DLSS 5: Has Nvidia’s AI graphics technology gone too far?

DLSS 5 is not just another iteration of Nvidia's graphics upscaling technology — it's a breakthrough moment in which the GPU chip giant decided to venture into territory where generative artificial intelligence directly modifies the visual aspect of games we already know and love. Nvidia announced its latest technology as "the most significant breakthrough in computer graphics since the debut of real-time ray tracing in 2018". It sounds impressive until you see the actual results. Then it turns out that what was supposed to be the future of gaming looks to many players like the latest version of "AI slop" — that phenomenon we've already come to know in photographs, films, and other digital media.

Reactions to DLSS 5 are decidedly mixed, but critical voices have dominated public discourse. Players aren't hiding their frustration: characters from Resident Evil Requiem look like they've been through an Instagram filter, distinctive faces have been "beautified" to the point where they lose their original identity. Even Virgil van Dijk, the real and well-known captain of Liverpool, in EA Sports FC looks like some other guy. This is not what players expected, and it's not what game developers intended.

The most interesting part of the whole affair? Nvidia CEO Jensen Huang, responding to criticism, simply said: "Well, first of all, they're wrong". This statement says far more about corporate mentality than about the technology itself. Nvidia has a problem — and it's not a technical one.

When upscaling became an "AI slop" generator

DLSS 5 is a fundamentally different beast than previous versions of the technology. Previous iterations of DLSS used machine learning to intelligently scale images from lower resolution to higher resolution, closing the gap between graphics settings. It was practical, efficient, and most importantly — invisible to players. It did its job, and we didn't worry about it.

DLSS 5 changes the rules. Nvidia describes it as "geometry-guided 3D neural rendering model", but those are marketing words. In practice, the technology applies generative artificial intelligence to rebuild the game's lighting and materials in real time. This means the game doesn't just scale the image — it changes it. It adds details that were never in the original code, modifies lighting, changes textures, and most importantly — changes character faces.

When you see a side-by-side comparison of the original version of a game with the DLSS 5 version, the difference is striking. The lighting is indeed more realistic, shadows more natural, materials more shiny. But there's one problem: everything looks identical. All games go through the same generative filter, and all come out with the same characteristic "AI look" we already know from other media. This is exactly what people mean when they talk about "AI slop" — that particular aesthetic that is the result of training algorithms, not artistic choices.

Sean Hollister from The Verge compared it to motion smoothing for televisions — a technology that manufacturers added to TVs to "improve" movies by interpolating additional frames. The problem was that it made everything look weird, unnatural, and completely changed the experience that directors and producers had planned. DLSS 5 does exactly the same thing — but instead of interpolating frames, it interpolates faces, lighting, and materials.

Artistic intent versus algorithmic "improvement"

This is where things get really interesting — and equally frustrating — the tension between what Nvidia says it does and what actually happens. Nvidia claims that DLSS 5 "preserves artist control" and "honors the intent of the original creators". Jensen Huang said that developers can "fine-tune generative AI" to their needs.

In theory, this sounds good. In practice? Problems start quickly. First, generative AI is a black box. Even if a developer can "tune" the model, they don't have full control over what changes the AI will introduce. This is a fundamental feature of neural networks — they are probabilistic, not deterministic. You can't exactly say: "Change this lighting in this specific way". You can only say: "Improve this lighting" and hope that the AI does it in a way you like.

Second, the examples shown by Nvidia suggest that "tuning" mainly means changing the intensity of the effect, not its direction. Grace from Resident Evil Requiem didn't become more realistic — she became more "beautified", more smoothed out, more like what the algorithm "thinks" should be beautiful. This is not artistic intent — this is algorithmic fantasy.

For Polish game creators, of which there are increasingly more in the industry, this should be a serious warning. DLSS 5 could be an Nvidia-imposed standard that changes the look of their games without their full control. If you want your game to look a specific way, DLSS 5 could be a problem, not a solution.

Nvidia lost touch with gaming reality

The most interesting thing about this whole story is how Nvidia approached it. Instead of promoting DLSS 5 as a tool for future games, a tool that would allow developers to create more realistic experiences from the ground up, Nvidia decided to show how DLSS 5 "improves" existing games. This was a strategic mistake.

