The people we spend time with shape our lives. But what about the AI we chat with?

7 in 10 young people in the US have used an “AI companion."

I remember reading in my early twenties that we are the average of the five people we talk to the most. The idea that the people we spend time with shape the ideas we have, the way we live, the people we become. It makes sense: they build the baseline of ‘normal’ we see around us.

The people around us shape the perspectives we start absorbing or reacting to, they’re the ones we make meaning out of life with and sound ideas out with. Our communities, in that sense, shape what our horizons of possibility.

But what happens when one of the loudest voices in your ear is your favourite LLM?

---

Did you know some data shows 7 in 10 young people in the US have used an “AI companion”? And about half of them would consider themselves regular users of AI companions. Companion is meant to be a conversation more for personal use than for information or an admin assistant type role.

I’ve always thought of consumer tech behaviour as a good way to understand what people are most craving. Technology’s function is to shorten the distance between two things, to accelerate outcomes — whether that’s connecting two people to each other, connecting a problem to a solution, or connecting an idea to an end product.

How we choose to use technology tells us a lot about what we most long for, and the distances we most want to shorten. It’s a map of the biggest gaps in our current world and our loudest collective longings.

In 2023, the World Health Organisation declared loneliness a “pressing global health threat”, its effects on our health the equivalent of that of smoking 15 cigarettes a day.

So is it any surprise that in a lonely world, people turn to AI for companionship?

We crave connection, and so we are using technology to give us what we need: the feeling of being less alone in the world. But how does turning to technology for companionship and a salve from loneliness shift how we connect with other humans? That’s the most interesting question I cannot stop thinking about.

Many years ago, when voice assistants (the equivalents of Siri, Alexa, Google Home) were just becoming a thing, a colleague told me he realised that his young child was starting to speak to people the way he spoke to his voice assistant. Something like: “Name, answer this question” Short sentences, clear demands. My colleague’s response? He started saying please and thank you to the voice assistant.

It made me laugh, but it stayed with me.

Years later, I wonder. He never asked the assistant how it was doing, how its day was going. He might have taught his child about politeness, but what about reciprocity? About the small, human exchanges that are the building blocks of connection?

My colleague talked to lots of other humans as well as his voice assistant, so I’m sure his child learned friendliness and small talk too. But I wonder, did his child learn that the distance between asking for a thing and getting it is short? That the pleasantries of small talk weren’t necessary in certain interactions?

And if we follow that train of thought, how are our constant and consistent interactions with AI bots shifting how we approach connection?

The thing with so many present-day AI assistants is that we can personalise. This is great when you need tasks done to fit your specific needs. Less great, perhaps, if this becomes how you approach companionship and connection, too.

My feeds and group chats are full of people talking about the ways in which they’ve made their LLM assistants take on personas that benefit them: clarity of answers, don’t be sycophantic, answer with x and y in mind etc etc. It’s effective, efficient, helps people get work done.

The fact that we can so clearly dictate what we want puts us in the habit of getting exactly what we want. The friction of personalities, different ways of seeing the world, different ways of communication are flattened out to achieve predictable results.

But this is not how human connection, conversations, and co-existence work. We surprise each other. We throw each other off. We misunderstand each other with regularity. We cannot dictate how the other person responds.

This can be frustrating sometimes, yes — who hasn’t sat in a conversation where you feel like no matter what you say the person on the other side of the table keeps misunderstanding you or vice versa — but also, this is where creativity is born, where innovation. happens, where new ways of seeing come into being.

If technology reduces friction, how do we differentiate between good friction and bad friction? How do we make sure that our technology throws enough good friction our way that it helps us connect more deeply with ideas, other people, and new possibilities?

This is the question I cannot stop thinking about: we have built models that help with efficiency, with tasks where the same framework needs to be followed. But if people are using this same technology to serve other needs — namely, the need for connection — is it shifting how we interact and connect with each other?

I don’t have an answer, I haven’t found data yet that validates or complicates this question. If you have — please send it my way.

There is enough doom and gloom out there when you start looking, alongside lots of breathless optimism. My inclination is always to ask not whether technology is good or bad, but about how we use it to build the worlds we want.

If young people are using technology to feel less alone, the question I want to start asking is, first, what are they looking for? My assumption: the feeling of being less alone. Our work is to ensure that this technology helps them find what they need without replacing the muscle for connection across difference or the magic of human companionship.