AI and rapid prototyping: the future of game design?

My wife came running into my office the other day: “OMG! You’ve got to see this!”. I follow her to a monitor where she shows me a demo video of the Pixicade app. There it is! Kids creating video games from drawings. Back in the eighties, I initially drew my first Yar on some graph paper, but it took weeks of work to see it flying around on a display. Yet here are kids at home, drawing their ideas, capturing them and then playing them… at least on a simple level. Impressive. I recognised this as a rapid prototyping tool, and it brought to mind the first time I heard about this in video games. It was at Atari.

When I share the development tools we used when video games were still in their infancy, modern programmers laugh. They laugh at the absurdity of life without decent dev tools, but they seem nervous too. Are they wondering if their work rises to the level of their environment?

Many was the day I’d wake up with a grand idea, only to see it wither when faced with the threshold of work, expense, and time required to ’try it out’. Vision is a fun fancy, but when the total cost for testing something is high, experimentation is reserved for those resourced enough (or crazy enough) to invest in a ’possibility’. Is commitment to vision enough? Neurotic devotion? Either way, the best path to expanding the horizon of possibilities is to lower the cost of the experiment. That’s what rapid prototyping is – an effort to lower the cost of realising and evaluating game concepts.


AI-made games don’t exist yet, but AI-rendered paintings do. This one sold for nearly half a million dollars at auction in 2018

Today, Unreal and Unity join a variety of tools available for rapid prototyping. Unreal is a great name for the engine, because tools like this were unthinkable back in the day… which is not to say they weren’t thought of, or more accurately, ‘wished for’. And like any entertainment company, there were people at Atari focused on making wishes come true.

Steve Wright, who was head of VCS development at the time, launched just such a project. His idea: instead of making a game, let’s make a system in which we can quickly approximate various gameplays to test them out. I’d like to say this was the birth of rapid prototyping, but it was merely the moment of insemination. This pregnancy would stretch out over many years before any viable birth would occur. 

The ability to try new game features, mechanics, and configurations to see if they feel worth pursuing was always the dream. But we couldn’t realistically imagine a system (or even an interface) that could get us there at the time. Game dev tech has come a long way. 


Howard Scott Warshaw at his Atari workstation. Rapid prototyping and AI-driven game design were mere dreams at this point

We’ve reached the point where AI and machine learning algorithms are simulating players to test (and sometimes predict) new game features. Non-human systems are assessing human engagement and entertainment value. It’s an interesting place to be. Touring through the Turing test, and unable to tell if our guide has gooey guts or gears? Where does this go?

Ultimately, the computer could map my brain and use that info to instantly generate a game ideally suited to my tastes and perfectly tuned to my capabilities. How fun would that be? Maybe not much. If we’re all playing unique games contoured to us, the competitive angle’s lost. How can we compete if we’re all playing different games? And isn’t competition one of the major motivators for playing in the first place? And what about being in first place? 

To support competitive gaming, we need a discrete number of different games, or at least genres. Then the brain map doesn’t have to go all the way to creating a special unique experience chosen just for me, it simply selects the best match from a limited pool of alternatives and away I go. Then we would have narrow casting. The downside is that new possibilities are limited. The upside is that we don’t have to wait for it since it already exists. The future is now! 

These algorithms are mainly about adding bonus elements or monetisation. But could it create new gameplays? Would it create new genres? And if it did, would these new elements be tuned to human satisfaction or algorithmic affirmation, which may or may not be the same thing? I believe AI neither takes over humanity, nor controls it. But machine learning algorithms could ultimately realise that humans are not worth accommodating and prefer machine entertainment over human satisfaction. What then? We spend our lives chasing the dream and it turns out to be the nightmare? Is our entire existence merely a Twilight Zone episode?

Perhaps. But let’s move back to the past we’ve already escaped. Back then, rapid prototyping was vapourware. Often at Atari, we’d wish for some magical synthesising ability to manifest our concepts with less sweat. That’s what people do on breaks. But some people seem to remain on break. They blame a limited dev environment for rendering their amazing ideas unreachable, thereby justifying both their ’genius’ and their inaction. That’s when Jerome, my dearly departed designer and friend, liked to quote George Herbert: “It’s a poor craftsman that blames his tools.”

There are two kinds of visionaries: some adapt their vision to the tools at hand. Others choose to wait around ’until the tools catch up’. At Atari, the best game makers did what they could (and sometimes more) with the tools available. That’s what you do when you can’t wait to create. You do what you can with what you’ve got. This way you don’t seem like such a tool.

Leave a Reply

Your email address will not be published. Required fields are marked *

More like this