When you show players that technology can "improve" their favorite games, you send the message that the original games look bad. That Capcom didn't know how to light a scene in Resident Evil Requiem. That the creators of Hogwarts Legacy didn't know how to make characters look natural. This insults the artists who worked on these games and frustrates players who love them.

Nvidia could have said: "Here's a tool for future games that will allow developers to do things that were previously impossible". Instead, it said: "Your favorite games look bad, but we'll fix them". This is exactly the opposite of what it should have done if it wanted to interest the gaming community.

Jensen Huang said that players "are wrong". But maybe Nvidia is wrong. Maybe players know exactly what they want, and it's not a game that looks like every other game that's been through the same generative filter. Maybe artists know exactly what look they want to achieve, and it's not an algorithmic compromise.

The technology itself is not bad — it's a matter of application

To be fair to Nvidia, the technology itself is not bad. Real-time neural rendering is a fascinating engineering achievement. The ability to modify lighting and materials in real time has real applications. In the future, when developers design games from the ground up with DLSS 5 in mind, it could allow for things that were previously impossible.

The problem is how Nvidia sold it. Instead of promoting DLSS 5 as a tool of the future, it promoted it as a tool to "fix" the past. And that is a fundamental communication error.

It's also worth noting that Nvidia has real control over how this technology is perceived. When it shows demos, it can choose which games to show, how to show them, and how to describe them. When it chooses games like Resident Evil Requiem and shows how DLSS 5 changes character faces, it sends a specific message: that AI can "improve" upon the artist. This is a dangerous message, especially at a time when artists are already worried about artificial intelligence.

Memes, backlash, and the gaming community

The internet quickly reacted to DLSS 5 in a way that Nvidia probably didn't anticipate. Memes spread like wildfire. Players compared DLSS 5 to Instagram filters, to AI art generators, to "yassification" — a term used to describe situations when AI changes something in a more "glamorous" and simultaneously less authentic way.

This is not just a technical backlash — it's a cultural backlash. Players feel that their favorite games are being attacked, that artistic intent is being undermined, that an algorithm knows better than the artist. This is a deep feeling, and it's justified.

For Nvidia, this backlash is a problem because DLSS 5 is supposed to be adopted by developers. If players are actively reluctant to have games use DLSS 5, developers may be reluctant to implement it. We're already seeing some studios declare support — Bethesda, Capcom, Ubisoft — but the question is: will these be major AAA games or more niche titles?

Industry context: where Nvidia is wrong and where it's right

Nvidia is not alone in trying to integrate generative AI into games. Other companies are experimenting with similar ideas. But Nvidia has a unique position — its GPU chips are used by almost all game developers. This gives it enormous power, but also enormous responsibility.

Where Nvidia is right: real-time neural rendering technology is indeed impressive from a technical standpoint. The ability to modify lighting and materials without having to rebuild the entire scene is real progress.

Where Nvidia is wrong: the assumption that players want their favorite games to be "improved" by AI. The assumption that artists don't know what they're doing. The assumption that everyone will be delighted when they see how DLSS 5 changes character faces.

These are fundamental errors in understanding what players and artists actually want. And they are errors that could have long-term consequences for DLSS 5 adoption.

The future of DLSS 5: is this a "GPT moment for graphics"?

Jensen Huang called DLSS 5 a "GPT moment for graphics" — comparing it to the breakthrough that ChatGPT brought. This is a bold declaration, but it's at least premature.

GPT was a breakthrough because people wanted it. People saw the value it brought and wanted to use it. DLSS 5 has a problem — people don't want it. At least not in the form it's currently being presented.

It's possible that DLSS 5 will have a future in new games, where developers can design with the technology from the ground up. It's possible that over time, as artists learn to work with AI, the results will be less "AI slop" and more artistic. But right now, DLSS 5 looks like a technology looking for a problem to solve, rather than a solution to a real problem.

For the Polish gaming market, which is developing dynamically with studios like CD Projekt Red, Techland, or Bloober Team, DLSS 5 should be a warning. Not every new technology that Nvidia offers is worth adopting. Sometimes the best solution is to stick with the original artistic vision, rather than let an algorithm "improve" it.

DLSS 5 is not a bad technology — it's a bad idea applied at the wrong time in the wrong direction. Nvidia had a chance at something truly special, but instead it created something that divided the gaming community and raised concerns among artists. This is a lesson for the entire AI industry: technology must be in service of artists and users, not against them.