Nvidia CEO Defends DLSS 5 Against “AI Slop” Criticism | Ars Technica

Nvidia CEO Jensen Huang is attempting to address mounting criticism surrounding DLSS 5, the latest iteration of the company’s Deep Learning Super Sampling technology. While some gamers have lauded the potential for improved performance, a significant contingent has expressed concern that the “generative AI” enhancements introduced in DLSS 5 are resulting in visuals they describe as “AI slop”—a homogenized, artificial aesthetic that detracts from the original artistic intent of game developers. Huang’s response, delivered during an interview with the Lex Fridman Podcast, frames the issue as a misunderstanding of the technology’s core function and a personal distaste for the kind of unguided AI generation he believes leads to these undesirable results.

How DLSS 5 Differs from “AI Slop”

Huang acknowledges the concerns, stating he “could witness 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 attractive, so… I’m empathetic towards what they’re thinking.” The core of his argument rests on the distinction between DLSS 5 and what he characterizes as purely generative AI models, like DeepMind’s Genie 3, which can create entirely new content from scratch. Genie 3, unveiled in 2025, demonstrated the ability to generate interactive simulations, raising questions about the potential for AI to fundamentally alter content creation. Huang insists DLSS 5 is different because it’s “3D conditioned, 3D guided.”

Essentially, DLSS 5 doesn’t invent new visual elements; it enhances existing ones. The technology works by leveraging the game’s original geometry and textures – the foundational elements created by artists – as a “ground truth structure.” Huang explains that DLSS 5 “enhances but it doesn’t change anything” on a frame-by-frame basis. In other words the underlying artistic vision remains intact, while the AI component focuses on refining details and improving visual fidelity. The process relies on a form of neural rendering, using AI to infer how features *would* look if they were more photorealistic, but always within the constraints of the original artistic direction.

The Root of the Backlash: Homogenization Concerns

The criticism leveled against DLSS 5 isn’t necessarily about the creation of entirely new, unwanted content. Instead, many gamers fear that the visual “enhancements” will smooth out the unique artistic styles of different games, leading to a flattened, homogenized aesthetic. Examples cited frequently in online discussions include changes to character appearances in games like Resident Evil Requiem, specifically the altered looks of Grace Ashcroft and Leon Kennedy. Tom’s Hardware reported on the widespread criticism following the GTC 2026 reveal, noting the vocal response on social media.

Huang dismisses this concern as a misunderstanding. He emphasizes that DLSS 5 isn’t a “post-process” applied to a finished game. Rather, it’s designed to be integrated into the artist’s workflow, providing them with a tool to leverage the power of generative AI. This integration, he argues, ensures that the technology serves to enhance the artist’s vision, not override it. The intent is to provide artists more control over visual fidelity and detail, rather than imposing a uniform aesthetic across all games.

The Role of Artist Control and Generative AI

The key distinction Huang draws is the level of artistic control. In his view, DLSS 5 isn’t about AI taking over the creative process; it’s about AI augmenting it. The artist retains control over the fundamental structure and textures of the game world, while DLSS 5 provides a means to refine and enhance those elements. This approach contrasts sharply with fully generative AI systems, where the AI has greater autonomy in creating content. The integration of DLSS 5 is intended to be a collaborative process, with the artist guiding the AI to achieve the desired visual outcome.

However, the extent to which artists will *actually* have control over the AI’s output remains a key question. The effectiveness of DLSS 5 will depend on the quality of the tools provided to artists and their ability to effectively guide the AI’s generative process. If the tools are too limited or the AI’s behavior is unpredictable, the risk of homogenization will remain. The success of DLSS 5 hinges on striking a balance between AI-powered enhancement and artistic control.

What Comes Next: Integration and Refinement

Nvidia’s immediate focus is on working with game developers to integrate DLSS 5 into their titles. The company will likely provide ongoing support and training to artists to help them effectively utilize the technology. The initial response from the gaming community will undoubtedly influence the direction of future development. If concerns about homogenization persist, Nvidia may necessitate to refine the technology to provide artists with greater control over the AI’s output. Further, the company will likely monitor feedback closely and address any technical issues that arise during the rollout process. The long-term success of DLSS 5 will depend on its ability to deliver on its promise of enhanced visual fidelity without sacrificing the unique artistic styles that define individual games.