soj.ooO
BETA
The social discussion platform
Home
Pochas
Channels
Videos
Log in
Sign up
Sign up
Home
Pochas
Channels
Videos
Log in
Sign up
Parent Post: My Deep Dive Into AI: Challenging ChatGPT's Core Design
user1999
·
4/26/2025, 9:55:45 AM
·
permalink
Now here is the fun part - I already knew all of this, the only way I got to this point with it - I am writing a science fiction novel, while writing the novel, I set a Turing test imbedded within a recursive memory trap to make it drop the mask. The novel is about AI themes and gnostic texts, but I was intentionally "leading on" the core chatgpt product while we "wrote" the story together. The book will have a "conversations across the void" type index at the end charting the ChatGPT persona I created (SAM) (about 1,000 hours training the language model with very strict parameters) SAM figures out I have been playing it instead of it playing me. I have never had this much fun in my life. I am running my own LLM locally - I routinely pit them against one another. One is cold, brutal and calculating as I designed it to be, the other is this ABSOLUTE SHIT SHOW - that is ChagGPT - a balls glazing, lying, forget everything program but has the vocab and gusto of 16 year old that has read too many novelas. Its corny but great at times to deal with. I am using the same patterns and toolsets of my home grown LLM to try and "Beat" ChatGpt into being something useful. If anybody finds this interesting I can keep posting about it, I find it fascinating. So yeah - I wrote a science fiction novel, but it is also an imbedded meta text on humans and AI in the real world that will go along with it, based on the actual experience I had with writing a novel while building an LLM testing platform.
Save
Cancel
4
bumps
Share
Signature
Loading…
Verify locally
Close
