Your Thoughts on Generative AI for Games
The Results of My Generative AI Survey from Mid-2023
Artifacts is a new YouTube series made possible thanks to the AI and Games crowdfunding community on Patreon as well as right here with paid subscriptions on Substack.
Support the show to have your name in video credits, contribute to future episode topics, watch content in early access and receive exclusive supporters-only content.
Earlier in 2023 I provided an overview of the current state of Generative AI for video games and highlighted both the potential and the pitfalls of this technology as it begins to make its way into gaming.
At the same time, I issued a call to all who read/watch AI and Games to share your thoughts on how generative AI intersects with your relationship with gaming. The survey received over 250 responses and today, I'm here to summarise the results and give my perspective on its findings.
So strap in as I share your thoughts on the matter, your excitement, your trepidation, your concerns and your anticipation on the future of generative AI for games.
The Responses
Now before we go into the results it's worth highlighting the sort of responses I'm receiving and where they're coming from. The response pool is largely reflective of my regular YouTube audience, with 90% of responses from those between the ages of 18 and 45, and 84% identifying as male. Similarly, 47% of all responses came from Europe, with a further 32% from North America. As such, it's important to take this into consideration given there can very well be perspectives and opinions that are not reflective of this particular demographic. And as such, please leave comments after watching the video if there are specific thoughts you have that you don't feel were captured by this survey.
Further to this, it's also worth breaking down the careers of our respondents and also the types of relationships they have with games. Only 22% of our respondents are those who actually work full-time in the games industry. This isn't terribly surprising, given we're asking some pertinent questions about the adoption of generative AI - and many developers will be concerned about violating an NDA. But I am nonetheless grateful for you all taking the time to fill it in, we'll dig into the results momentarily.
32% of responses came from hobby game developers: people who don't work as game developers but enjoy making games on the side. While 31% was from people who enjoy games but don't develop them in any way. Rounding this out was students at 7%, lecturers and researchers at 6% and finally content creators - the majority of which being YouTubers - taking the final 2%.
While all participants were asked about their thoughts on generative AI as a game player, each demographic had their own custom questionnaire asking them about how generative AI may intersect with their careers when it comes to games. So let's dig a little further into what each group had to say, and where possible, share some insights from these wonderful people*.
*Note: all quotes are presented anonymously
Game Developers
So let's start with the game developers. For the survey I really wanted to build a broader perspective on three key aspects:
What the sentiment is like in different games companies with regards to adoption of generative AI.
How, if at all, generative AI is being adopted in studios.
Lastly, whether it is through the use of third-party tools out there such as GPT, Dall-E, and Co-pilot.
The latter of course being a big deal given the ongoing legal issues surrounding many third party systems, be it their building of datasets without artists consent, and of course the lack of copyright that can be attributed to outputs.
The responses captured a fairly broad spectrum of the games industry. Ranging from indie studios all the way to AAA's like Blizzard, Creative Assembly, Frontier, Ubisoft, and Xbox Game Studios. Plus, unsurprising given this channels audience, the majority of submissions came from programmers alongside designers. Though it's worth stating we also had concept artists, animators, producers, and narrative designers in the mix as well.
The results presented on adoption of generative AI paint a picture of caution and enthusiasm for their use in game productions. 18% stated that they are not permitted to use generative AI at all in their games - as dictated by the heads of the studio. Conversely, 7% stated that they are told to use them by the studio. 40% of developers said they're using it as a personal choice, while the remaining 35% said they're not comfortable using them. So already, we see that despite this generative AI boom, over 50% of those participating stated that right now they're cannot or will not use generative AI in their day-to-day workflow. And a part of that is very much the legal ambiguity that surrounds it. To quote some of the written responses I received...
"We got a message explicitly forbidden the use of generative AI due to the dubious legal situation surrounding them. Since we usually do contracting work this goes doubly so."
"I am not aware of any available generative machine learning models / services / etc. that have been trained ethically and so can not feel comfortable using something created by essentially stealing other people's work without permission and reimbursement."
