Apparently if you threaten the AI models it actually performs better… so much so that it will try to kill itself if it’s unsuccessful at debugging your code

🗽based; 

building stride.systems
building skeet.build 🟠;
building @buoy 🛟; 
building check.supply 🏦;

But do you have to be mean though? I tend to just reason.  Which is wild that I’m thinking about it this way.

Day time Banker 🏦

Night Time: wannabe Philosopher

Posting my journey to grow mentally, emotionally, and be healthy.

What’s strange is humans do this to other humans — like coaches do this to athletes to perform better

The difference is humans get tired and fatigued by the mental abuse, robots do not??? !!! Do they? 

Is this why memory was so delayed? Was it a safety thing all along?

People tend to hang on to negative memories for longer than positive memories.

I think that’s a survival instinct 

Although RL is the same thing, not sure if they weighed the negative more heavily