AI is one of the hottest topics of 2023, and for good reason – it has changed many industries significantly in the past year alone! We recently sat down with our CEO, Dan White, and CTO, Alex Stone, to hear their unique insights on how AI will change the gaming industry, education, and learning games. In the first part of this interview, we focus on how AI will show up in games, educational and commercial, and how AI may evolve into more efficient tools for game developers and designers to make games.
via Giphy
Join us as we explore how AI technology can effectively mirror professional practices within a game environment, the incredible benefits it offers to players seeking enriching, immersive learning experiences, and so much more! Take a second to give us a follow on Facebook and Twitter so you never miss an eye-opening interview like this one.
This interview has been edited for length and clarity.
How do you see AI impacting the landscape of learning game design?
Alex Stone: One of the main impacts will be how we give players feedback. Right now, all of our feedback systems are super heavily designed and specific to each learning outcome. I think with AI, specifically generative text AI and multimodal deep AI, they’re going to be able to give so much more contextual feedback to every possible thing a player might do in the game. By default, AI will offer more flexibility than what we might be able to craft.
Of course, it’s not just about offloading game design to AI – instead of designing something and then testing it, we could say, “all right, we turned on the feedback engine we use for every game, we play-tested it, and we’ve identified the refinements we need to make for this game.” In this way, AI may be able to flip the design process. We’d be able to test with an AI system and then refine it as opposed to designing a really refined system, testing it, and then identifying additional tweaks. That’s how I anticipate AI will change game design. We’re already seeing this slowly filter into that tech space. I think next year we will see this more often in the gaming industry.
Dan White: That’s really important, too, because developing tutorials for games takes a large number of development cycles. They’re typically a significant part of the budget. If that can be, in part, offloaded to AI, that’s a significant pickup.
Additionally, whenever we’re using learning games as an intervention, we have to teach the game before the game can teach the player. To the extent that the AI can accelerate the process of teaching the game, that means more time for the game to do what the game is designed to do, to have the educational impact that it’s meant to have, which should translate to higher efficacy for game-based learning interventions.
You can look at it like a balance sheet. In the “liabilities” column you have the time that it takes to train the player how to play the game and in the “assets” column, you have the impact of the game. The net impact should be more positive if training the player to play the game itself is less of a liability because the AI makes that training more efficient and more effective.
AS: Opening up a whole other avenue, I also think AI will change how we communicate about game design, both internally and externally. This communication can be massively boosted by having big text to visual generation, and even just text assistance. All of the writing tools that help the general public also help designers communicate and potentially move faster, allowing them to write more ideas and try more ideas because there are just more aids to their brain available to them. That’s not unique to game design but it’s already impacted all game designers in the world.
DW: File this under a huge question mark, but what happens if you ask an AI to design game mechanics or game systems for you? I imagine they’re bad at designing something fun but they’re probably really good at taking an existing system that has been designed by a human and helping balance things out. A lot of the front end of a designer’s job involves a lot of creative thinking about gameplay systems and then the back end is a lot of time in spreadsheets and tweaking and fine-tuning and I imagine an AI could potentially improve both the efficiency and the quality of that back end.
AS: Right now in Google Sheets, there’s AI that can run in the background to automatically try to run statistics on your sheets for you to gain surface insights, but there are no collaborative tools to build out from. Rules like that are not in code. There are lots of tools that will collaboratively write code with you that our programmers are using already. But in terms of people that are outside of programming, the tools are not there yet. There are no integrated workbenches with AI unless you’re a data scientist and you already have a bunch of advanced math statistics skills.
DW: Tell me if this is practical now, or potentially in the future, Stone. Where my head is going is toward rapid prototyping of game design ideas, and if AI can increase the rapidity with which rapid prototyping can happen, that would allow designers to flesh out more ideas more quickly.
AS: What’s the prototype artifact? Is it a functioning game? Or is it a video of a game? Is it screenshots and wireframes of a game, or is it GDD descriptions of a game?
DW: It could be any of the above, but which one of those is the closest to becoming a reality?
AS: Right now, AI can help designers flesh out the written description for a prototype, such as written distributions of gameplay and game mechanics. Designers are already using it for that today. Soon, we’ll be able to have it generate still images, wireframes, and user interfaces in a representation of gameplay mechanics.
A little bit further out, but very still in the near-term horizon is having AI generate video footage of the finished game described by that design. Then much further out is a functioning game based on the description of the design. That’s probably still years away. But you can watch Twitch footage of someone playing the game that you wrote a minute ago – that’s probably gonna happen next year.
DW: So it sounds like there’s at least something to be said about co-design.
AS: But not co-development. I think AI for this purpose is going to go in two different directions. The generative text AIs are gonna add more and more media, and they’re going to get more and more fluid, so designers are gonna get closer to the code that way. And then Unity and other game engines are going to add more and more tools to help programmers and artists generate content in the engine. Someday those will merge. But who knows when that could be, it might still be five, eight, ten years out.
I also think that as a junior Unity developer, you’ll be able to go to Unity and make a more interesting game by yourself. Then again, everyone will be able to do that, so it doesn’t actually give you a competitive advantage in any way. But I do think that it’s possible, one day, that someone could say, Hey, I made this in Unity yesterday, and wow, it’s a finished, playable game with nice art.
