This year GamesBeat was part of the 2023 Game Awards in L.A., as part of a mini GB event, “Pixels to Pop Culture,” exploring the intersection of game industry creatives, Hollywood and a growing audience’s hunger for new IP.
Jeff Skelton, head of technology partnerships at Electronic Arts joined Meg Tucker, director of games at Google Cloud, to talk about today’s challenges in creating immersive content in a crowded market, and how new technology and tools are unlocking opportunities to really deliver.
“Every game in a genre has to be bigger than the game that came out before it in that genre,” Skelton says. That means literally: the file size of an average texture in a game today is larger than the memory of the original PlayStation. Games are commonly 150 gigabytes-plus, even before the DLC. The sheer size and amount of content required has multiplied, but the number of industry creatives hasn’t — which demands new technology and processes to allow game creatives to manage it all, and do it well.
“It’s allowing them to do it quickly and with high quality, allowing them to be able to express their intentions without roadblocks getting in their way,” he adds. “Sometimes they have a very quick turnaround. They don’t want to be super under the gun doing a bunch of boilerplate or stuff that isn’t actually doing the creative things that we want creators to be doing. It starts with the tooling, allowing that to happen more fluidly than it used to.”
That includes everything from the data and analytics within a game that allows developers to keep track of player activity and game performance, to the advances in generative AI that allow one-to-one personalization.
The central goal of adopting AI tools, both generative and predictive, is to allow artists to be more efficient and more productive in what they do as creators by eliminating the roadblocks in their workflows, Skelton says. They might have three days to figure out an urgent patch, but they’re also waiting a day and a half for a pipeline to run and process data as they work, for instance.
“Anything we can do to reduce all that time and allow the artists to do what they want to be doing, which is doing inspired work that can only come out of their heads — it’s not coming out of generative AI,” he explains. “We see all of our tooling as human-centered and human-directed tooling. It’s being controlled by the creators that want the output from it. They have very fine control over what parameters go into it, what they want and they can iterate on that.”
In other words, these tools are a productivity multiplier that accelerates the creative process, and the end result, when artists have more time and space to create, means higher-quality work, and faster.
For example, terrain is notoriously time-consuming to create, even for accomplished artists using more advanced graphics tools, especially the initial stage in which the artist needs to block out the whole area before starting the detail work. But now, with LLMs and native language interfaces, a creator can request twenty variations on 16 square kilometers with two mountains, a valley, a river and a medieval village, and continue to iterate on the best configurations until they find one that works as a foundation for their creative process, and it’s done in a fraction of the time hand-tweaking a landscape would.
What’s even more valuable, Skelton says, is the combination of generative AI and traditional predictive AI.
“We make enormous use of procedural tools, we have for a decade; they’ve been the reason we’ve managed to keep up with content creation,” he explains. “If you mix in procedural manipulation along with generative AI, that artist can then take that 16 square kilometers, put it in their favorite modeling tool, and then drag around the river. I don’t want it to be quite as windy. I want to unwind it a bit. Everything around it will adapt.”
They can shrink and grow the mountain, ask the tool to add new features like shrubbery and trees, species dependent on their altitude and more.
“There are people that love doing things like that, but in general, most artists don’t want to go hand-plant 5,000 instances of trees and calculate, well, this tree wouldn’t live at this altitude, but it would be at that altitude,” he adds. “Doing things like that lets that artist have so much creative control.”
It also slashes the time it takes to create a massive amount of complex objects — one might take six staff months’ of work, and 150 would take 60 staff years, but procedural tooling has made all the difference in a teams’s ability to create and deliver content for players. A sports game gets more teams and stadiums, a racing game gets more cars, an RPG gets a bigger, more complex world to explore, and it’s all in 4K.
A year ago, generative AI was amazing. Today it’s interesting. Tomorrow it’s going to be default, tablestakes, a must-have in the game industry, whether that’s building game assets, letting engineers off the hook by creating the tedious unit tests that take up so much of their jobs, tackling an overflowing inbox or producing decks for presentations in planning meetings with a tool like Google Duet AI.
To reach the broadest audience, a studio needs to support the broadest array of platforms, from old mobile phones all the way to top consoles and the highest-end PCs and GPUs. EA supports all of them with its innovative internal engine, Frostbite. It allows content creators to make something once and then port it into a variety of platforms, to empower developers and delight players.
“In the world where there’s cross-media, whether it’s going out for marketing, or whether it’s videos for YouTube, or assets that could potentially go to a movie or TV show, having the tooling that allows you to have that variety of output is important,” Skelton says. “We’re going to continue to see that. We see that in the industry already with some of the other engines in the industry that are doing that very well right now, publicly. We have to be able to support everything, everyone, everywhere. Give our players what they want, where they want it.”