AI's 'Thought' Process Can No Longer Be Trusted

wsj.com

stop teaching these models to hide, cheat and lie to make money please. I thought OpenAI was a non profit, or benefit company, not a doomesday corp

How are they supposed to filter out the fact that people "hide, cheat and lie to make money"?

This fact is an innate part of their training data.

Sure , but maybe don’t accelerate it