This article is about a search for a new model of game development. It will not deliver a model ready to go, or even a formula for finding a new one. We know the object of our search exists because progress exists, but practitioners are more likely to find it than authors. However, the previous article showed that there is a need for a new model. There are indicators that the conditions are in place for a new one to emerge, and an example is a good way to get people looking in the right direction. Often these kinds of innovations are hiding in plain sight. A natural place to start is in a place that shows potential that is currently being thought about in unproductive ways, which is why this example will be about AI.
AI can mean just about anything now and the definition here similarly broad, including generative AI and deep learning models that have shown promise in terms of prediction (in the statistical sense of fitting a set of inputs to a classification or outcome, not necessarily forecasting). This is a promising area to look because it has demonstrated ways in which it can enhance productivity, but all the attention has been on narrow applications, if not simply rejecting it outright.
Rejecting AI is the right call if it is not a good fit for a project, but the amount of moral grandstanding attached to this rejection suggests the kind of blind spot the target of our search might live in. This blind spot is the productivity equivalent of targeting a game’s outreach to other developers. There are social benefits to talking about how much AI sucks and how you should never use it, but maximizing social approval is more appropriate for a rhetorician than it is for a game developer. Such misplaced priorities recommend a closer look to see what was missed.
While hostility towards AI may be fashionable within indie social networks, the hype needs to come from somewhere and there is an enthusiastic group promoting the adoption of this technology. If you are not already hostile to AI, the most evangelical of this group will fix that.
The biggest problem with the way that incorporating AI in gaming is being discussed (for or against) right now is that it is focused on replacing parts of a system rather than proposing a new system. The most obvious example is the worry about AI replacing jobs. The worry goes something along the lines of: ‘Generative AI can create art for less than it costs to hire an artist, therefore artists will be fired.’ Not only does this communicate a lack of imagination, it tends to be presented more by artists than people who hire artists.
There are, in fact, efforts to replace jobs with AI, just not in gaming. IBM has been clear about their intentions in this regard for example. Gaming has also experienced a large number of layoffs, but none can be attributed to AI, even though there would likely be financial incentives to do so (short term increases in share price). This is likely because it is quite difficult to replace a job with AI. AI is often great at doing one specific part of a job someone is hired to do and is terrible at all the rest. If the AI can’t perform every part of the job, the employer is left with the costs of the AI and the worker.
When AI has been adopted it has often been in the more sensible case of augmenting work someone is already doing. For example, marketing can benefit from AI’s predictive strength, leaving the marketing team better equipped to make decisions. The result is a more effective marketing campaign without requiring layoffs. A lot of applications for AI work like this and likely more will emerge over time.
These applications are still limiting though. They take the existing model as given and aim to make a certain aspect of it more productive. Whether we’re trying to make environment artists irrelevant or effectively connect with potential buyers, the division of responsibilities remains intact and the process stays the same. This is like watching Unity and the App Store come into existence and only seeing it as an opportunity for existing developers to save on engine costs and distribution.
AI should not be limited to augmenting the existing model but instead provoke a reimagining of what modern game development is. Such a model will likely combine other factors, but, whatever the catalysts are, the greatest gains will come from asking more fundamental questions about what the technology enables and how development can be organized around these newfound capabilities. This work is eminently suitable for indies who should already be accustomed to arranging the factors of production in uncommon ways.
There is also some indication that there are developers thinking along these lines. Embark Studios has demonstrated its use of AI in animation with animators guiding and fine tuning the results. They’ve also automated their photogrammetry process. This has allowed them to compete with Fortnite and Apex Legends with their first game (The Finals) and with fewer resources than their competitors. It seems like a company focused on making its employees as productive as possible by questioning how much of the existing processes are actually essential.
Remember that the present focus on AI is just an example. It works well as an example because it shows that people are looking in the wrong place, but the ultimate goal is to provoke a search for a new development model, not a particular implementation (even if AI does show promise). Such a search is not easy, but it is necessary, and we should aspire to move indie developers from “are we going to make it?” to “who will spark the indie renaissance?”
SystemTalk 