even though many people may see this as disrespectful to artists who work with games, I don’t think that’s entirely true. I do agree that it looks a bit exaggerated and has that AI-generated appearance, but as stated in the dlss5 notes, this is an adjustable filter for developers (so the intensity of that effect can be reduced)

in addition, in my opinion this doesn’t seem to significantly deform the original mesh, but rather works as a filter to improve lighting and resolution. in the end even if many people don’t want to accept it, the hardware required to run something with similar lighting (path tracing) is extremely expensive. running this in real time while also saving hardware resources using AI (nru) is impressive and can help in many aspects of games, not only with lighting but also with the realistic deformation of organic materials, which is extremely difficult to achieve in a fully realistic way even for major industries

I would reject the idea if it took a completely blocky, low-definition model and turned it into something entirely different, but this is just a lighting correction filter. I believe that over time this will become something essential for the advancement of rendering, not only helping in games but also in 3D in general, where it will be possible to save render time, achieve better results and create more realistic materials more easily (and over time it will likely lose that “AI-generated” look)

rejecting nru, in my view, is overlooking a clear technological advancement. 3D modelers who value quality in their work, when comparing these two images and not noticing any improvement, may be overlooking some important points or simply being more resistant to this kind of change, and following that same logic it would also be fair to question the use of cycles (path tracing) in game teasers, since that is not real-time rendering either. saying that nru is wrong while cycles is acceptable ends up being a comparison that to me doesn’t make much sense, which is why I don’t fully understand all the backlash

in the end this is just my opinion for now, I’m still analyzing this technology further but depending on the use case it could be very interesting to see how it evolves in practice

I might get hated by other developers for saying this, but I don't deny that the quality of real-time rendering will become much more interesting from now on. :p

Note: Without the artists' base, the AI ​​can't improve the image, so well-made models are still absolutely necessary. Therefore, it's more of an upscaling process than content generated entirely by AI. I really don't understand all the hate.

Note2: The only problem with DLSS is that every frame looks like a perfectly edited photo, but it would be nice to see it adjusted with a more realistic tone and less contrast.