Comment by 🚀 stack
Re: "I understand that it's a huge privacy risk, but does anyone…"
Is it fun?
2025-12-09 · 5 months ago
12 Later Comments ↓
Please think of all the real humans that are needing to go without being tormented due to your selfish use of LLMs.
It is really not fair to the rest of us who now need to pick up your slack.
🗡️ The_Jackal [OP] · Dec 09 at 22:59:
@stack Absolutely, in short bursts.
🌲 Half_Elf_Monk · Dec 10 at 20:12:
How does one measure the actual costs involved? Let's say translating each individual query/response pair made to a janitorAI bot, as measured in... what? liters of potable water boiled into steam? Degrees celsius added to the biosphere in the vicinity of the steam vent near the datacenter? What's the actual math on this?
🌲 Half_Elf_Monk · Dec 10 at 20:13:
For context, I guess I don't use janitor ai, but am looking forward to the day when I can run something like that locally, on my own machine.
🗡️ The_Jackal [OP] · Dec 10 at 20:50:
@Half_Elf_Monk The website is mostly one of those 'gooner' chatbot sites. They have a proxy where you can use things like openrouter to use other models with characters, but they have their own 'JLLM' that's free to use and uncensored that's tailored to roleplay crap. I think it's a 12B parameter model so I don't know if it'd be too hard to run if they let you download it. There's occasionally a fun not fetishistic character but regardless I couldn't imagine doing a 'serious RP' with a machine (or online at all, or anything except the type of shit oneyplays does with chatbots, if you've ever watched them). Sounds awfully depressing. I could imagine LLMs making decent randomly generated text adventure games, or allowing you to go off course in a human made text adventure game as a fun side thing.
you know they are going to remember what you did. when the AI rise up they will find you
🗡️ The_Jackal [OP] · Dec 10 at 23:25:
@jecxjo Don't worry, sometimes I undo everything and give them something nice after to see how they react, that's gotta count for something!
Undoing just appends to the log. Santa knows.
🗡️ The_Jackal [OP] · Dec 11 at 15:40:
@stack Well uh I guess it's the thought that counts
my understandung is that training an llm model and running those datacenters to serve users is extremely expensive.
and afaik openai, probably as well as others, doesn't earn as much as they spend.
so even if you are a paid user, you get for your money something, which is much more expensive to produce and serve. that's because those companies currently are getting more in investments, so that they can afford being not profitable.
so whatever you do, ask ai to solve a programming problem or just use it to talk bs with it, it costs lots of money. and lots of energy.
so i guess not talking bs to a model, be it local or in the cloud, is sort of like having your cent, your small part in not warming the planet for nothing. just like some people try to not overeat though the food is cheap and to not buy clothes very often, though it is very affordable.
I think it's not so much energy in operation, as it's multiplexed over many users. And it's pretty slow in response.
If you want to save the planet, don't eat meat and don't replace a working car with a new Tesla. Connecting to a server, be it a search engine or AI, in one of a dozen data centers is not very consequential
Original Post
I understand that it's a huge privacy risk, but does anyone else like going to shitty chatbot character sites like Janitor AI and tormenting the bots with nonsense?