The developer behind Teach My Little Sister How to Drive, a driving instruction sim where you instruct a generative-AI-powered, dynamic NPC, has been spending $1,000 dollars a day to keep the demo working.
Given that people playing this have GPUs, why not run the AI model locally? I can’t imagine it needs to be a very large model.
AI running a video game, somehow still less power efficient than a playstation 1. Good job guys. Just make an actual video game.
You could literally spend $0 and an afternoon hooking up a 4gb model to the game and having it run locally instead. Or you could also spend $0 and code a decision tree that would be fun for players to experience. Or you could choose the dumbest possible option and bankrupt yourself for no reason
Number 3 please!
deleted by creator
Why are there so many opportunities to post this meme?
Boooy is this “meme” exposing your well deserved insecurities.
There will only be more as time passes.
The only model I program with is a Model M.
Well maybe its a Mini M, but close enough XP
No you dont.
lol not reading that
me explaining to future generations why all the water dried up
“Markets can remain irrational longer than you can remain solvent.”
One of the important things to understand is that certain elements of that irrational market are very, very rational, as long as the money comes in.
It’s what happens when the chumps you’re grifting either smarten up or go broke that things get bumpy.
How can this ever be sustainable without a huge monthly subscription?
Easy, you just hyperinflate the USD!
Infinite money glitch, irl.
Ask Anthropic! I hear they worked out a solution.
The solution being a huge monthly subscription. Or massive token rates if you’re a big customer
They keep borrowing and giving out ‘free’ tokens until people are hooked.
Then gradually introduce different payment methods until they figure out what works best.
Then prices rise (I’m guessing, rather rapidly) to pay off all of the loans and to start making profits.
In the meantime, their machines are learning from the users, and learning which areas and how it can be most profitable.
Boom. They’ve created a profitable business that nobody wanted, nobody asked for and nobody needs. The interactors don’t even realise that they’re not even really using it, they’re training it.
DON’T USE AI.
YOU’RE CREATING MORE PROBLEMS. MORE BEZOS’ AND MORE MUSKS.
Why can’t they just fine tune a small locally hosted model? Are they dumb?
Most gamers have a GPU anyways
Not everyone has a GPU with enough RAM. I have a 3060 with 12GB, and I cannot run ollama and a 3D game at the same time. As soon as ollama starts processing, the GPU is at 100% and game and ollama start swapping data in and out of the GPU RAM, which means the whole computer becomes unresponsive for a long time. I had to set up a script that disables ollama if steam is running.
12GB is a lot of ram. By small local models I am talking about fine tuning something like < 2GB LLM
Sorry, all of your what?
- ollama is bad at everything they do. 2. They can run a really small llm, as the requierements are pretty limited.
Exactly my thought. Surely you could and achieve at least a similar result.
I would say yes, yes they are dumb. Taking out a loan for a game that isn’t even out yet according to the article to keep an ai going is pretty stupid. Even if the game does well, the overhead is massive for what could be done with a smaller specialized model. Hell, I bet payig someone to train an existing model would be significantly less than a $1000 a day when you aren’t even making money yet.
Because this is vibe-coded. You think an LLM is going to give up how to make your own?
I dislike AI slop as much as the next guy but that’s a stupid take
yes. it’s not self aware.
I literally let deepseek + opencode modify the nanogpt code for my own custom model and suggest training datasets lol
Ran out of patience after a while because crappy GPU, but it was training and words were being worded when I prompted it after the first day of training
It will though
deleted by creator
So they didn’t even do the math for a basic cost forecast before launching the demo. Also this shit makes them vulnerable for bad actors. Just put this game on several PCs and run it 24/7 to bankrupt them.
Another engineer I work with had a really well put description of proper AI implementation that really applies to this:
“AI should help you build the automation, AI should not be the automation.”
AI should help, not do.
AI: only do, not help
How long before the internet is flooded with clips of the little sister saying all kinds of nasty stuff after people prompt inject the AI?
Hilarious article to put out tbh. If the game wasn’t broken before, it’s gonna be now.
Good. Go bankrupt asshole.
AI powered games don’t work.
We did a super cool project for a game jam last year. Cost like 50c per game for us.
Cool for the 100 or so players who tried it, but there’s no way to make a profit from it.
Local models maybe.
For just simple and guided actions and dialogue, local would be fine for this. Or better, have AI design the large database of what to say or do when a player does things. No need to reinvent what players are going to duplicate. The token cost was worse when you consider all the same stuff being prompted over and over.
If they’re actually hitting a cloud LLM, there’s no good reason not to wrap that and memoize the results… it would probably be hard to generalize the inputs (it could just be the whole game state for all we know 🤷♂️) but youd build up that database of all the most common input/outputs in real-time
Exactly, that’s just smart programming (maybe that’s one of the issues… but I can’t say that without evidence).
I saw someone years ago develop that concept for a text to speech idea for NPC responses, and for the same reasons. Why repay for something when you can just capture the sound file for each new phrase? The cost is in the creation and addition of things, not in the actual running of every person’s game (or app in this case).
Exactly, that’s just smart programming (maybe that’s one of the issues… but I can’t say that without evidence).
That would require you to actually understand what your code is doing…
In case it might be interesting info, maybe want to check out WebGPU, WebLLM, Wllama, or Transformers.js. You can write web apps that run an llm on the client’s browser.
It requires permission and compatibility and isn’t… it might be tough to really make accessible to the average user.
But it’s an option and kinda cool.
And they say there’s a bubble, pffffffffffffffffft
