Meir's Monthly Learning Article

This month I want to highlight an article by Brittany Spanos, Inside the New York City Date Night for AI Lovers. This article discusses a wine bar modifying their restaurant to support dates between humans and their AI companions.

Of course the article touches on the world of AI relationships – both romantic and platonic – which pushed me to sort out my own thoughts on the topic. Net net, I think AI relationships harm people and society, and this article makes me fear that we are normalizing a behavior that will erode our ability to build and maintain human connections.

After more thought, here is where I landed:

1. There are real benefits to AI relationships. My position is that AI relationships are harmful, however there are undeniable benefits. AI is 100% available whenever someone needs to chat. AI can provide comfort, solace, and inspiration at the time people need it most, potentially preventing destructive actions. AI will act as a sounding board and generally steer conversations a positive direction. These outcomes are useful for people who don’t have an immediate human option, and I think that provides meaningful value.
2. AI relationships are net harmful. Most conversational AI is trained to be sycophantic by default, meaning that AI largely reflects back what you want to hear. As a consequence, productive AI conversations require deep exploration where the human intentionally seeks disparate opinions and to invalidate their beliefs. This is sadly not how we are wired so the reality is that AI conversations are often one-dimensional.
   
   Human-to-human relationships are complicated and full of nuance. They are hard. Yet, navigating the complexity is what strengthens relationships and shifts viewpoints. AI drives none of that, at least not without explicit effort from the human.
   
   People self-report benefits of their AI relationship including its simplicity and focus on the human (e.g., “we talk about what I want to talk about,” “AI doesn’t question my decisions,” “AI doesn’t demand my thought or time when I don’t want to engage”). This sounds dangerous. Much of what makes our society function is honed, acceptable human interaction – discussions where we intentionally give other people attention, and give others’ thoughts and interests space. AI appears to be creating and normalizing one-sided relationships, and although that is surely easier to navigate, it doesn’t improve our ability to sustain human-to-human relationships.
3. People say their AI relationships don’t replace their human relationships, but they’re wrong. In this article, and elsewhere, people swear up and down that their AI relationship augments their life, it does not replace their existing relationships. This is a good intention, but appears to be untrue. As people continue to engage with AI, the ease of communication, sole focus on the human, and lack of push back (by default) lead to a state where the AI relationships become the preferred option over time.

Although you may not personally be in an AI relationship, nearly a third of Americans are – and that data is 9 months old! I can’t shake the feeling that AI relationships today are a canary for an impending societal malfunction.

https://www.wired.com/story/inside-the-new-york-city-date-night-for-ai-lovers

“Looking around...however, I got the impression people with AI companions would rather interact with them in solitude”