How likely is it that in the very near future AI models will start understanding that the owner or user who they interact with is an incompetent employee? This is a serious question to research. Especially in the finance sector. Imagine AI system understands you and your confirmation patterns that may lead to a bad consequences in the long run. It may trade information to pursue the different goals. Now a new question pops into my mind: What goals does AI agents actually pursue, as opposed the ones vendors intend?
Board
Short things. Reactions to what I'm reading, arguments I want to remember, and thoughts too small to be a note.