you’ve spent years building the thing, you kind of need somewhere to put it. Not necessarily because games were crying out for it, but because the technology exists and needs a reason to be there.

So, who asked for this?

The answer, as usual when a bad idea shows up, is the people selling it. Not players, who were already happy with how games looked. Not developers, who spend years carefully crafting those visuals in the first place.

This isn’t a response to demand. It’s a direction being set from the top down.

DLSS 5 doesn’t solve a meaningful problem. It takes something that already works, something built on deliberate artistic choices, and runs it through a system that flattens those choices into something more uniform and more “acceptable”. It replaces intention with approximation.

Perhaps I’m being cynical, but this doesn’t feel like progress to me.

Image credit: Unsplash