Nvidia CEO Jensen Huang has once again commented on criticism of DLSS 5 technology, attempting to dispel the concerns of players and developers. In a new interview with Lex Fridman, he admitted that he understands the audience's dissatisfaction with the abundance of monotonous AI content.
According to Huang, he himself does not like the so-called «neuro-slop».
I understand the point of view of critics, because I myself am not thrilled with such content.
He also added that he regrets such a reaction from the community.
According to the director of Nvidia, DLSS 5 works differently. The technology is «conditioned and guided», which means full control of artists over the geometry of the scene. It is these data that are used by the system to generate the image, while the structure of objects remains unchanged in each frame.
Additional details about the operation of DLSS 5 were previously revealed by Nvidia employee Jacob Freeman. He reported that the technology uses a 2D image of the game and motion vectors as input data, based on which it forms the final frame.
At the same time, despite possible visual changes, the basic geometry of objects is not affected. Freeman also noted that the previously demonstrated examples — including the scene with characters from Resident Evil Requiem — are only an early version of DLSS 5, and its final capabilities may change significantly.
