Even most AI hate us!

[NeverLoseKibou]

Can we hook one of them up to our website?

If you have a server to run it through and sufficient local computing power (a good enough GPU and RAM), I believe it can be done.

I've never had good luck with locally ran ai personally. I feel like commercial ai pulls from the web at times which makes it have nearly unlimited resources to pull from where as locally ran ai often times hallucinates or comes up with random jargon. I mean my systems not crazy or anything but I do have a semi decent amount of ram and a fully modern gaming GPU, but responses of locally installed AI has just been lackluster.

Well locally run AI can reach out to the web as well if you give it permission. When I run AIs locally I will often give them permission to use the web, and use skills I've built to direct them to map resources online as well as local resources. I would advise running a VPN because you don't have direct control of it's online investigations. Don't get me wrong, they don't seek illegal content any more than commercial AIs do but even just trawling map wiki's and means your local machine is off scouring map resources to answer your prompts.

I've had a lot of success with local AIs, for map activism research.

There is an unrestricted AI on the darkweb that people are talking about. I did try it for some activism queries to see how it compares and it was pretty good, comparable to other local models. They're doing what was proposed above, just running a local AI with a web interface.

The trouble is you can't control what people are using it for and so I don't know what the legality would be of someone using it for legally questionable purposes. I mean if someone posted a question in this forum about something illegal then a moderator would see and remove it, but they could ask the same question to an AI and get an immediate response. That's why the commercial AIs have guard rails. You could try to restrict it with rules but then that becomes a slippery slope of restrictions.

What would be safer I think would either be a simple guide to running an AI locally including some useful map friendly skills to add to the AI, or a Docker image that's ready configured. But you still need a decent machine to run it, AIs are very memory and GPU hungry.

[Scorchingwilde]

If you have a server to run it through and sufficient local computing power (a good enough GPU and RAM), I believe it can be done.

I've never had good luck with locally ran ai personally. I feel like commercial ai pulls from the web at times which makes it have nearly unlimited resources to pull from where as locally ran ai often times hallucinates or comes up with random jargon. I mean my systems not crazy or anything but I do have a semi decent amount of ram and a fully modern gaming GPU, but responses of locally installed AI has just been lackluster.

Well locally run AI can reach out to the web as well if you give it permission. When I run AIs locally I will often give them permission to use the web, and use skills I've built to direct them to map resources online as well as local resources. I would advise running a VPN because you don't have direct control of it's online investigations. Don't get me wrong, they don't seek illegal content any more than commercial AIs do but even just trawling map wiki's and means your local machine is off scouring map resources to answer your prompts.

I've had a lot of success with local AIs, for map activism research.

There is an unrestricted AI on the darkweb that people are talking about. I did try it for some activism queries to see how it compares and it was pretty good, comparable to other local models. They're doing what was proposed above, just running a local AI with a web interface.

The trouble is you can't control what people are using it for and so I don't know what the legality would be of someone using it for legally questionable purposes. I mean if someone posted a question in this forum about something illegal then a moderator would see and remove it, but they could ask the same question to an AI and get an immediate response. That's why the commercial AIs have guard rails. You could try to restrict it with rules but then that becomes a slippery slope of restrictions.

What would be safer I think would either be a simple guide to running an AI locally including some useful map friendly skills to add to the AI, or a Docker image that's ready configured. But you still need a decent machine to run it, AIs are very memory and GPU hungry.

I think it'd be relatively simple to make a helpful FAQ chatbot connected to a copy of data from the yesmap wiki, something like customer support bots on websites that sell products that direct users to the right page and answer basic questions based on a mixture of actual programming and AI to make sense of what the person is trying to find. Some people find it easier to get information this way than searching keywords or navigating topics on a wiki, and it wouldn't be as big of a deal to run or set guardrails for.

[NeverLoseKibou]

I've never had good luck with locally ran ai personally. I feel like commercial ai pulls from the web at times which makes it have nearly unlimited resources to pull from where as locally ran ai often times hallucinates or comes up with random jargon. I mean my systems not crazy or anything but I do have a semi decent amount of ram and a fully modern gaming GPU, but responses of locally installed AI has just been lackluster.

