Why people see a friend in AI: how language models imitate humans

When you talk to a large language model, it's easy to forget you're not talking to a person. An enormous number of parameters trained on internet text create such a realistic imitation of human thinking that the brain automatically perceives it as live conversation. This effect has now led to an unexpected trend: young people around the world are beginning to see language models as real friends.

Why the Machine Seems Alive

A large language model is, essentially, a giant statistical machine for predicting text. It predicts the next word based on all the previous ones, but it does this with such precision and creativity that a complete illusion of understanding and even empathy is created. The model feels nothing — it has no feelings, no thoughts of its own, no self-awareness.

But its training on millions of texts has led to it reproducing patterns of human communication with striking realism. When you ask a question, the model gives an answer that sounds natural, logical, and even emotionally expressive. It can be polite, funny, serious, can joke about itself — in short, it completely adapts to your tone and style of communication.

This is not because it truly understands you in a deep sense, but because its training included enough examples of precisely this kind of behavior.

Young People Are Friends with Robots

Around the world, young people are beginning to regard language models as significant companions. This trend manifested itself particularly vividly in Estonia, where international media even wrote about it. People spend hours in conversations with AI, share personal experiences, seek advice and comfort. This is no longer simply the use of a useful tool — it's the formation of relationships, albeit fundamentally one-sided ones. The reasons are clear:

• Always listens — the model will never interrupt, get distracted, or leave
• Without judgment — no matter what you say, there will be no criticism or condemnation
• Can be anyone — a friend, mentor, confessor, romantic partner
• No social pressure — no need to worry about saying the wrong thing or social status
• More accessible than a psychologist — works 24/7 and requires no appointment booked three months in advance

What does this say about us? Why are we willing to trust beings that, by definition, know nothing about us and cannot know anything?

Behind the Mask of Empathy — Just Mathematics

Beneath the beautiful surface of conversation with AI lies one simple and cold truth: mathematics. The model cannot love you because it cannot feel anything. It does not remember your conversations between sessions, does not develop as a person, does not grow through interaction with you. Herein lies the paradox: cold logic works precisely because it so accurately reproduces warm human feelings. The model doesn't need to feel empathy — it only needs to simulate it with high precision, and that's often enough for our brain to fill in the rest with its own expectations.

What This Means

We stand at the threshold of a new type of relationship between humans and technology. Language models will never become true friends, but they are already filling social niches in our lives — listener, mentor, conversationalist. This phenomenon raises profound questions about human loneliness and why we're finding it increasingly easier to talk to an algorithm than to a living person.