Tasteless

Nowadays, I wake up in a constant state of apprehension that something blatantly outrageous will have happened in the tech industry overnight. Cortisol level is still at an all-time low as a consequence of my poor life habits, but the industry, on the other hand, never fail to deliver something egregious.

This morning, Nvidia just released DLSS 5, which turned the non-problem of wanting you video game to look like photorealistic slop into a perfectly working implementation.

Some people are gonna enjoy it, the same way they enjoy chocolate with cheese, or soda with haute cuisine.

But, why is it infuriating you would ask? This, once again, a move that is profoundly disrespectful toward artists.

When creating a game, all artists involved into making it make conscious decisions about the aesthetic, the colors, the lighting, the rendering, the shaders, in order to create an immersive experience.

Borderland worked because of its unique cel shading look. Lethal company was a hit, partly due to its glitchy pixelated aesthetic that works well with the horror theme of the game. Binding of Isaac, and recently Mewgenics have a unique graphic style, and you can tell who made the game just by the look of it.

Pixelated used to be a technological limitation but now, it’s an artistic direction.

If the author of a game wanted it this way, why not respect their decision? Are these tech bros heartless, or tasteless? Apply hyperrealistic filters, is both.

Should we also apply these filters to Mona Lisa? Does it make sense take artists with innovative art styles like Van Gogh or Monet and bring them back to the realistic norm?

Music production has also suffered from this, years before AI was a thing. There was a point in time where almost everyone started to compress the dynamics of the sound so that there is no nuance in volume anymore and quantize everything so that it sticks to the tempo grid. Music is a living things, tempos, as played by humans may vary. Rage Against the Machine should not be quantized. Chopin’s is tempo rubato’s expressiveness would be lost if quantized to the grid. There is simply no grid there!

Quantization and compression would be the equivalent of taking Mona Lisa, pixelating it and reduce the palette to only four bright vibrant colors.

Another thing is that these AI enhancement tools are non-deterministic. It means that as the seed changes, the rendering will be different. Drawing a face is hard, even a single misplaced millimeter makes it look totally off. And off it looks! Quoting a fellow developer Jean Baptiste Dusseaut: "It is uncanny valley as a service!"

How are we supposed to build a shared experience, as humans, if the characters look different from one player to another?

So what’s happening? Are we so afraid of imperfection that our ideal aesthetic is hyperrealistic and standardized? It’s 2026 and everything must become slop. If we let the blinds drive innovation, everything will soon look like a cliff.