In terms of tutorials, tutors, and guides, how can AI revolutionize the way players learn and progress in games?
AS: There are two aspects to it. There’s the actual feedback the game gives the player. The goal of learning games is to give feedback that keeps players on the challenge curve outside of being frustrated, but still learning. If a game can do that, and can also tailor the experience to the learner’s specific needs, that’s the dream.
That’s the dream of all differentiated instruction and adaptive learning models – to be available to everyone, and not locked behind companies that have huge teams or departments that study data from students and refine their adaptive models. Instead, the dream is that every piece of software will be adaptive.
But the question that isn’t quite answerable yet is “How do we embody that?” How does the player interact with that? Because right now, you can put a chatbot in a game that gives you really useful tips. But how does it show you feedback instead of telling you? How do we have embodied AI agents that are NPCs or embodied in the game in general?
In the future, characters that are using generative AI could give you a completely unique game experience that no other player is going to have. I think AI will affect all entertainment games in the future, and probably every entertainment game that’s starting to be developed now. I could see one of their headlining features being generative AI characters. That’s all probably super secret, under-wraps type stuff.
For us, we will definitely hope to look at the options for generative AI and use it to show players what to do in the game or lead them along the path of success. Microsoft just released a tool called Type Chat which solves a problem for us as developers: we want to use a large language model that’s trained on text, and we want it to give us data that the game can then render. Typically, that’s been a problem because the programming interface is just about sending a text prompt and then receiving a text response – but coders use structure and data models for things. Type Chat is going to allow us to say “Here’s a game. Here’s how the data works for this game that we’re already rendering. Now you, large language model, generate it to these specifications,” and we can just start plugging stuff in. I could go back into our catalog of games and just start plugging Chat GPT into them. That’s where we’re starting to get to, we could use ChatGPT to start offloading certain game mechanics.
It’s going to be just like any other system, and then we can start integrating things like Type Chat into the Scripting API, and we’ll be able to call up this function in our code and not even think about the fact that ChatGPT is out there on the server somewhere. I think that that will make it really practical for a lot of people to be using AI throughout the process of creating a game.
DW: The powerful thing about tutors is that, in theory, a good tutor keeps the player or the learner in the zone of proximal development. So the student with that tutor scaffold can be taking on more challenging content than they otherwise wouldn’t be able to in the absence of that intelligent tutor. That’s a really big deal. It’s a big part of the reason why even the best games, in my view, should be paired with a good instructor in an ideal scenario.
With good AI tutors embedded in games, more advanced and complex learning games should be manageable for a broader GenEd audience than would otherwise be impossible. Students, in theory, should be able to grapple with more complex game systems and game topics than they would be able to without that tutor keeping them in the zone of proximal development. They’re learning progression should be more efficient and effective because of that tutor.
We’ve said before that games enable teachers to be less of the “sage on the stage” and more of “the guide on the side,” and the same applies to AI. AI in learning games will increase the ability of teachers to provide students with in-depth learning experiences and personalized instruction.
Even the best teacher is one teacher serving a classroom of 20 to 30 students. AI tutors in games can make it so that a tutor-based scaffold can be available to every student even when the teacher is not able to interact with every student at the level they need. That will make it so the in-person teacher can really focus their time and energy on one student or one cohort of students at a time, thus speaking to their individual needs in a way that an AI probably won’t be able to any time soon.
Another thing is the power of AI from an equity perspective. AI in learning games means that people who are playing learning games in informal settings or at home where there is no instructor (and maybe there’s a parent but the parent is disengaged, or the parent doesn’t have subject matter expertise) now have access to more effective learning solution because of that AI tutor, and they can be learning on their own time at home, or in an after-school program, or any other situation where they don’t have an instructor to help scaffold their learning. That’s really great from an equity of access perspective.
How can AI assist in mirroring professional practices within a game environment? What benefits does this offer to players?
AS: Imagine if we could take Contents Under Pressure, the chemical safety game we created with Rowan University, and make it come alive with an AI that was three chemical engineers bantering in their lab, and the number of real reactions they could have if something goes wrong. That would potentially change the immersion factor of the epistemic-type experiences we make.
I also think of “Shirt Mart” from MSI Retail Sim. Instead of having a couple of prescripted customers, we could generate an endless stream of AI customers with a different mix of concerns, needs, demands, attitudes, and moods. The depth of a simulation can become much more robust when you’re bringing in that level of agency.
DW: Yeah, for sure. Certainly, the aspects of most if not all professional practices that involve interacting with and negotiating with other human beings will be vastly improved. I would imagine that it’s the same or very similar to the same technology that we were talking about earlier in regard to commercial games that will show up in NPCs for every RPG that’s being developed in the future.
As AI becomes more sophisticated and integrated into gaming experiences, the landscape of learning is set to change fundamentally. Our CEO and CTO have more thoughts on AI though, so be sure to keep an eye out for part 2 of How AI Changes Learning, where we’ll discuss how AI will affect future-facing skills and the types of learning outcomes emphasized in education.
Want to contribute to the future of learning games with lasting impact? We’re educational game developers with 18 years of industry experience, and we know how to create playful experiences that improve people’s lives. Get in touch for a free consultation! Let’s work together to change education for the better.
More industry insights: