AI in gaming

I don't know why I'm writing this. But here I am, writing this. And I'm writing it while sitting in my nicely secluded summer family spot, far away from the bustle and bristle of any urban landscape. My two boys (3 and 2 years of age) are watching Blippi while some jazz vibrates from an overly-bass-heavy speaker next to me. We're so remote that internet is provided through 5G on our phones rather than actual broadband. We're by the sea. It's lovely. And my laptop is decidedly offline.

I manage 3 little websites with distinct flavours of content. This one, as you probably know, is about gaming. While I have a personal site and one about the existential crisis presented by climate change. All of them are static sites that have no actual UX for me. I just have a litany of markdown files managed in three Obsidian vaults.

Why am I blathering about file management on a games website in a post with an AI-adjacent title? Because over time, I've gone from hyper futuristic tech requirements to paired-back and 'just working.' What might tempt me back to hyper futuristic stuff is said futuristic stuff helping the management, updating, versioning, plugins and other nonsense. Today I do very little of that because the stack is static and simple. And I wonder if that's the promise of AI.

Because AI, in reality, will be like a hyper intelligent, coffee-addled intern that's eager to help and smart enough to be dangerous. With a human boss, it'll be amazing. Without that, it'll just be stuffing text prompts out with lies and nonsense.

In gaming, though, I think this is still true. The industry is currently under water with spiralling budgets for games (Spider-Man 2 cost $300m to make, more than any of the franchise movies) and an increasingly wary, or just flat broke, audience, who with the prospect of €80+ per title, are just buying less and less. Or not buying at release date, but instead waiting for a sale or for enough media to suggest that it's worth the cost of entry. All while publishers like EA fill their boots with executives who clearly don't play games trying to find the next Fortnite "game as a service."

AI could step in to lower costs. Think of AI being trained on datasets by a few artists so it can replicate the style and theme of the title in order to scale the artists' work. Similarly in a spiralling game like Spider-Man or any other open-source RPG-esque title, having NPC interactions feel more rich because AI stepped in with dialogue writing, character design and choice election under the watchful eye of an actual person making sure it doesn't derail the main plot or quests.

This is a great promise for the big studios as well as indies and soloists. AI could potentially step in to reduce costs, time-to-publishing and even make a title more ambitious than first thought because scale is available. But all of this is predicated on human supervision. Not withstanding the option for AI to write some complicated code and skip a lot of the pre-debugging phase in a game engine.

I'm not as hyped up about AI as the rest of the world appears to be. It's a tool in the technological arsenal. It's also an absolute disaster for the climate, as companies like Microsoft literally drain freshwater lakes to cool data centres designed for shit LLMs. But there is promise there. I suspect we'll land somewhere good with it, but it will take time. Both for engineers to get to grips with what they're meant to do with AI, and for AI models to stop focusing on bad chatbots.

I suspect in the next few years we'll see engines like Source2 and Unreal build AI models directly into them, widening the scope of any project. When that happens, costs should stabilise, release schedules should be less crunchy, and creativity should take the fore over endless remakes designed to make a 10 year old title actually profitable.