Within hours, NVIDIA unveiled DLSS 5. The technology transforms images in real-time, injecting new shadows, lighting, textures, and material properties through AI inference. In theory, democratized photorealism. In practice, something uncanny.
The results divided immediately. To some, inevitable progress—performance headroom, visual enhancement, the next benchmark to chase. Digital Foundry, influential technical analysis outlet, championed it enthusiastically. Their framing treated games as products to be optimized: frame rates, fidelity metrics, efficiency gains.
To others, the technology looked lifeless. The DLSS 5 filter produces renders that read as technically proficient yet spiritually vacant—photoreal without being photographic, detailed without being deliberate. The shadows fall correctly by statistical inference, not artistic intent. The lighting obeys physical plausibility without emotional purpose.
This fracture exposes a deeper tension. The debate over whether videogames constitute art has exhausted itself into irrelevance. More productive is recognizing that two distinct frameworks compete for dominance in how we make, evaluate, and consume interactive entertainment. Call them Games as Art and Games as Product—not as mutually exclusive categories, but as gravitational centers pulling development in different directions.
The Art framework treats the game as an authored experience. Every element serves intentional expression: lighting crafts mood, mechanics embody philosophy, the player’s receptive attention completes the circuit. This aligns with what writer David Dennen calls the “aesthetic attitude”—a contemplative stance where the object guides perception and “overt doing is reduced to a minimum.”
The Product framework treats the game as a commercial object. Success metrics include attach rate, monetization efficiency, hardware attach, and benchmark performance. The player here is active, goal-directed, engaged in what Dennen terms the “ludic attitude”—but instrumentally, as a user rather than a witness.
Photorealism sits at the intersection, apparently neutral. The Art camp can deploy it for immersive worldbuilding, controlled atmosphere, emotional verisimilitude. The Product camp can deploy it for market differentiation, “next-gen” justification, premium pricing. Both pursuits look identical on screen. They diverge only when the cost structure fractures. DLSS 5 is that fracture made visible.
The Catalyst: The Easy Button
A Hollywood VFX shot gets minutes, hours, days to render. A game gets sixteen milliseconds. Bridging that gap with brute force—throwing more hardware, more people, more years at it—has driven budgets into the hundreds of millions and timelines into the absurd. Final Fantasy VII, one complete game in 1997, now spans three separate releases. First dropped 2020. Second, 2024. Third? Who knows. Naoki Yoshida actually apologized for this. “The release intervals have gotten longer,” he said. Younger players who grew up on Fortnite’s constant churn encounter Final Fantasy as these expensive, isolated events—years of silence, then a $70 monument to visual fidelity they didn’t ask for.
Square Enix didn’t rethink the ambition. They just announced they’d use AI to sustain it. Same appetite—Advent Children made playable, the photoreal dream that Spirits Within nearly bankrupted them pursuing—now chased with automation instead of craft.
And they’re not alone. Every AAA publisher is staring at the same math. DLSS 5 looks like salvation: cut render costs, shorten timelines, get back to annual releases like the old days. But there’s always a cost. We’d get games sooner, sure. They’d just be hollowed out—soulless husks, though plenty would argue we’re already there.
The jobs disappear first. Concept artists who establish the visual identity. Lighting designers who actually feel their way through a scene. Composers recording with live ensembles. The art is where it starts. Songwriters don’t begin with an orchestra—they start with a guitar, a piano, a melody. Those foundational roles? Replaced by programmers babysitting AI systems, tweaking sliders on outputs they didn’t create.
Voice actors are fighting this now—striking to keep ownership of their own voices. AI companies want to record once, model, then cut them out. Same curve for music: synthetic orchestras hitting the same emotional beats at zero marginal cost. The extraction logic spreads. Surface stays, substrate thins. Doomer shit, maybe. But this is Square Enix we’re talking about. Remember: this is a business.
NVIDIA calls DLSS 5 “the GPT moment for graphics.” Telling. LLMs generate plausible text without understanding a word of it. DLSS 5 generates plausible images without understanding beauty. Jensen Huang promises “control for artists,” but the control is a joke—sliders on top of inference, masking tools for lighting the artist never placed. The “ground truth” render becomes theoretical. Nobody sees it. Nobody plays it.
This isn’t liberation. It’s extraction wearing innovation’s face.
The Killshot: The Confession
AI models, when pressed, concede the same limitation: they do not understand beauty. They emulate, interpolate, pattern-match. They generate outputs statistically likely to be perceived as beautiful based on training data. But apprehending beauty—the receptive, contemplative stance Dennen describes as essential to art—remains inaccessible. Correlation, not resonance. Plausibility, not purpose.
DLSS 5 is this limitation rendered in pixels. It does not know why subsurface scattering matters in this scene, this moment, this emotional beat. It knows only that certain vector inputs correlate with certain photoreal outputs. The shadows fall correctly by inference, not intent. The lighting obeys physical law without artistic law. NVIDIA promises “the GPT moment for graphics.” The framing is honest: GPT moments are moments of sophisticated non-understanding. Statistical coherence without meaning. Surface without depth.
The Product framework does not require beauty understood. Only simulated sufficiently to trigger purchase. And so games chase photorealism through tools explicitly incapable of apprehending what made photorealism valuable—not as metric, but as potential vessel for human intention. The extraction completes. The player receives an image no human fully authored, accompanied by voices no human fully performed, set to music no human fully composed. The “game” functions. It benchmarks well. It sells.
The machines were not built to know beauty. They were built to approximate its appearance. The rest of us are forgetting the difference.
Finale: Which path will you choose?
The rest of us are forgetting the difference. DLSS 5 fuels this forgetting. It renders games as product more efficiently than any prior technology—not by failing to work, but by working too well for one framework and not at all for the other. The Product camp gets what it needs: frames, fidelity, function. The Art camp gets a simulation of what it wanted, statistically likely to resemble beauty without ever being authored as such.
And some players genuinely prefer product. You see them every Game Awards cycle—the eSports competitors bitching that their games don’t get recognized alongside Expedition 33 or Astro Bot. They don’t feel because the game gives them feelings. They feel because they create feelings through competition: the clutch, the outplay, the ranking achieved. The game is a system to master, not a space to inhabit. When that’s the case, all that matters is output. Does it run at 240Hz? Does the hit registration feel crisp? Does the dopamine hit clean?
They don’t sit with the ending of Hollow Knight. Don’t ruminate over choices in Baldur’s Gate 3. Don’t hope Kratos finds redemption. They want the points, the goal, the shot. And they are not wrong. The ludic attitude is valid. Dennen was right about that. But they also will not care if humans made their games. AI-generated textures, synthesized announcers, procedural arenas—so long as the loop functions, the origin is irrelevant. The framework is Product all the way down, and DLSS 5 serves it perfectly.
The question is whether you live there too. Whether your games are things you use or things you receive. Whether you benchmark your experiences or sit with them. Whether you can name the last time a shadow made you feel something, and whether you cared if a person placed it. The fissure is not in the industry. It is in you. The machines have chosen their side.
What about you?
About Author
