51–60 of 90 posts
(iOS - https://apps.apple.com/us/app/ai-game-master-dungeon-rpg/id6...) (Android - https://play.google.com/store/apps/details?id=com.aisuccess....)
This is roughly my 5th startup, going through Computer Vision -> Special Needs -> Cannatech -> Hacking RNA -> AI -> Gaming with AI
AMA :)
> The video game audience has decided, almost unanimously, that they don’t want generative AI in their games. Nah, it's just a very vocal minority making lots of noise. We're slowly starting to see saner minds prevail, with steam backtracking on the ban of aigen and so on. People don't care if the end result is good. And the "but muh art" is just virtue signalling. I've used this example before, for me "art" is the f…
It's always this massive cop out. "Things just got started. The future will be amazing". I heard this with IoT, Wearables, Cryptocurrency, VR, now Gen AI. Forgive me if after a while every vague gesture that "now it is for real" just becomes noise.
No no. The present is already amazing. If you don't see it yet it's on you. I've been having the most fun I've had in 25 years of doing this thing.
Earlier quoted context omitted.
It's always this massive cop out. "Things just got started. The future will be amazing". I heard this with IoT, Wearables, Cryptocurrency, VR, now Gen AI. Forgive me if after a while every vague gesture that "now it is for real" just becomes noise.
> "Things just got stated. The future will be amazing". No no. The present is already amazing. If you don't see it yet it's on you. I've been having the most fun I've had in 25 years of doing this thing.
Your reply does not inspire much confidence. A personal anecdote that you love something is exactly the sort of response people got during the cryptocurrency hype cycle.
Earlier quoted context omitted.
I think there's a certain antipathy between "hustle culture" and gaming https://components.news/the-gamer-and-the-nihilist/ that is is, people who are caught in AI FOMO are performatively trying to appear to be productive and that's the opposite of fun.
Anyone who has worked on a game knows it is a long, painful slog to the finish line. AI dev is promising the exact opposite: minimal prompts and the agent does all the slog. Even if AI can whip up a quick demo or prototype for a game, it is the long-tail of tedious details that a passionate person has to hammer away on that separates what ships from what dies. I'm guessing most AI opportunists are looking for quick w…
> The video game audience has decided, almost unanimously, that they don’t want generative AI in their games. Nah, it's just a very vocal minority making lots of noise. We're slowly starting to see saner minds prevail, with steam backtracking on the ban of aigen and so on. People don't care if the end result is good. And the "but muh art" is just virtue signalling. I've used this example before, for me "art" is the f…
It's always this massive cop out. "Things just got started. The future will be amazing". I heard this with IoT, Wearables, Cryptocurrency, VR, now Gen AI. Forgive me if after a while every vague gesture that "now it is for real" just becomes noise.
Just yesterday i ran Devstral Small 2 under koboldcpp with a MCP server that exposes a few simple set-file-contents, get-file-contents, show-file-tree commands and i asked it "fix the formatting in foo1.h using foo2.h as a reference and add doxygen comments using the documentation from API.md" and it did it, all running on my own old PC with everything fitting in a 4 euro USB stick.
To me this is amazing. Sure, there are a bunch of issues, but it far from anything reated to IoT/Wearables/Crypto/etc (VR despite being a niche has a lot of fans so i'm not sacking it with the rest). I couldn't even see myself caring about any of those, but i could see several ways how i could use AI/LLMs before i was even able to run one (though admittedly, not all things panned out as i expected :-P).
We are still early in the tech. 2021 LLMs were not fit for any real purpose outside zany Madlibs. The tech has only really gotten fit for purpose since GPT4 and coherent models only got cheap and fast enough since llama 2 (July 2023). Even then smart models only got affordable around Deepseek v2 (May 2024) which was the first gpt4 level model to consistently serve subdollar per million tokens. You need at least a GPT4 level model to make a really interesting game. That's like a blink of time in game making time, indie games have 2-3 year cycles at best and AAA are like 5+. Even now fast, tool native interaction is only just coming online and there are no cheap models for that. Really fun game playing AI needs something like that to be magical.
There are AI games they just don't look like AI games neal.fun's Infinite Craft made a boat load of money and was very popular and it was powered by a Llama backend. Character ai and it's dozen copycats are like online storytelling/roleplay things. These are fairly popular. That could be a game or like a game platform or maybe it's more like fanfic. It has no fail state but people have made variants with fairly complex rules and states. Skryim and a few other games have pretty popular AI mods that let you add in NPC interactions that can talk, see, and even change their interactions with AI. The Skryim one Mantella has more than ~100k downloads.
We don't really have a northstar yet. Making a good game is really hard and usually takes someone doing something weird and clever. Minecraft is obvious in hindsight but many games on the way failed to make 'legos on the computer' fun. Incorporating AI breaks all the 'rules' of gaming. It operates slowly, it has high potential to break the rules and it has nearly unrecoverable failure states. Indie devs haven't figured out the best way to handle this and I suspect most big companies have just washed their hands of the whole thing until the AI gets faster, cheaper, and more stable.
I did a conferencr talk maybe three years ago on using AI for games. Since the talk, code, audio,and imagery have progressed leaps and bounds, almost all of the points on active AI use haven't changed much. You have more capable smaller models now that means you could at least run local models in games, but you still don't get to see what it is going to do before the user does. Developers are acutely aware of the cli…
I built a little mafia game a few years ago, but it was never very popular. http://gptmafia.io/
and someone i know made one too: https://talktomehuman.com
maybe the games market is just hard in general, but i dont know that ai games are super popoular even though people are building them. playing mafia against the ai feels somehow tiring, even if it is also occasionally fun
Having recently considered playing Dwarf Fortress(and quickly rejecting it for UI issues and overall tediousness), it seemed like a game that would benefit from an LLM backend. I'm curious what sorts of world history an LLM could come up with. It likely would be a lot less efficient than the current system, but interesting nonetheless. I think it would be easy for devs/modders to add a way to supply an API key to an…