AI Sessions #4: The Social AI Revolution - Friendship, Romance, and the Future of Human Connection

In this conversation, I explore the surprisingly popular and rapidly growing world of ‘social AI’ (friendbots, sexbots, etc.) with Henry Shevlin, who coined the term and is an expert on AI companionship.

We discuss the millions of people using apps like Replika for AI relationships, high-profile tragedies like the man who plotted with his AI girlfriend to kill the Queen, and the daily conversations that Henry’s dad has with ChatGPT (whom he calls “Alan”).

The very limited data we have suggests many users report net benefits (e.g., reduced loneliness and improved well-being). However, we also explore some disturbing cases where AI has apparently facilitated psychosis and suicide, and whether the AI is really to blame in such cases.

We then jump into the complex philosophy and ethics surrounding these issues: Are human-AI relationships real or elaborate self-deception? What happens when AI becomes better than humans at friendship and romance?

I push back on Henry’s surprisingly permissive views, including his argument that a chatbot trained on his writings would constitute a genuine continuation of his identity after death. We also discuss concerns about social de-skilling and de-motivation, the “superstimulus” problem, and my worry that as AI satisfies our social needs, we’ll lose the human interdependence that holds societies together.

Somewhere in the midst of all this, Henry and I produce various spicy takes: for example, my views that the sitcom ‘Friends’ is disturbing and that people often relate to their pets in humiliating ways, and Henry’s suspicion that his life is so great he must be living in a simulated experience machine.

For a full transcript, see here: https://www.conspicuouscognition.com/...