For those who are using generative AI, it largely fell into three main camps: concept art (32%), coding tools (23%), and a combination of NPC dialogue and story (31%) - much like what I've showcased earlier this year with the likes of Convai and Inworld's offerings.
But digging into it, it seems that while the latter is very much for the purposes of production, in the first two camps it's more for use in ideation, rather than use in the final product. As one developer stated:
"I don't really ever see anything AI generated surviving into a game wholesale, without a person touching it and changing it."
This sentiment appeared a lot in comments, with the idea to bootstrap productivity, rather than to replace people in critical roles:
"We've been deploying it mostly as a tool (like mentioned in your video) to enhance productivity and effectiveness. It doesn't appear to be something that's even capable of replacement at this time, nor is leadership seeking to do so."
Now these two comments are reflective of an overall trend I observed: if generative AI is being used in games it's often as a first-step to get something in the game, followed by iteration to improve it. After all, getting the first pass in to build from is often the trickiest part of any creative endeavour. It's only with time do we iterate, and at that point the creators imprint is so thoroughly embedded in the final product that issues of copyright are largely going to be resolved. But also, the other key aspect of this was that it's being used by developers to allow them to do their job faster. Not replace them entirely.
This was a theme that I raised in some questions: are developers concerned that the hype is making an impact at a managerial level that they would consider reducing the number of developers on staff. This is particularly relevant in 2023, in which the games industry has had a rough time with layoffs. So much of the generative AI hype has centred on the idea that AI can do all of these things better and faster than a human, but the results suggest that cooler heads are prevailing, with the idea being to find tools that speed up workflow, rather than fundamentally alter it. This is, to my mind, the inevitable future of generative AI as tools are built to support this approach, working within established development processes - rather than trying to upend them - and enable developers to build games more efficiently, rather than simply replace them.
The final outcome was one of mixed enthusiasm and trepidation, while around 25% of developers stated that generative AI won't make any impact in their job, and a further 49% felt it had the capacity to make it easier for them in the long run, there are still concerns, and much of it lay with those in charge...
"While I do believe that such tech could be utilised to improve the lives and work of many roles across most disciplines I do not trust that it will not simply be used to push for even lower salaries than are already present through devaluing labour and threatening redundancy."
This is of course a sentiment I have voiced and agreed with largely over on AI and Games Plus as I discuss the impact on the sector. But it's interesting to observe how developers worry about the expectations of senior staff being mismanaged as a result...
"Management can be overconfident about the scope of a project when generative AI is involved."
"A director needs to communicate their vision, which can lead them to using Generative AI tools to quickly create a prototype (image, text, visualization, etc) of their vision. However, they can then be so in love with that idea and that first iteration (prototype) that the additional input and changes that a team provides will be thwarted and crushed completely."
It's worth stating, as our closing fact: 40% of game developers were confident that management would make the right call with generative AI. 28% thought otherwise, and the final 32% remain unsure at this time.
Hobbyists and Students
While the adoption of generative AI among our games industry participants isn't quite as pervasive as the narrative may suggest, the difference appears very much when we look at hobbyists and students. Many hobbyists who responded work on their own, or with a handful of partners. Given the scope of work they have to complete independently, over 50% of these developers stated that they're interested in using generative AI, while another 36% stated they're interested in finding ways to use it.
The range of applications, when compared to a professional game developers, was much broader with everything from code to concept art, dialogue, textures and sprites, character models, animations, sound and music and more. The responses were also broader in how they perceived generative AI, from those who saw it useful as a tool to speed up tasks they find mundane, or simply help them get better at making games. Though there were still some that found it could only really be useful to provide placeholder assets, or even in the worst case that it would hinder their ability to get better at game development, given the generative AI would be doing the bulk of the work.
Meanwhile 59% of all students who participated stated that they've used generative AI in their university coursework, either off their own back or in the case of 17.6% of them, because their instructor told them to do so. The majority of which was the usage of GPT for essay writing or coding. While in most instances, it was to bootstrap their productivity, there were some perspective that I personally had not considered, namely the potential for it to make you more employable:
"I am waiting to see what techniques become industry standard or used in my field (tech design) before investing time into learning AI tools."
