Jensen Huang Backtracks on DLSS 5: “I Don’t Love AI Slop Myself”

Jensen Huang just blinked. Days after telling gamers they were “completely wrong” about DLSS 5 at GTC 2026, Nvidia‘s CEO appeared on Lex Fridman Podcast #494 and took a markedly softer tone — admitting he understands the backlash and doesn’t like AI slop either.

DLSS 5 Controversy: Timeline at a Glance

What Exactly Did Jensen Huang Say?

On the podcast, Huang said: “I could see where they’re coming from, because I don’t love AI slop myself… all of the AI-generated content increasingly looks similar and they’re all beautiful, so I’m empathetic towards what they’re thinking.”

He then quickly pivoted, insisting DLSS 5 is structurally different — “3D-conditioned” and “3D-guided,” meaning it enhances visuals without disrupting artist intent or geometry. Huang also suggested that developers could use it to create wildly different looks, even non-photorealistic styles, through customisable model training.

Why Gamers Aren’t Buying It Yet

The controversy was ignited by a DLSS 5 demo over a Resident Evil Requiem scene featuring Grace Ashcroft — her face was rendered with smoother skin, sharper features, and an uncanny look, and even Capcom’s artists were reportedly seeing it for the first time at the event. Nvidia had better comparison screenshots available that showed DLSS 5 preserving character facial structure, but chose not to lead with them — handing critics a ready-made example of exactly the AI slop they feared.

The deeper tension is a technical contradiction. Huang claimed DLSS 5 operates at the geometry level, not as a 2D filter — but a Nvidia GeForce Evangelist confirmed the model uses “a 2D frame plus motion vectors as input,” directly contradicting Huang’s framing. Until DLSS 5 ships in actual game engines with developer-tested results, gamers remain sceptical. Read our analysis of Nvidia GPU news and RTX 50-series updates on TechnoSports.

FAQs