Nvidia’s DLSS 5 isn’t a tool. It’s an invasion

DLSS 5 isn’t really DLSS at all

What we now consider “AI” began as generative AI art, where users asked services like Midjourney to produce computer-generated images via descriptive prompts. The results aren’t “art” in the traditional sense, yet the output is still technically impressive.

I’ve never looked at AI art as something to value. My home’s walls are full of art that we’ve bought from real local artists, not drawn by a computer—but I can still appreciate the way AI breaks down and analyzes writing in the same way that noir borrows heavily from icons like Dashiell Hammett, Hitchcock, and The Big Sleep. I’ve always appreciated the technical ability of generative AI to create images, but I always understood I wasn’t creating “art.” I was commissioning content.

In the meantime, of course, “AI” evolved into actual tools, like command-line instructions via Claude Code and various features within Adobe Photoshop. Now, even Nvidia uses it.

Nvidia

But as a number of my colleagues pointed out on the most recent The Full Nerd podcast, Nvidia’s first mistake was charactering DLSS 5 as a tool. It’s not.

While Nvidia’s Deep Learning Super Sampling (DLSS) feature is synonymous with performance improvements. You might not care how DLSS features like upscaling and frame generation work, but with those the AI is designed to make games feel smoother (“fake frames” or not). But with DLSS 5, that’s not the case at all.

Instead, DLSS 5 is merely a visual enhancement. Nvidia seems to want us to appreciate DLSS 5 with all the technical admiration we’d have for a generative AI art service like Midjourney, Udio, or Runway, but also to think of it as a practical, useful tool. It’s neither.

Nvidia’s examples suggest that DLSS 5 adds additional detail via generative AI where the original rendered graphics either suggested it or left it out altogether. In reality, the early demonstration—and yes, it’s just a demonstration—have added an “uncanny valley” commonality to familiar video game characters, prompting calls of “AI slop.”

And those are just the examples Nvidia supplied. Could Far Cry 3‘s Vaas end up with dimples? What about Darth Vader with rouged cheeks or lipstick? Shao Khan with dyed hair? AI can make mistakes, we’re reminded. Maybe that’s hyperbole… or maybe it’s not.

Nvidia’s CEO Jensen Huang has countered that users have gotten it completely wrong and that game developers remain in control of all of the creative elements they’re used to.

PCWorld’s Adam Patrick Murray, who saw the DLSS 5 demo first-hand, also seems convinced that we’re all wrong, and that the additional AI lighting and textures won’t detract at all—possibly the opposite, in fact.

That still doesn’t answer the question of who exactly benefits from DLSS 5 being turned on in the first place. And whose fault is it if something visually glitches, especially if that glitch varies by PC?

The artistic cost of DLSS 5

Whether you like it or not, AI is becoming more and more integrated into the PC. Microsoft and chip vendors are laying the foundation for AI to become integrated inside PC graphics. Intel, AMD, and Nvidia backed up Microsoft in introducing the next evolution of DirectX, the API responsible for playing video games on PC (and soon, the next Xbox). AI is also being used to develop shaders—”small model” instructions for rendering scene elements and “large model” for entire scenes.

Instead of compressing textures, Microsoft wants AI cores (like the Tensor cores in Nvidia’s GeForce RTX 5000 GPUs) to extrapolate what those textures “should” be. Same goes for lighting: AI will try and guess how light rays “should” bounce off objects. Instead of actually modeling light as it bounces off one object to other (path tracing), AI will guess.

Nvidia

Nvidia already uses the “large model” technology in existing versions of DLSS, where frame generation technology looks at the entire scene itself and tries to determine the path of a ball, for example, as it moves behind trees. This interjects “fake” frames, sure, but performance increases.

In this, I’m no purist. If Nvidia’s DLSS or Intel’s XeSS can make games accessible to more users simply by increasing frame rates, terrific! I think that’s a marvelous use of AI.

But games aren’t just technical achievements. They’re art, too. Art directors design games with purpose. The way a character looks defines who and what they are; environments are lit to lead and conceal and suggest. Would Expedition 33 be the same game with a nu-metal soundtrack and a design inspired by Pokémon? Of course not.

That’s hyperbole, I know, but my point stands: Even altering a game in subtle ways can have a profound impact.

The AI invasion continues

The fact that Nvidia is doing this via AI matters. Transforming familiar characters into plasticky, well-lit influencer clones is one thing. Games, however, feel like a safe haven against AI’s blitzkrieg into our daily lives… and so this kind of encroachment feels all the more violating.

I still think there’s a place for AI-generated content, even if it isn’t art. Art is what you pay for at the theater, or listen to, or read; content is the illustration accompanying the pamphlet on new linoleum that’s left in your mailbox. Art requires effort and insight and discussion; content is the machine summary of a Teams call you were too sick to attend.

Nvidia / YouTube

Nvidia thinks its additions will strengthen the artistic contributions of game developers. But Nvidia is wrong. DLSS 5 is simply the sprinkling of AI content on top of games, devaluing them in the process.

A number of my family members (including my wife and cousins) are creators: writers, journalists, voice actors, and stage actors. Fortunately, AI still can’t replace a human performance, especially to a live audience. But that performance isn’t just the work of the actor upon the stage, but also the set designers, the directors, the creatives who plan and run lighting and audio. Providing those creatives with cheap, effective tools is one thing—taking over their outputs is quite another.

Buying a video game means paying creators to give you a tour of their creative vision. It’s a contract and an agreement. One where you shouldn’t have to let Nvidia bully its way in and dictate to you what parts of the tour you see, or what the tour itself should look like.

Yes, you may be able to turn DLSS 5 off. But until then, I absolutely will defend the rights of consumers and creators to stand up, draw a line, and refuse to let Nvidia cross it. It’s a line that we actually can draw right now, and that in itself still feels meaningful.