Artificial Intelligence in Open Source Game Development

Gaming
1

:bookmark: ITT we discuss the ethics and applications of artificial intelligence in open source game development.

While researching #gaming:red-eclipse, I came across an interesting discussion on their forum regarding the use of AI generated content in the game. As the author mentioned, this is a polarizing topic that sparks heated debates; however, the technology could be of tremendous value to projects that depend on many hours of volunteer work. [1] As a server administrator for open source games, I am interested in using AI generated content to offer custom experiences through mods and integrations. In many ways, the potential benefits and ethical concerns raised in this discussion mirror both my own and those of society at large.

Potential

Creating a game with high quality scripting, artwork, and content requires a diversity of skills that often necessitates a team of people putting many hours into specialized roles. Employing AI models could allow an individual game developer to not only access skills they otherwise wouldn’t have, but also do so in a quantity that would be impractical even for large teams. This potential is amplified further in games with active modding communities, such as #gaming:luanti. Beyond asset generation, artificial intelligence could be used to create custom in-game experiences in real-time through, for example, NPC dialogue, questlines, and lore. Rather than expounding upon this list here, I open this discussion to the thread to share additional examples of AI content in gaming.

Ethics

Artist compensation was the main ethical concern raised in the original Red Eclipse discussion and its citations. [2] Absent AI, artists would be paid hourly or by commission to create the necessary assets for the game. The necessity for iterative development in a software environment can make this a fairly lucrative position for artists, though it has also been reported to lead to strenuous relationships between engineers and artists. [3] AI poses a threat to artists in this relationship in that it is both cheaper and less temperamental. Many people find replacing artists with artificial intelligence in large corporations as a cost cutting measure objectionable. I don’t think this perspective is wrong - just futile. I also share the sentiment that artists will still be needed in the creative process, as outlined by Udonis:

However, rather than replacing artists entirely, AI is more likely to shift their role toward curation, refinement, and creative direction rather than raw asset creation. [4]

When animated cartoons transitioned from movie theaters to television, production companies began using artists only for the major scenes of a cartoon while outsourcing the in-between shots to cheaper third-party companies. Corporations exist to make money. If they can find a way to cut costs, they will.

FOSS

The FOSS community is generally regarded as being committed to philosophy and ethics, [5] making it simultaneously more leverageable and less impactful in this debate. The community would be more inclined to take action based on an ethical argument; however, these projects are often dependent on volunteers and are maintained for a communal good rather than a profit motive. It seems counterproductive to impose further limitations on those who are already committed to good ethics. In so far as open source projects wish to apply the same philosophical principles to their use of artificial intelligence, they can simply use open source AI models, as discussed below; however, it seems that only a minority of players are particularly concerned with AI in gaming, as reported by MIDiA:

The prevailing narrative is that players are also staunchly against gen AI in their games. While some are, MIDiA data shows that this is likely the vocal minority. The majority of gamers – even daily players – are neutral about developers using gen AI, with the tech rarely impacting purchasing decisions. [6]

It would be interesting to see the survey results for gamers who also use FOSS - perhaps they are that vocal minority.

Open Source AI

If open source game developers want to make use of AI generated content while still upholding the ethos of free software, one option could be to use open source AI models, such as the ones released by Mistral AI. [7] Ideally, these models will have been trained on content that is either already in the public domain or has been obtained with permission and compensation to the creators of the original works. At the same time, open source projects are already disadvantaged in regard to budgeting and resources compared to those backed by corporate incentives; further, the end product of their efforts are freely available to the benefit of all. Imposing further limitations on open source projects, in my opinion, seems to be missing a bigger point. Given that corporations are already using proprietary models trained on copyrighted materials to eliminate workers and increase margins, there is an argument to be made that FOSS projects should be even more willing to draw from the complete library of humanity’s creative output than their profit-seeking peers.

  1. ↩︎

  2. https://www.reddit.com/r/gamedev/comments/16dh6re/use_of_generative_ai_in_games_backlash_in_code_vs/ ↩︎

  3. ↩︎

  4. ↩︎

  5. https://www.freeopensourcesoftware.org/index.php/FOSS_Philosophy ↩︎

  6. ↩︎

  7. ↩︎

2 Likes
2

It is interesting to hear from this side of the debate, as ultimately, I had to abandon any notion of using generative AI due to the fact I would likely lose most of the contributors to the game. I think this highlights the distressing position open source developers are put in given this situation.

Red Eclipse is a community first, and a game project second, so I had to protect that.

1 Like
3

It’s hard for me to fathom the scope of the decisions required to reach the development “finish line” (while also being aware that the finish line tends to move as a function of time).

We all have people and things we value. When the water rises too fast you protect the people first and worry about the things later.

And good grief, the AI advancements are really making the water rise fast. :man_rowing_boat:

4

That is disappointing to hear, and from an external perspective it sounds like the idea was shot down before it even got off the ground. The Turing Test turned out to be useless in determining consciousness in machines, but it is true to say that we often cannot distinguish between human and AI content anyway. Maybe a rogue developer introducing some AI content through an optional modpack would change some minds…

That is a good ethos for an open source project. If the users don’t want it, they don’t want it. The game really is excellent already, so I wouldn’t say that generative AI is necessary on this (or really any) project. It is a pretty interesting case study, though!

5

It’s a delicate balance, for sure.

I actually hired a studio using all of our available funds on OpenCollective to create a human produced version of the song and we’re about to receive the first draft. They’re going to handle distribution of the song in a wider sense (where the project will end up receiving royalties) as they believe it’s a hit that could give them some much needed exposure.

In a way, this seems like a happy ending despite the speed bump, but it is worth noting that I tried to use Upwork and they terminated our original contract arbitrarily because I was discussing with the freelancer how we would go about paying them from OpenCollective. They then tried to keep all the money but I threatened legal action.

Nothing is ever easy, especially in game development, but I’m hoping to elevate Red Eclipse above what people have come to expect from an open source game.

1 Like