I have a contrarian take on AI in games.

Some people who see game development from the outside have a strong stance on AI. The nuance gets thrown out of the window, and that any and all GenAI (especially for art) should be banned.

I think this is a case of throwing the baby out with the bathwater. It stems from a lack of understanding of what the game development process is like, and where true bottlenecks lie.

Game development is mostly technical tasks

The actual fun tasks like concepting, scultping, and creating FX systems is something artists ENJOY doing. However, most game development tasks over the course of a few years, end up being routine technical tasks, whether you see it or not.

A few examples:

• Retopology for character and environment assets,
• Routine cleanup processes for animations (Foot cleanup, arcs, etc),
• Configuring material layers to work correctly in engine, for assets to look more detailed and be composed well,
• Configuring physics simulations (like RBANs) for cloth and other accessory movement (like capes, reins, etc)
• Generate enough variants of audio through pitch and other attributes,
• Configuring your VFX to be scalable — hit impacts working well even for enemies of differing scale, global wind speed influencing how much foliage or other actors sway, sampling the ground when you apply a crack impact decal on the ground etc,
• Rigorously QAing gameplay code and debugging to catch bugs early in development,
• Wiring up back handlers, tabbing between panes etc with something like Common UI (in Unreal Engine), so that your UI works flawlessly across all input devices (keyboard and mouse, controllers etc),
• Profiling for performance issues regularly.

Other examples that may not directly impact the development directly, but are important:

• Being able to quickly read and summarize e-mail, (Important especially when you’re talking to so many vendors, publishers, and contractors at any given time)
• Quickly arranging documents to fill up routine forms (Game platforms like Steam, events, etc)

Game developers, especially artists in this field, need to understand that some of these are tasks that AI has gotten okay at over the years, and gives an incremental speed boost. If you hyperfixate on “AI bad overall”, you are really removing nuance.

None of the above is generative art.

A question to ask yourself, if you’re so fixated on this stance:

• Should we also throw procedural generation out of our workflows, because it is the closest thing to AI in terms of what it does (automating large biome creation, as well as rapidly create variants of small props etc)
• Is GenAI used for retopology…bad? Think carefully about this (For those who don’t know, Retopo eats up 60-70% of an artist’s time, probably more. Sculpting, concepting and putting the character in game are what artists actually find fun!)

It also turns out that companies like Capcom have mentioned something similar. Their official stance was:

We do not incorporate content generated by generative AI into our game content. However, we plan to actively utilize this technology to improve efficiency and productivity in the game development process. To that end, we are currently exploring ways to apply it across various departments, including graphics, sound, and programming.

It’s unclear how the industry will be 5 years from now, but let me leave you with a simple message:

A big AAA game costing $600m+ (like Witcher 4), $1b+ (GTA 6), or even smaller ones costing $100m+ is not sustainable if the industry has to grow. Every studio will start to play it safe - and you will not find uniqueness in games. You will see more games like Marathon and Concord.

A large portion of those costs go into the routine technical tasks I have mentioned above, and is what causes unnecessary delays and bloated teams. Games costing this much lead to centralization (i.e lack of democratization in game dev), which means game developers (including artists) don’t get to decide their future, and will continue to complain about the industry at large because they misundersand the true bottleneck.