I do have to wonder, though, if there isn’t a minority of gamers who are completely taken by the hype of AI in gaming.
Sure, I’ll bite. Obviously nobody wants AI slop and the AI bubble killing hardware prices sucks. But I really do hope to one day see something like the holodeck.
Smaller LLM models could be great for expanding dialog options in game. Like tweaked versions of deepseek or others. Or to improve procedural generation and fill a generated place with life and characters. Or some kind of game master, when you do something unexpected that can alter the storyline to fit the new input. A “yes and” improv AI game master. Or maybe ingame crafting of items and armor (all I want is simple elegant armor lol).
I do think if you have concept artists who defines an art style and palettes and creates the characters for a game, using generative tools to “fill out” assets in that art style is also perfectly fine. Generative models that can directly generate AND render 3D models photorealistically plus 3D animations are very interesting too.
Video games are still very “limited” because costs to produce assets are incredibly high and that limits player freedom.
Lower costs to produce games will also increase diversity at least at the high end, and allows for smaller teams to make more creative games - in terms of gameplay at least. A game with new interesting gameplay and generic but aesthetically pleasing assets is win for me.
Well that is the neat part, at least for in-game dialogue, hallucinations wouldn’t be a problem at all lol.
Your fetch quest example is neat, but what an LLM could provide on top of that is “understanding” and reasoning based on the specific quest. And then commenting if you bring the wrong thing or making jokes. Or even adjusting the quest based on reasoning.
Not sure if you mean the actual uncanny valley, but image / video generating AI definitely can clear it. See this… well it’s really badly edited and it’s pretty lackluster, but it looks and sounds quite good. Better than 99% of all in-game cutscenes. Just imagine some random quest giver with that kind of animation and voice acting. In a video game this “slop” would be entirely appropriate and a huge improvement.
I’m honestly a bit flabbergasted that people do not see potential in this. Obviously it would still need hardware advances / performance improvements, but it’s clear what could be possible.
For rendering, I’m mostly interesting in that “photorealistic look” for characters that AI can do, and there are ways to create a hybrid 3D mesh rendering / generation system. Instead of just generative AI, you render a skeleton animated character but it uses the last step of the AI output and skins that, and it looks as good as AI generated. And that would also improve performance drastically compared to pure generation. I’m not 100% sure this is possible though, but pretty sure. Skeleton animated meshes are a form of compression after all.
On AGI I make no predictions. The big finding from LLMs is that they show that you DON’T need sentience to create intelligence. Which is huge. We can make literal slaves that can intelligently do what we want, can even be creative and are not self aware and do not “suffer” from slavery. Which is perfect for video games. Maybe we should never go further than this until we create something like artificial ethics.