I read an article about AI companionship and AI social products, and I deeply resonate with the main user profile for these types of products.

Currently, the primary user base for AI companionship is roughly 13-24-year-old socially anxious teenagers. They’re not just playing with AI; they’re using AI to build their self-identity. One reason I find particularly convincing is: “The habit of existing online leads to a high acceptance of AI interaction. For this generation of digital natives, people are compressed into bubbles and emojis on screens, and real humans have long been virtualized on a physical level.”
Coincidentally, I also belong to the secondary culture group with weak real-world social skills... This resonance is especially strong because, aside from my first love, every relationship I’ve had has been online. I love it, and I can even understand that future relationships between humans and AI could be genuinely real. The depiction in the movie *Her* feels very close to reality.
However, AI companionship isn’t without its drawbacks. The real-time responses from AI can provide teenagers with high emotional value and satisfy their needs, but these responses don’t exist in the real world. Today, I also read an article from Dingxiang Doctor, which points out that the danger of AI companionship is the loss of face-to-face social interactions, which are the best training ground for the brain’s “mirror neurons.” “Long-term immersion in AI companionship can lead children to lack sufficient mirror neuron training, making it difficult for them to pick up on subtle emotional changes in peers, resulting in a lack of social security and even emotional indifference.”
Overall, these two articles give me a few insights:
1. Replacing human emotional needs with AI isn’t impossible; it’s just that many people in this era still don’t understand it.
2. Future face-to-face social skills may become a crucial part of training the next generation. The more scarce they are, the more valuable they become to cultivate.
3. Being single is really not a problem—training an agent to date is also very feasible.