Generative AI bots have been taking the internet by storm, allowing anybody with a network connection to put in a prompt and watch a piece of software complete it. The latest trendy bot is ChatGPT, developed by OpenAI, which has generated conversations, scripts, text-based games, and even fully-fledged articles to varying degrees of success. DALL-E 2 and Midjourney are other popular programs that produce images based on text prompts, and there are countless more.
There are a lot of ethical questions surrounding the use of AI-based tools in creative work, but in video game development, they’re only getting more popular. According to George Jijiashvili, an analyst at Game Developer sibling company Omdia, AI tools will be “the hottest topic in games tech” in game development over the next year, with startups launching to fill the space. One company is modl.ai, which produces AI-assisted tools that can create bots for multiplayer games, detect bugs, and can design levels in match-3 games.
I was scheduled to speak with Christoffer Holmgård, co-founder and CEO of modl.ai, before ChatGPT, AI, and machine learning became the talk of the internet for the upteenth time. So the topic of what these types of tools can do for game developers was at the top of mind. One of the key things that Holmgård mentions about his company’s tools is that they won’t get rid of (human) work. They’ll just change what the work looks like and where developers can put their focus.
“When the digital spreadsheet came out, [it] may sound boring, but before that, you did everything by hand… So I’m sure that a lot of people thought that a lot of the kind of accountancy and computation work would go away,” he explained. “But what actually happened was we had an explosion in the industry for financial forecasting and analysis and things like that.”
“At the end of the day, I would say that I think this kind of automation probably leads to upscaling more work and may change the nature of the work, but it doesn’t necessarily lead to replacement.”
modl.ai bots aim to reduce the friction in game development
The goal of modl.ai is to “revolutionize” game development, according to its website, with developers’ perspectives at its core. Holmgård himself studied organizational psychology and statistics before going into game development at Die Gute Fabrik, an indie game studio known for the recent Mutazione, before founding modl.ai and combining his video game and psychology interests. The three tools modl.ai currently offer are seemingly unrelated to each other, but they were inspired by what the team saw as key areas that could use improvement.
“I think some of it is motivated by our own game development experience, and just thinking, ‘what are the frustrating or difficult parts?’ Or ‘where would you wish you had more time or more resources when you’re building a game?’ When we started the company and we came up with some of these ideas, we tried mapping it to, ‘what are the things that artificial intelligence could actually do for us right now?’” he said.
The team also consulted with game developers outside the company and looked at the limits of what AI and machine learning could do. That turned into tools that can replicate human behavior—to an extent.
Take, for instance, modl:play, which helps to create bots for multiplayer games. Holmgård showed off an example of the tool in motion on our call. I watched as this avatar ran into walls and obstacles in a simple dungeon-styled map in Unreal Engine. It might look clunky or useless from the outside, but the bot, which was built without machine learning, is learning the space first. It needs to learn what the game looks like before it can start interacting in a meaningful way. Once that is done, then it learns how to use weapons, to attack and defend, and other skills through machine learning. It can replicate player behavior using data from outside users, so as people play more, the bot becomes more complex.
This is a tool that developers can use during something like early access. Not only can the bots simply fill out a multiplayer space so early users can interact with them, but they can also become more advanced over time and provide data on how everybody is playing. In the example, the bot learns how to combine using a flamethrower and a melee weapon because it’s something it picked up from players.
“Normally, you would have to scrape your way or you’d have to write code, a behavior tree, or some other AI model, AI algorithm, or script to take care of that. But the benefit [here] is that you can do sort of like parts of this automatically by just observing players doing it,” he said.
Holmgård says the bot’s behavior was dictated over around 10 hours of gameplay, but that time can vary depending on the size of the game. Obviously, something advanced like Call of Duty would take longer than something a lot smaller or with fewer moving parts.
But it’s not totally automatic. You can’t just leave the bot alone and have it do all the work. modl.ai’s bot tool still allows for developer control. You can adjust skill levels, for example, or turn the proverbial nobs when adding in new mechanics to tell the bot specifically what it has to do. If you add in more data, you have to control the bot to learn that new data and apply it. This is in contrast to other tools like DeepMind in Starcraft II that take full control.
“One of the things that we learned from the multiplayer bots that we’ve been building so far is we need to have this balance, where the designer still needs to have some levers and knobs and direct values of input so that they can realize their vision. And that needs to work to live in concert with the machine learning. So you get the best of both worlds,” Holmgård said.
Doing the things developers don’t want to do
Modl:test is an AI tool that is designed to look for glitches, bugs, and performance issues that can affect a game at launch. Then, there’s level design for match-3 games, which often involves developers playing through a level 15 or so times to ensure it goes toward release. That’s where modl:create, which can help design levels for match-3 games, comes in. Both of these tools replace time-intensive but vital tasks in game development, freeing up QA testers and designers to do more meaningful work.
This was a key place where modl.ai wanted to create a tool, and it seemed obvious for what it wanted to do. It’s a vital area where people work, so an AI bot tool can fill in, replicate human behavior, and contribute to game development.
“A lot of the QA functions and capabilities are still executed by having people actually play the game for large amounts of time to different degrees of complexity. So for us, if you were able to replicate some amount of human behavior and put it into the development process, that’s a place where we could make a difference for game developers,” Holmgård said.
From there, devs can work on other tasks or learn new skills that can contribute to more areas in places like QA or level design. He calls this the “upscaling” of the work rather than the replacement of it.
He brings up the invention of open-source engine tools like Unreal Engine and Unity. Before this software hit the market, video games had to build an engine from scratch. However, with a base engine being readily available even for smaller developers, the work moves from building the engine to tweaking it to fit specific needs and different genres. The option is still there for a developer to build something from nothing as well. There are too many video games for a one-size-fits-all AI tool option to take over.
The AI bot industry will grow into 2023
The AI tool industry is bound to expand thanks to increased interest in the technology and in looking for ways to streamline work in game development. modl.ai is just one of many companies that are creating these tools specifically for developers. Holmgård refers to the industry as the first big change in how we create games since the advent of open-source game engines over a decade ago. Game development has been moving towards AI tools for a while, so it was only a matter of time.
In the future, modl.ai hopes to create more tools and expand on the ones they already offer. Holmgård notes wanting to help take the bot from one game and move it over to another; allowing devs to download an already-intelligent bot and put it into their game, along with other assets. AI tools are growing fast, so this and similar applications feel like a possibility.
But even as the industry continues to grow, Holmgård wants to maintain a balance between automatic aspects of game development and letting users take control.
“You’ll always need to put in a little work to get exactly what you want, but it’s about lifting the point of entry so that everybody comes in at a higher level,” he summarized. “Advanced AI, in particular, is expensive to build, and it’s expensive to maintain for many developers. So if you can lift the whole industry up to a place where you can get this relatively easily, and then you just do the tweaking, we think that would be super impactful.”