Conversation

Notices

1. Embed this notice
   NO NAME (jeff@soapbox.hackdefendr.com)'s status on Saturday, 31-Aug-2024 03:51:03 JST NO NAME

   Sure are a lot of #AI haters on #Mastodon.

   With the hate I am also seeing the same level of shortsightedness and a complete disregard for the benefits of AI already working in the medical and health fields.

   Sure there are some out there with a point. Some are just walking around with their panties in a bunch. Maybe you are scared of losing your job to AI? Maybe you are just paranoid? Maybe you aren't as smart as you thought you were and AI exposed it?

   Whatever it is, you are stomping on the lives of people that actually found good use for AI. So maybe just keep your negative thinking off social media. If ya don't like something then don't boost or like it. You can even filter words or phrases to avoid seeing anything about the topic in your timelines.

   </rant>

   • Embed this notice
     BeAware :fediverse: (beaware@social.beaware.live)'s status on Saturday, 31-Aug-2024 03:51:02 JST BeAware :fediverse:

     in reply to

     @jeff I just wish people that advertise it would stop advertising it as "knowledgeable" or containing knowledge at all...

     That's not what these things are for and they wonder why it "hallucinates".🤦♂️

     If you're asking it questions and expecting a factual answer that's outside of the scope of "does this sentence structure seem correct?" Or "summarize this 10 page document", then you're asking for it to give you wrong answers.

   • Embed this notice
     BeAware :fediverse: (beaware@social.beaware.live)'s status on Saturday, 31-Aug-2024 04:03:12 JST BeAware :fediverse:

     in reply to

     @jeff right, the model has to be "within scope" for it to be factually knowledgeable.

     ChatGPT is trained on too much data to be considered knowledgeable and as it's the most used model, that's what I refer to when discussing things like this.

     Sure, if I train a model on a paragraph of a specific book, it'd probably be pretty damn good at knowing the information of that chapter.

     On the contrary, if I train it on 100 books, it probably won't know a damn thing about a specific one and would get jumbled up, quite often.

     It's all about scope of training in those cases.

   • Embed this notice
     NO NAME (jeff@soapbox.hackdefendr.com)'s status on Saturday, 31-Aug-2024 04:03:13 JST NO NAME

     in reply to

     • BeAware :fediverse:

     @BeAware

     The knowledge a model has depends on how a model is trained. If the proper care is taken at the training phases then the knowledge within a model is more accurate when used for those specific use cases.

     As for how people are choosing to label or describe AI today. I can't do a thing about them and so I just filter them out.

     There are good uses for AI and some people are quitting research because of constant ridicule - this needs to stop.

   • Embed this notice
     BeAware :fediverse: (beaware@social.beaware.live)'s status on Saturday, 31-Aug-2024 05:24:12 JST BeAware :fediverse:

     in reply to

     @jeff alt text is different from knowledge about a particular subject.

     I don't do any official research or anything but I've been dabbling in these things for around 3 years now.

   • Embed this notice
     NO NAME (jeff@soapbox.hackdefendr.com)'s status on Saturday, 31-Aug-2024 05:24:13 JST NO NAME

     in reply to

     • BeAware :fediverse:

     @BeAware are you speculating or do you have real world experience in AI research? I’m just asking.

     My roommate and I have done some work, as stupid as the subject matter is, with meme identification and alt text updating for the visually impaired here on Mastodon. We use open source models for the identification piece and a simple language model for the description writing. These models work in tandem with each other using custom scripting that reads from a OneDrive where each meme subject is its own folder that also maps to an account on our Mastodon instance.

     Now this just scratches the surface of a necessary use for AI. We continue to do our research using every new technique discovered along the way. There is really a lot more that could be done with properly trained models that specialize in one subject, but people are scoffing the whole idea of AI before it has been given a chance.