Basics

Some chat bots such as ChatGPT “remember” what you have asked it/said in the past etc
This is SUPPOSEDLY done to make it better at anticipating your needs/faster
- and to be fair google is doing this a bit too recently
  - Sometimes, for technical information etc it can be handy, essentially bypassing Disambiguation Pages / multiple rounds of making sure the google search is written right/subtracting unrelated words etc
  - Other times it gets in the way/preventing me from seeing what i need to etc (basically all use cases BUT looking up specific instruction manuals/specific stuff like that with keywords)
HOWEVER, this can also be used for Manipulation via Agreeability/Telling you what you WANT to hear, rather than what you need to hear/what is more true
- Given OpenAI is No Longer Open or Non-Profit , their main motive is getting people hooked + Maximizing Engagement etc, all of that is more important than truth or actually helping people research

Potential Remedies

Ie using them only when especially stumped as a “hey let’s see if it finds a paper?” Type solution
Hierarchy of Controls etc

Essentially how often should the data be purged to keep it from getting AI Brain Rot from recirculating LLM Answers into Prompts (that sort of feedback loop is the main issue, akin to Needs More JPEG etc

A 2025 Article Titled “ The tech bros are making themselves sick”
- Essentially AI Induced Psychosis type stuff where the LLM’s agreeability latches onto something and leads someone down a BAD death spiral
- That and mentioned the neat concept of LLM Chatbots as an Internet Shuffle Button (akin to the “i’m feeling lucky” button
  - If you’re lucky you'll get a summary of a useful reddit post answer
  - If you’re unlucky you’ll get Mecha-Hitler (a reference to…THAT incident with Grok ), an AI Summary of every edgy bro/cyberbully/neo-nazi on 4 chan etc