Generative AI is already being used in developer workflows. We expect AI to further augment content creation pipelines by generating images, audio, video, and code.
Generative AI is faster than most consumer products and attracting a high level of VC investment, even in today’s tough funding environment. The applications for generative AI are vast, for us at Remagine Ventures, naturally the impact it can have on gaming is really significant. So, will generative AI play a role in gaming? How is it being used today? Will it work? At GDC, the Games Developer Conference, in San Francisco next week, generative AI is sure to be one of the hot topics.
Personally, I believe that game development is a creative endeavour that includes both technical skill and a level of creativity and originality that would may be one day be possible for an algorithm. For example, take Grand Theft Auto. It would be hard to algorithmically generate the gameplay, the storyline, the cohesiveness of it all as one cool story/experience. But I believe that we’ll absolutely get there one day. In the meanwhile, AI is here not to replace game developers, but to be a helper, writer, tutor, debugger, sketch artist, etc. It’s a step on the way to creative automation. I’m sure this will be a hot topic in GDC next week.
In this post, I’d like to point out a few of the way that Generative AI could impact gaming, feature some of the leading resources on generative AI in gaming and a few of the live examples out there. Let’s dive in.
The landscape of generative AI startups in gaming is rapidly evolving
The first market map of generative AI in gaming was published by the A16Z gaming fund in November and the team published an updated version today. The landscape today includes 57 startups. Notable companies include Character.ai, which recently raised about $200M to create chatbots that can impersonate a persona.
In my small data sample (of Israeli gaming startups) a lot of startups in the generative AI gaming space are either missing in this map, or are choosing to remain in stealth until they have a solid demo out. Advances in AI like GPT-4 or Midjourney 5 will accelerate the development of this space and I believe that in a year’s time the size of this landscape will at least double if not triple, considering that traditional gaming companies and studios are incorporating generative AI into their existing products.
Jon Radoff expanded on this a bit more in his mapping of generative AI companies who build virtual worlds, published in Feb 2023.
Create 2D and 3D content with Generative AI
The games industry spends about £54 billion a year on content creation. The more complex the games, the more detailed design work they require. Generative AI can dramatically reduce the cost and the speed of creating 2D and 3D content for games, but there are a few nuances worth keeping in mind, see below section of the challenges of generative AI gaming startups. There are a number of companies operating in this space including Scenario, Kaedim, Sloyd, 3dfy.ai and Israeli startup Genie Labs.
Text to game
There are a number of companies in stealth taking the generative process beyond assets, to full fledged (short) games. I’ve seen a few demos and while there’s still work to be done, it’s passable for a short session. It’s interesting to see how this takes different forms with the various genres, from hypercasual through casual mobile games and all the way to AAA games. Companies operating in the space (not in stealth), include Latitude, Ludo (2D games) and Israeli startup Sefi AI.
Create a 3D game like Doom with GPT-4
GPT-4 is out from this week and it’s an extremely powerful tool. A single prompt can be use to generate code in GPT-4:
In this thread, you can see how it’s possible to generate a 3D game like Doom based on a prompt.
It’s a multistep process and we have to set the expectations low, but still, a couple of years ago this would look like science fiction, and today it’s within reach. Can’t wait to see how this tech develops.
Everyone can become a creator
When it comes to user generated content, Roblox is king. As a company, Roblox is responsible for building the virtual world where 50% of US households spend over 3 hours a week playing. But the content, the games inside the virtual world, are 100% reliant on user generated content. Rather than try to resist generative AI, Roblox is leaning in: building tools to allow creators to develop integrated 3D objects that come with behaviour built in.
In a post on the Roblox blog Roblox’ CTO Daniel Sturman said:
Today, Roblox provides creators with a platform that enables end-to-end tools, services, and support to help them build the most immersive 3D experiences. With Roblox Studio, creators have everything they need, out-of-the-box and for free, to build their experiences and publish immediately on all popular platforms, reaching 58.8 million people daily worldwide. With the advent of generative AI techniques, however, we are seeing an opportunity to revolutionize creation on the platform, both by augmenting Roblox Studio to make creation dramatically faster and easier, and also by enabling every user on Roblox to be a creator.
