Nvidia CEO tries to clarify why DLSS 5 isn’t simply “AI slop”

Final week, Nvidia’s public reveal of DLSS 5—and its “generative AI” enhanced glow-ups of gaming scenes—drew widespread condemnation from the gaming neighborhood. In a podcast printed Monday, although, Nvidia CEO Jensen Huang tried to distinguish the know-how’s elective, artist-guided graphical enhancements from the “AI slop” that Huang says he’s not a fan of.

As a part of a virtually two-hour-long interview with the Lex Fridman Podcast, Huang was requested to clarify the “drama” round DLSS 5 and “the players on-line [that] have been involved that it makes video games appear to be AI slop.” Huang responded that he “might see the place they’re coming from, as a result of I don’t love AI slop myself… the entire AI-generated content material more and more appears related and so they’re all stunning, so… I’m empathetic in the direction of what they’re considering.”

On the similar time, Huang stated DLSS 5 is decidedly separate from that type of “slop,” as a result of it “is 3D conditioned, 3D guided.” The artists behind a recreation are nonetheless those creating the in-game structural geometry and textures that type the “floor reality construction” that DLSS 5 works from, Huang stated. “And so each single body, it enhances but it surely doesn’t change something,” he stated.

For probably the most half, although, players haven’t been apprehensive about DLSS 5 creating trippy new content material from the bottom up like some generative AI world fashions. As an alternative, the concern is that DLSS 5’s visible “enhancements” might find yourself smoothing out many disparate video games towards a single, flattened, homogenized photo-realism customary.

That’s a misunderstanding of how DLSS 5 works, Huang stated. It’s not a know-how the place a recreation ships in a single state and “then we’re gonna post-process it,” he stated. As an alternative, DLSS 5 “is built-in with the artist, and so it’s about giving the artist the instrument of AI, the instrument of generative AI.”