Image : Mark Hachman/IDG
Jensen Huang, Nvidia’s chief executive, appeared on stage to answer questions from reporters at CES, a day after launching Nvidia GeForce GeForce 50 series graphics cards.
One of the most important questions: In a future where AI is increasingly being used to create or interpolate frames, will the end result be a PC graphics world that is completely AI-generated? No, Huang replied.
We asked Huang this question for a reason. Nvidia claims that DLSS 3 can inject AI-generated frames in between every GPU rendered frame, but DLSS 4 is able to infer three frames from a single traditional one, as Brad Chacos pointed out in our previous report on the GeForce F50-series reveal.
AMD was asked the same question a day earlier. “Can I tell that in the future every pixel will be ML [machine learning]-generated? Yes, absolutely. Frank Azor, AMD’s chief architect for gaming solutions, responded.
Huang disagreed. “No,” he answered Adam Patrick Murray of PCWorld.
“The reason is that when ChatGPT was first released, we said Oh now, let’s generate the bookBut no one expects it. Huang continued, “And that’s because you need credit,” Huang said. “You have to give credit — that’s what condition is. Now we condition the chats or prompts by adding context. You must first understand the process before you can answer any questions. The context can be PDF or a web search. You could have told it what context you want, right?
And the same goes for video games. You must give a context. The context of video games must not only be relevant from a story-telling perspective, but also be relevant in terms of world, space, and spatial relevance. You can condition it by giving it early pieces of geometry or textures. It will then generate the rest.
The context in ChatGPT is called Rapid Retrieval and Augmented Generations (19459042), the context that guides the textual output. He said that in the future 3D graphics will be 3D ground condition generation. Huang said that Nvidia GPU rasterization engine renders only one out of four forward-looking frames in DLSS 4. “So out of four frames with 33 million pixels we only rendered two [million]” “Isn’t this a miracle?” “But it has taken that long for everyone to realize that generative AI really is the future. But you need condition, you have to ground with the artist, the [and the] intent.”
Mark Hachman Senior Editor, PCWorld.
.
Mark is a technology writer who has been writing for PCWorld since the last decade. He has written over 3,500 articles, covering PC microprocessors and peripherals, Microsoft Windows, and other topics, for PCWorld. Mark has written for PC Magazine, Byte and eWEEK as well as Popular Science, Electronic Buyers’ News and Electronic Buyers’ News. He also shared a Jesse H. Neal Award with Popular Science for breaking news. He recently gave away a collection consisting of several dozen Thunderbolt Docks and USB-C Hubs, because his office has no room.