For years I’ve been interested in the idea of AI delegates—agents that could learn your values and advise you on votes you’re taking, or even vote on your behalf.
"But AI seems vulnerable to a particularly adversarial kind of framing that humans may not be."
Not sure exactly what you're suggesting. Maybe it's uncomfortable because they are vulnerable to adversarial framing differently that we are? Maybe we feel better about being tricked because somehow that still shapes the world in terms of our (unconscious, manipulated) preferences so it feels more familiar or in control?
"But AI seems vulnerable to a particularly adversarial kind of framing that humans may not be."
Not sure exactly what you're suggesting. Maybe it's uncomfortable because they are vulnerable to adversarial framing differently that we are? Maybe we feel better about being tricked because somehow that still shapes the world in terms of our (unconscious, manipulated) preferences so it feels more familiar or in control?
This is an excellent piece about something that not enough people are thinking about.