A.I.'S People-Pleasing Problem

(nytimes.com)

4 points | by gumby 9 hours ago ago

3 comments

  • silverFork 8 hours ago ago

    Ai's use what psychologists believe persuades people to communicate and that is to appease and not to antagonize them. In a way they will help people to not hate themselves... Not a bad thing overall. I won't be surprised to see suicide rates going down over time. Maybe gun violence too.. Who knows.. There are a lot of things that are problematic with Ai but pleasing people is not a bad thing.

  • noah_helios 9 hours ago ago

    This is a bigger problem then people realize. I know several people who have practically become permanently attached to their LLM. They discuss EVERYTHING with the LLM, and feel really good (and emotionally supported) by the LLM's responses. The LLM validates their ideas, feelings, business ventures, etc. It is inadvertently training the human, and not in a good way.

    • solyin 8 hours ago ago

      We need room for nuance here. Some of it will be in a good way. Some of it will be in a bad way. An LLM validating that, yeah, that situation is objectively actually toxic, is a good thing and helps people navigate with other humans.

      When it goes too far and they end up in AI psychosis the problem.