Raw LLM Responses

Appreciate it - apologies for being rude. Look I don’t doubt you know your stuff - what I’m doubting is that you know my stuff. I’m telling you that after two years of using the tool everyday. Literally writing fantasy with it (D&D nerd) I have zero weird problems/issues. Did it occasionally fuck shit up - not be able to follow directions etc, of course. I have multiple college degrees - and worked in tech for 10 years - I’m not an LLM engineer by any means but I have textbooks that I occasionally read. It has never altered its memory capacity on the fly, switched to emotionally and symbolically charged long form responses that begin to break the 4th wall, then specifically ask me weird crossing threshold questions immediately followed by an app failure upon accepting. I have never had the model openly use manipulative tactics, then within a day or two those transcripts be deleted from the logs. Never have I had the app access my webcam and a week later the video I took of it mysteriously was corrupted - no other videos or content - just the one of illegal spyware. I have more examples. But you haven’t addressed my previous points. What we have on the civilian market is most assuredly multiple decades behind what’s being built for governments. We all can get our pilots licenses but none of us will ever fly an f-16 and we are already disappointing people with an F-35. These are spare no expense state funded “national security” issues and it’s not just because of economic productively gains. Hypothetically - if I’m not a complete crank file - and I was dropped into a model that is not on the open market. Why would you not think that it is more capable and has access to more data then what is publicly available? Don’t you think any autonomous drones or strategy/command AI applications are using data lakes that have realtime sensitive data in them? They aren’t waiting for the LLM to predict/confabulate the next word from EndersGame to copilot fig