According to NVIDIA and its CEO Jensen Huang, the AI PC era began with GeForce RTX and the 20 Series of desktop GPUs in 2018. For PC gamers, this meant using AI to make real-time ray-tracing a possibility with the arrival of RTX hardware and AI Tensor Cores. DLSS, or Deep Learning Super Sampling, is the not-so-secret ingredient.
Fast-forward to 2024, and AI-powered DLSS upscaling and DLSS 3 Frame Generation (which uses AI to generate entire frames) have pushed PC gaming performance to new heights across a wide range of hardware. With games like Cyberpunk 2077 and Alan Wake 2, we're now at a point where full Path Traced lighting is a reality.
At Computex 2024, we saw two examples of how NVIDIA plans to empower game developers with new AI technologies. NVIDIA ACE helps create digital humans with whom you can interact. Project G-Assist uses AI and a general model trained by NVIDIA to provide in-game assistance, gameplay guidance, and even performance optimization, all through natural dialogue or text inputs.
It's fascinating (and impressive) stuff. During a roundtable discussion with Jensen Huang at Computex 2024, we asked about the future of game development and AI. With DLSS and Frame Generation solely focused on performance and bringing impressive ray-tracing effects to PC gamers, what about the creative side?
"You might have seen the work that we did with Getty on generative images and textures. It's trained," Jensen Huang tells me. "We've created a world-class generative model but trained completely on licensed content. You can now use that model because it's been trained on licensed content to create other content that you can resell. And so we call that model Edify."
"Edify is utterly world-class," Jensen continues. "You could use it to create all kinds of content; you can also use Edify to create 360. So, environmental domes for reflections and image-based lighting (IBL), that kind of capability could be possible with this dome rendering, not just two-dimensional, but 360 renderings, photorealistic objects with HDR."
AI-generated textures, art, and environments? Check. However, the big hurdle is AI's ability to create 3D objects with accurate physical properties and dimensions, which NVIDIA is focused on evolving.
"You probably have seen some of the early works we've done, from text to 3D," Jensen Huang explains. "You describe a 3D object, and we generate a 3D object you might have seen, and that's called Edify 3D. You might have seen some of the amazing work that we've done with NeRFs to render, infer, and generate 3D views. And so you could see NVIDIA doing a lot of work here in generative content."
Jensen Huang adds that the earliest real-world examples of 3D models created by AI will come via the content creator space. Still, once the technology and AI models become "more sophisticated over time," they will "literally be inside the game."
His and NVIDIA's vision for gaming's AI future is impressively ambitious. From generating higher resolution textures on-the-fly so games take up less physical disk space to AI co-op partners that join you (and remember) you on your next adventure.
"We're going to get richer and richer games in the future," Jensen Huang says. "Inside the game, all the characters will be AI. If you go into a battle with six of your colleagues and you know all those six colleagues, maybe two of them are real people, the other four are AIs. And they've been playing with you for a long time, so they remember you. And so you're going to have AIs that you're playing along with, [and] you'll have AIs that you're adversarial with."
- Read more: NVIDIA Project G-Assist - your in-game RTX AI Assistant for guides, tips, and overclocking
- Read more: NVIDIA App updates adding major improvements to ShadowPlay and 'one-click GPU tuning'
- Read more: NVIDIA announces 'SFF-ready' stamp of approval for RTX graphics cards
- Read more: Got a GeForce RTX GPU? Well, you have a powerful GeForce RTX AI PC
- Read more: First digital human running locally on a GeForce RTX PC is here
- Read more: RTX Remix goes fully open-source to remaster as many classic games as possible