- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
Is there any AI tool that simply acts like a tool?
Posted on 1/6/24 at 10:38 am
Posted on 1/6/24 at 10:38 am
I’ve been trying various AI Chat GPT, BARD, BING, GROK….but all seem to have filters that seem to purposefully hide/filter/manipulate responses.
I tested a few things where I only change the subject (romans, vikings, celts, gauls, native americans) and when I get to the native Americans response the responses are a complete 180 from the others. crucifixion impalement beheading..etc.. detailed on the first group however, when it comes to Native Americans, the response is all of a sudden “restorative justice to bring harmony with nature”.
Anyway, I tested it again today, giving out a list of celebrity names and asking it to summarize political campaign contributions that are public record, and all platforms claimed that it couldn’t for privacy. However, I turn around and give it another list of names, and each gave complete details with sources. (The first was a list of celebrities, the second was a list of business people who supported trump, it was a test)
Anyway, I would just like AI to be an educated browser search and not a filter. Is there any test environment that you people are aware that claims to be completely neutral, and has shown so?
I tested a few things where I only change the subject (romans, vikings, celts, gauls, native americans) and when I get to the native Americans response the responses are a complete 180 from the others. crucifixion impalement beheading..etc.. detailed on the first group however, when it comes to Native Americans, the response is all of a sudden “restorative justice to bring harmony with nature”.
Anyway, I tested it again today, giving out a list of celebrity names and asking it to summarize political campaign contributions that are public record, and all platforms claimed that it couldn’t for privacy. However, I turn around and give it another list of names, and each gave complete details with sources. (The first was a list of celebrities, the second was a list of business people who supported trump, it was a test)
Anyway, I would just like AI to be an educated browser search and not a filter. Is there any test environment that you people are aware that claims to be completely neutral, and has shown so?
6
Posted on 1/6/24 at 2:00 pm to Kujo
Just wait. There's a lot going on with open source models right now and where things are headed is you'll be able to install an LLM on your computer and tune it however you want in the not too distant future. Technically you can do that kind of thing today, but it's a little involved, most guides on how to do this start by assuming you already have python installed on your system. That's not hard to do, but this isn't aimed at beginners.
Also, for now most of them require a GPU. Within the next 6 months to a year I'd expect the hardware requirements to come down as well as have some more user friendly installers.
Also, for now most of them require a GPU. Within the next 6 months to a year I'd expect the hardware requirements to come down as well as have some more user friendly installers.
Posted on 1/6/24 at 8:42 pm to Kujo
(no message)
This post was edited on 1/23/24 at 10:54 am
Posted on 1/7/24 at 1:41 pm to TigerinATL
There are cloud hosting services available you can run the llm on.
And uncensored models are alrdy available.
Not necessarily beginer friendly, but with chatgpt4’s assistance and a youtube video guide… i was able to get it up and running without too much fuss
And uncensored models are alrdy available.
Not necessarily beginer friendly, but with chatgpt4’s assistance and a youtube video guide… i was able to get it up and running without too much fuss
Posted on 1/8/24 at 10:20 am to Kujo
quote:
Anyway, I would just like AI to be an educated browser search and not a filter
THAT'S...THE...POINT
You will get the information deemed appropriate, nothing else, and you will like it.
Posted on 1/10/24 at 9:48 am to Kujo
It doesn't know anything other than the inputs given to it. GIGO.
There are only two options: program it to stay within certain parameters, or leave it wide open and have trolls sabotage it like with the TayTweets debacle.
There are only two options: program it to stay within certain parameters, or leave it wide open and have trolls sabotage it like with the TayTweets debacle.
Posted on 1/12/24 at 3:22 am to Kujo
Well, “Native American” is a catch all term for dozens of tribes. It’d be like typing “Europeans” and wondering why there’s no mention of Nazis or Mussolini.
Have you tried narrowing it down to specific tribes?
Have you tried narrowing it down to specific tribes?
Posted on 1/12/24 at 4:42 pm to Kujo
quote:
Anyway, I would just like AI to be an educated browser search and not a filter.
Until we get a solid open source AI, you will get answers that won’t hurt the bottom line of the company running the AI you are using.
And, yes there are many that act like a tool. We have one built in to our document management system. It will spit out the best answer based on what you searched. Good news or bad news. But we are paying for the tool to search our data w/o building in any PC or Woke restraints. Tainting the data is great for getting investors and keeping the masses happy that you don’t give racist, sexist, etc answers based on the data you have. It’s not so good when your bottom line depends on truly accurate results.
Page 1 of 1
Popular
Back to top
Follow TigerDroppings for LSU Football News