Comment by kurthr

The disturbing thing to consider is that it might be bad acting AI with human systems. I can easily see a situation where a bad acting algorithm alone wouldn't have nearly so negative an effect, if it weren't tuned precisely and persuasively to get more humans to do the work of increasing the global suffering of others for temporary individual gain.

To be clear, I'm not sure LLMs and their near term derivatives are so incredibly clever, but I have confidence that many humans have a propensity for easily manipulated irrational destructive stupidity, if the algorithm feeds them what they want to hear.