"At the very least it’s an important short term skill to have. Getting a job is tough, and being able to put “Generative AI experience” on your résumé is a buzzword that will get you through the door to start."
The Future of Generative AI for Games
The final part of the survey, was to discuss their thoughts on the future of generative AI as a game player. This segment was open to everyone, though it's worth saying that this includes the roughly 31% of participants who have no professional links to games. This was the only part of the survey that they completed.
The questions for this segment were to address two things:
First, regardless of the issues that surround it, what is the overall sentiment about generative AI as part of games as whole? Are people excited about the idea of playing games that even use generative AI at this time.
Secondly, what are the issues that people have surrounding their use in games. While this largely apes the questions raised to developers, I was conscious that a significant proportion of those participating wouldn't have any professional links within the industry, and as such do they share those perspectives?
So looking at overall excitement or interest, just over 50% of those surveyed said they were excited about the potential of generative AI for games. With the most interest focussing in areas of AI for dialogue, story generation and, interestingly, AI powered textures. But also with a lot of suggestions for other areas, such as NPC behaviours, level generation, and even finding ways to improve replayability. Whether these are actually attainable, well that's for the future to tell at this point.
Plus two related questions, was whether you would be interested in if generative AI is used in a game and whether everyone thought that this would speed up development. The result for interest in generative AI, is a fairly even split. 27% don't care, 23% are excited, and the remaining 49% are largely indifferent. Meanwhile regarding speed, 42% felt that games will be developed faster, with 13% thinking they'll take longer, and the remaining 44% feeling like nothing will change.
Reading through the written responses, there is an overall enthusiasm and excitement for how this technology will be employed:
"Many big-budget games in recent times have had to involve huge amounts of soul-crushing busy work. I am particularly interested in the improvements that could be made by being able to more easily when this busy work is outsourced to generative AI tools and artists can spend more time on things they actually care about."
"Being able to make something that looks good despite having no artistic knowledge, an actual artist will be able to find issues with my AI generated images that I wouldn't even consider, or take way less time to get something usable, but the difference between "I will never have anything" and "That's pretty good" is insane."
"As a consumer: not much. As a developer: I look forward to being able to automate the boring, mundane parts and allow focussing on the more interesting and creative aspects."
But, perhaps unsurprisingly, the issues continue to mount up, and we had significantly more people write about their issues with generative AI compared with their excitement:
"By (partially) removing the human aspect of it, if not done properly, it could potentially reduce the quality of the final product. In the end companies are all about money, so they will always try to reduce costs as much as possible, even if that means assigning less human labor than the minimum necessary."
"A huge issue already visible in generative AI projects today is that it has revealed just how much executives and managers are willing to accept low-quality work if it is done at an incredibly low cost. I think there is a huge danger in people's work being systematically devalued as generative AI tools are used to first mine their work for training and then deployed to produce low-quality copies at relatively no cost compared to human workers' wages."
"General iteration speed will allow project managers to further reduce allocated development time (This is already happening at my job). Also, generative Ai will take away most junior/intern positions, making it harder for new talent to access both AAA and Indie workforce."
Closing
Thank you to everyone who participated in the survey. Personally, the results did not surprise me all that much. It was very much in-line with my own expectations, particularly the difference between engagement with generative AI tools across game studios versus hobbyists and solo developers. As I've said in both here, and elsewhere on the internet, adoption of generative AI is going to be a slow and gradual process particularly in professional studios. But there is enthusiasm mixed up among the many legitimate concerns about its deployment, be it in overall production quality, mismanagement, and the legal and ethical concerns.
I'd be interested to see how these opinions evolve over time, and particularly as the tools improve as AI companies realise they're better off creating tools to optimise existing workflows, rather than seek to replace them (far from a unique situation in AI circles let me tell you), or developers build up the expertise and do it themselves. Not to mention with legal frameworks going to more better defined, this will also help I feel in the long run.