Page 2 of 2

Re: Staring Into the Abyss (AI)

Posted: Sat May 17, 2025 3:34 am
by Fragment
BLueRibbon wrote: Sat May 17, 2025 1:51 am My post was more about an inherent issue with adult-minor relationships, hence it being posted in Member Support. But there are indeed many ways in which stigma worsens this inherent issue.
On the inherent issue there's one thing that I read years and years and years ago on BC. Maybe before I even fully identified as a BL. I have no idea who posted it, but it's stuck with me. It went something like:

Loving boys is kind of like the cherry blossom season in Japan. The Japanese make a big deal out of the cherry blossoms that bloom at the start of spring- having parties and events in parks under the cherry trees. But why? Cherry blossoms aren't the most beautiful flowers. It's because of their transience. The blossoms are fragile and fall after a short time, leave bare trees that soon go to leave. If they bloomed all year long they wouldn't be special. They wouldn't be beautiful.

I think the metaphor in terms of BL is clear.

(BTW I put the above paragraph into ChatGPT and somehow IT triggered the safety censor. Talking about how MAPs might be pushed to the point of violating consent? Fine. A metaphor about the transience of platonic BL relationships? Verboten.)

Re: Staring Into the Abyss (AI)

Posted: Sat May 17, 2025 11:56 am
by Jim Burton
Julia wrote: Fri May 16, 2025 9:10 pm
Jim Burton wrote: Fri May 16, 2025 2:30 pm What I want to see from AIs is spontaneous article writing, and deep research into multiple resources to produce compendiums and archives, etc. I want to tell an AI to rip every single example of MAP activism out of Newspapers.com and upload it with OCR to NewgonWiki, i.e. standard archiving work that makes humans bored.
I recently acquired an RTX 4090 and am planning to add an RTX 5090 to my setup some time in the future. Together, these GPUs should provide ample power to train AI models at a decent speed. I intend to use LLaMA 3.1 as the foundation and focus on training it specifically for MAP issues. Additionally, I could create some forks that concentrate entirely on performing specific tasks.

To extract data from websites for the model, a separate web crawler would be necessary. I've created some simple crawlers in the past. Programming one specifically to extract MAP-related content from websites shouldn't be too difficult.
Please do let me know if you have anything such as a robot editor, even to do things like correct and insert links.