The DLSS 5 lie...

A tongue-in-cheek header image...

 

DLSS 5 has exploded into our lives the last couple of days and it's been one hell of a ride. However, I think a lot of this comes from not understanding the tech because Nvidia's CEO keeps lying...

 

 

 Let's get to the central conceit of this post - Nvidia's Jensen Huang has said the following:

 

"Twenty-five years after NVIDIA invented the programmable shader, we are renventing computer graphics once again. DLSS 5 is the GPT moment for graphics - blending hand-crafted rendering with generative AI to deliver a dramatic leap in visual realism while preserving the control artists need for creative expression."

"[...] DLSS 5 fuses controllability of the of [sic - Tom's Hardware] geometry and textures and everything about the game with generative AI. [...] It's not post-processing, it's not post-processing at the frame level, it's generative control at the geometry level. [...] This is very different from generative AI; it's content-control generative AI. That's why we call it neural rendering."

 

These statements are nonsensical and contradictory and this is for a simple reason - there is very little truth here.

 

DLSS 5 doesn't touch geometry, it doesn't touch the textures...

If you look at the actual technical information provided by Nvidia in their release and the short summary image above, you can see that DLSS neural rendering is simply a more advanced neural net guided lighting filter. It's a more advanced version of ENB and Reshade because it is fed the data from the frame (and prior frames) just like the DLSS upscaler and ray reconstruction. It's generating data as much (or maybe slightly less) than those two technologies do and that's why it's stable - that's why it's not mush that is uncontrollable, which real Gen AI is. 

 

DLSS 5 is Machine Learning (ML).

 

From this perspective it's obvious why this is placed under the umbrella of "DLSS" instead of having its own technology nomenclature.

 

The big problem has been that Jensen's priority is not gaming or communicating to gamer: his priority is keeping the AI investment economy chugging along for as long as he can. Through that lens, his comments and messaging makes complete sense. 

 

Gen AI has no commercial product - it is not making money. Jensen, re-labelling this tech as Gen AI gives him something to point to, "Hey, we're shipping this! Look, it's a real product! Gen AI has real applications! It's not a bubble - we have infinite growth!" He's talking to investors, the market, Wall Street, etc. 

 

Chat GPT isn't Gen AI. It's an LLM. Yes, it can be linked to Gen AI through agentic means but it's not Gen AI. Generative AI can't be hand-crafted or controlled. If it could be, it would be stable, repeatable, and not require multiple generations to get what you want as an output.

 

 

Look at the flow description of DLSS 5 - it's literally a post-processing effect. It needs the colour, the textures, the heightmaps, buffers, etc. to be in place for it to work. The higher the quality of those things, the better it will be. What some people are freaking out about is not on the cards, at all... It's literally post-processing at the frame level.

 

"DLSS 5 takes a game's color and motion vectors for each frame as input, and uses an AI model to infuse the scene with photoreal lighting and materials that are anchored to source 3D content and consistent from frame to frame."

 

The problem is that not only are these lies and misdirections from Jensen hurting the brand in consumers eyes, they're creating confusion as to what, exactly, this technology is. Twitter has been a cesspool of people hating on this technology simply because he used the words "generative AI", people are completely unable to tell that the detail is being brought out of the underlying models and textures.

 

Worse still, it's exposing how poorly people understand lighting. I've been an amateur photographer for 20 years and I worked at a Newspaper Techdesk, editing photos for print for 2-3 years. I'm not an expert but I have a lot of grounding in how things look and work from a lighting standpoint and the images shown of DLSS 5, while not perfect, are more correct from a lighting standpoint. I've seen many people nitpicking the images and most of them are wrong/not considering all of the facts of each scene. 

 

I could spend an inordinate amount of energy trying to convince them by refuting some or all of their claims but it's not worth it for me, plus I have things going on in my life that demand more attention. At the end of the day, we will see proper implementations towards the end of the year and then the proof will be in the pudding. So, I won't waste time arguing about half finished tech demos which can easily be fixed.

 

 

Conclusion... 

 

Coming back to the point of this blogpost, Jensen is lying in his public statements in order to keep the AI bubble from popping, the stock price of Nvidia high, and his huge bonuses rolling in. He is saying things that are contradicting the technical information released by Nvidia on their website and in-person to press and users at the GTC reveal event.

 

Even looking at the available information allows people with some technical background to decipher what DLSS 5 actually is. However, many are ignoring the information and not using their brains. I think this technology has a lot of potential - just like how ReShade and ENB have been incredible tools for enhancing the look of older games. We now have a superior tool that can be more deeply integrated into game engines for developers to use that enhances the available art in the game - and requires higher quality, more detailed art for better results.