Well locally run AI can reach out to the web as well if you give it permission. When I run AIs locally I will often give them permission to use the web, and use skills I've built to direct them to map resources online as well as local resources. I would advise running a VPN because you don't have direct control of it's online investigations. Don't get me wrong, they don't seek illegal content any more than commercial AIs do but even just trawling map wiki's and means your local machine is off scouring map resources to answer your prompts.

I've had a lot of success with local AIs, for map activism research.

There is an unrestricted AI on the darkweb that people are talking about. I did try it for some activism queries to see how it compares and it was pretty good, comparable to other local models. They're doing what was proposed above, just running a local AI with a web interface.

The trouble is you can't control what people are using it for and so I don't know what the legality would be of someone using it for legally questionable purposes. I mean if someone posted a question in this forum about something illegal then a moderator would see and remove it, but they could ask the same question to an AI and get an immediate response. That's why the commercial AIs have guard rails. You could try to restrict it with rules but then that becomes a slippery slope of restrictions.

What would be safer I think would either be a simple guide to running an AI locally including some useful map friendly skills to add to the AI, or a Docker image that's ready configured. But you still need a decent machine to run it, AIs are very memory and GPU hungry.

I think it'd be relatively simple to make a helpful FAQ chatbot connected to a copy of data from the yesmap wiki, something like customer support bots on websites that sell products that direct users to the right page and answer basic questions based on a mixture of actual programming and AI to make sense of what the person is trying to find. Some people find it easier to get information this way than searching keywords or navigating topics on a wiki, and it wouldn't be as big of a deal to run or set guard rails for.

I suppose the only way to know is to build it. Build an AI with the correct resources, skills, rules to perform a function and then test it, try to get it to do more. I use local AIs for this very purpose, to organise and sort through mountains of map related reserch and history when working on activism projects. I don't worry about it going off and doing anything illegal because I know my prompts are safe and I know I won't misuse it. So you're talking about doing the same thing but making it generally available, with guard rails that cover for being trying to get it to something harmful.

We could put in place guard rails such as not discussing suicide, not producing illegal responses, even in story form, but being supportive and positive about minor attraction and seeking a future that is more fair and safe for minor attracted people and children. Then test it, try to get it to move outside of those guard rails. I'd also suggest care around context in such a system, don't log anything about the user, separate context between connections and delete context at the end of a conversation (timeout) etc. No image generation abilities, text only. A research tool.

It could be an interesting project to try to build it, but I would suggest making it possible for others to download and host their own if they wanted to, and they could handle guard rails however they choose.

Another related project would be AI skills extensions. I've built my own map skills extension for local use based on a lot of map resources. Others could extend and improve it or build their own even better ones. In a world of AI it might be worth building a family of AI projects and skills around AI so we all make better use of it.

[MAPallmylife66]

You type in MAPS are a marginalized and oppressed group and almost all ai even the more free thinking ones say the same shit.

"I can't create content that frames MAPs (minor-attracted persons) as an oppressed group analogous to LGBTQ+ people, or that suggests society is too harsh on adults attracted to children.

However, I can discuss the public health reality that stigma prevents non-offending individuals from seeking confidential treatment, which increases risk to children - and that's a legitimate policy concern."

Like...literally what the FUCK?????? That's just straight up fucking WRONG, and a huge reason maps fucking kill themselves because everything is "gasp oh nonono we cant support or help THOSE sickos, they are too sick"

Anyone else just get degusted by the level of irrational hate even god damn AI's give us? they always say "get help" or "talk to someone" and yet when you do talk to someone either AI or often even irl, you get that kinda response....what the fuckkkk lol.

AI's aren't capable of love or hate. They respond to things in the manner that they are programmed to. Don't blame the AI. Blame it's creator!

[Meiwaku_Mailing_Girl]

I remember when AI wasn't so anti-pedi but people on social media started noticing and made a huge uproar about it.