Discussion What unrestricted ai do yall use (21 Viewers)

[Cheat]

u know u can just ban him right

ok im sorry pls don't ban me, ill end up in mental asylum without .gg, i redact everything i've said... Biomaxx Godveil Heir

 

[Circadex]

It's like stepping into Ukraine. Or russia? Forums full of euro fags probably pissed off one side

 

[Dexter]

spicychat.ai

 

[framer]

Gpt got another safety update and groks getting worse and worse

I need a little slave to fetch me roid sources

Pleaaaassseeee

just go on meso rx or t nation

 

[XvideosDemon]

Gpt got another safety update and groks getting worse and worse

I need a little slave to fetch me roid sources

Pleaaaassseeee

Non ion want the Jews knowing I want to remove diddys clothes

 

[Nardicus102]

If you high IQ and know how to script, use terminal and some python here:

https://github.com/p-e-w/heretic

This question is better asked to people who make AI porn, as those people have this down to the biggest tea which is the hardest restriction

 

[Nardicus102]

You can just download the ai and remove the safety features

DO you use your LLM locally on your computer ? which model, cause I know you can do the but its very model dependent

 

[Nardicus102]

Downloading the model doesn't work because, first of all, you need to remove the restriction that requires actual technical knowledge.
Secondly, you need a supercomputer to process the information. Your RTX GPU won't do anything and will only produce a horrible, inaccurate result.

Yea this is true for the most part, tho there are GitHub repos that unlock restrictions if you look hard enough n have decent technical knowledge, and with newer more accurate LLM's it become harder

 

[Biomaxx]

If you high IQ and know how to script, use terminal and some python here:

https://github.com/p-e-w/heretic

This question is better asked to people who make AI porn, as those people have this down to the biggest tea which is the hardest restriction

 

[Godveil Heir]

Yea this is true for the most part, tho there are GitHub repos that unlock restrictions if you look hard enough n have decent technical knowledge, and with newer more accurate LLM's it become harder

the GPU, the training, & the data it needs to be fed isn't going to allow it
you NEED a supercomputer to host an LLM.

perplexity works just fine for this
90% of the time you won't face any restriction

 

[Nardicus102]

the GPU, the training, & the data it needs to be fed isn't going to allow it
you NEED a supercomputer to host an LLM.

perplexity works just fine for this
90% of the time you won't face any restriction

This super computer logic only applies when your actually training a model, like configuring neural networks(Which most people here dont even have a mathematical background to even accomplish) But you can run/modify local models via terminal with scripting.
Now in terms of locally configuring a huge LLM than yea an expensive GPU does come in handy but I dont think bro owns a law firm or is doing massive data analysis to even garnish the extra load.

I haven't really looked in to preplexity but ill try it out as I feel most of the time when you use word like "hypothetically and in theory" when asking a question its more likely to give you a awnser your looking for so its mostly prompt engineering on users end

I have tried scripting and configuring LLM's locally and the biggest problem is running is making sure it still preforms (via giving the right answer) even after restrictions are removed. I gave up on this. But have some peer's at Princeton who are physics majors n have this down to a tea.

n honestly Larger models only matter if the task in hand is complex enough to demand said model which in case a nicer GPU is needed, majority of time if you have to ask said question, than the task at hand isnt even complex enough to begin with

 

[Godveil Heir]

I haven't really looked in to preplexity but ill try it out as I feel most of the time when you use word like "hypothetically and in theory" when asking a question its more likely to give you a awnser your looking for so its mostly prompt engineering on users end

it's not about hypotheticals at all
just directly ask the question and don't mention crime and don't try to use it for reasoning
for search, it works fine

This super computer logic only applies when your actually training a model, like configuring neural networks(Which most people here dont even have a mathematical background to even accomplish) But you can run/modify local models via terminal with scripting.
Now in terms of locally configuring a huge LLM than yea an expensive GPU does come in handy but I dont think bro owns a law firm or is doing massive data analysis to even garnish the extra load.

nope
you can run it for sure; I've done that
but it won't give nowhere near accurate response
it will just give you misinformation.

 

[Godveil Heir]

n honestly Larger models only matter if the task in hand is complex enough to demand said model which in case a nicer GPU is needed, majority of time if you have to ask said question, than the task at hand isnt even complex enough to begin with

this is also false
the LLM works by predicting the next word

it's going to be ass, if it hasn't received top tier training and not allocated extreme processing power