Generative AI on Roblox: Our Vision for the Future of Creation
Take a game like Grand Theft Auto, a single player game where you’re the leading character. A lot of your experience is determined by the interactions with NPCs, or non-playing characters. The development of AI NPCs in gaming is now a rapidly growing area, with companies like InWorld AI at the forefront of the field (InWorld AI recently raised $50M series A). AI NPCs have the potential to make games more immersive and engaging experiences for players, create realistic dialogue, generate new content dynamically and even adapt to the level of difficulty.
Also, interesting thought by A16Z games, on how smart NPCs can become a bigger part of the game if they’re intelligent:
Rapid prototyping, ideation and creative boards
Generative AI tools allow game developers to quickly and easily create new game assets, characters, and environments without spending hours manually designing and building them. As a result, prototyping can be greatly accelerated, enabling developers to test out new ideas and concepts much faster. Additionally, generative AI can be used to create interactive creative boards, where game developers can quickly visualize and iterate on game ideas and concepts in a collaborative and interactive way.
Anything World is an example of platform that enables rapid prototyping for creating and consuming 3D content. A key feature of the platform is its automated rigging and animation system, which lets creators request 3D models rigged and animated. This alone enables a new wave of 3D creation, since it greatly reduces the time and effort required to create a 3D character.
What will users do with the power of creation? most likely, create! As generative AI creates exponential growth in 3D content, it becomes more important to have a management layer to properly serve this content across platforms (VR, PC, Mobile, etc), get model level analytics and reduce the size of the apps by compressing the content. Echo3D, a Remagine Ventures portfolio company, works with developers to manage their 3D assets in the cloud and make the content work smoothly across platforms (VR, PC, Mobile, etc) and get model level analytics.
Where the rubber meets the road: the challenges of Generative AI Gaming startups
While I love the innovation as a user, as an investor I have to consider wether a startup can become a large company. A lot will depend on the team’s ability to get traction with their target customer: be it gaming studios (of all sizes) or the consumer directly. And in both cases that might be a tall ask. As Andrew Chen from A16Z’s gaming fund reflected on this thread:
There are many skeptics out there when it comes to generative AI in gaming, including insiders.
For example, Julian Togelius, co-director of the NYU Game Innovation Lab, who has multiple papers on the topic of generative AI in gaming, said to Wired:
…People aren’t being realistic about what it could do. Sure, AI could design some generic weapons or write some dialog, but compared to text or image generation, level design is fiendish. You can forgive generators that produce a face with wonky ears or some lines of gibberish text. But a broken game level, no matter how magical it looks, is useless. “It is bullshit,” he says, “You need to throw it out or fix it manually.”
In a recent interview, Imre Jele, co-founder of Bossa Studio, also touched on the issues of reliability as well as copyrights/ defensibility of the IP
“However, there are still challenges ahead. For example, the legal framework for training sets is up in the air with some risk and exposure depending on the outcome of some court cases. So there are some big questions one needs to answer before developing a game using these tools.
Ultimately, even though this is a very fast developing space, we only have very few actual production ready tools that a company can depend on. So arguably, we’re still a little while from fundamental changes to how we make games.”
So in conclusion:
- Tough to penetrate gaming publishers as customers.
- Also, it’s hard to change habits in existing workflows
- IP concerns (who owns it? is it protectable?)
- Reliability – is the tech good enough for prime time?
I enjoyed following the writing of Jon Radoff on this topic, especially his five levels of generative AI for games (similar to the level of autonomous vehicles), and recommend the A16Z Game Fund One latest blog post on the UGC gaming market and democratisation of game creation.
Generative AI has the potential to revolutionise game design and development. By using generative AI to automatically generate game content and test different design variations, developers can save time and resources while also improving the quality and variety of their games. This has been demonstrated by companies like Blizzard Entertainment and Ubisoft.
While the use of generative AI in gaming raises important ethical and regulatory questions, the potential for generative AI to enhance the gaming experience is clear. As game developers continue to explore the possibilities of generative AI, we can expect to see even more innovative and exciting use cases in the future.
We look forward to continue investing in this space at Remagine Ventures and will be able to share more news about exciting deals we’re already doing in this space.