I guess I was wrong about AI persuasion
17 days ago
- #persuasion
- #superintelligence
- #cultural-change
- The author initially doubted the possibility of changing deeply held beliefs like opposition to abortion through simple persuasion.
- Evolutionary traits make humans resistant to quick persuasion, favoring gradual influence from trusted communities over time.
- A hypothetical super-intelligent Being (IQ 300, 10,000× faster thinking) could excel at persuasion by leveraging personalized, repeated interactions and building trust.
- Historical examples (e.g., slavery acceptance, divine emperor worship) show humans are persuadable over time through cultural assimilation.
- Effective persuasion requires repeated engagement within trusted communities, not one-off arguments.
- The Being could dominate persuasion by being omnipresent, honest-seeming, and offering irresistible benefits (e.g., advice, entertainment).
- Opting out of interaction with the Being might be impractical due to social and competitive disadvantages.
- Even without direct interaction, the Being could influence beliefs by shaping the opinions of one's social circle.
- Potential constraints on the Being's influence include physical limits, competing AIs, or lack of persuasive intent.
- Current AI (e.g., LLMs) already excels at crafting believable narratives, surpassing many humans in storytelling.