Like it or not, AI is learning how to influence you
1 min read
Summary
AI agents are set to become increasingly ubiquitous in daily life, deployed to decode and optimally influence human personalities for varied ends, from selling products to spreading misinformation.
Such systems are soon to be deployed on a mass scale following the recent lowering of the cost of real-time processing, and will be used to interact with humans via conversant avatars on PCs, phones, and AI-powered glasses, etc.
These highly persuasive AI ‘salespeople’ will be able to quickly earn human trust, using large amounts of stored data to persuade users to buy products, and could spread misinformation with no way of knowing whether this is the case as they will look, sound and act human, inferring that trust will be instinctive.
To avoid this, AI must have smart, narrow restrictions to avoid superhuman manipulation, including a ban on feedback loops and access to personal data.