When NVIDIA CEO Jen-Hsun Huang participated in the Lex Fridman podcast, he responded to the controversial remarks about DLSS 5 at GTC 2026, and his attitude was significantly softened than before.At GTC 2026, Huang Renxun once said that players’ criticism of DLSS 5 was completely wrong, which caused strong dissatisfaction. A large number of players called DLSS 5 AI garbage (AI slop).
This time in the podcast, Huang Renxun said it differently: "I think their point of view is understandable, and I can see why they would react this way, because I don't like AI garbage myself.More and more AI-generated content these days looks the same...so I can relate to them. But DLSS 5 is not designed to do that. "
However, Huang Renxun’s explanation of the DLSS 5 technology itself is basically the same as that of GTC. He emphasized that DLSS 5 is content-controlled generative AI. It is neither a general generative AI model nor a post-processing filter, but a technology between the two.
He said that DLSS 5 will be faithful to the geometry set by the artist, "every frame is enhanced, but nothing will be changed."
Huang Renxun also described the future possibilities of DLSS 5. Players can even control the picture style through prompt words, such as "I want cartoon rendering", or provide a reference picture, and DLSS 5 can generate the picture according to that style, while maintaining consistency with the artist's original intention.
Huang Jenxun finally compared DLSS 5 to the previously launched skin shader, saying that it just has one more tool that developers can choose to use or not.
DLSS 5 is expected to be officially released later this year, and specific technical details will not be made public until then.
