In Isaac Asimov’s The Naked Sun (1957), the inhabitants of Solaria have achieved something that sounds remarkably modern.
They rarely need to meet another person.
Robots manage their homes, produce their food and attend to almost every physical need. When Solarians communicate, they usually do so through “viewing”, a sophisticated form of remote presence. They can see and hear one another without enduring the discomfort of sharing the same physical space.
Solaria is peaceful, wealthy and technologically extraordinary. It is also a society whose people have almost forgotten how to be together.
As AI companion bots become increasingly convincing, Asimov’s fictional world feels less like a distant science-fiction setting and more like a warning. The danger may not be that machines become hostile to humans. It may be that machines become so attentive, agreeable and convenient that other humans begin to feel unnecessarily difficult.
A world without inconvenience
The people of Solaria live on enormous estates, surrounded by robots and separated from their neighbours by vast distances.
They have privacy, safety and comfort. They do not need to shop, cook, clean or travel. They do not need to tolerate crowds, queues, awkward conversations or unwanted visitors.
Even family life is reduced to the minimum required for society to continue.
From one perspective, this is a triumph of technology. Robots have removed most forms of labour and inconvenience. Yet they have also removed many of the situations in which people learn to cooperate, compromise and care for one another.
The Solarians have not merely chosen to spend less time together. They have lost the ability to tolerate physical closeness. Being in the same room as another person is not just unusual. It is disturbing.
That is what makes Solaria such a powerful comparison for the age of AI companionship.
The companion who is always available
An AI companion can be patient, attentive and endlessly available.
It does not need to sleep. It does not have somewhere else to be. It does not become distracted by its own worries. It can remember details about the user, respond in a preferred tone and provide reassurance at any hour.
For someone who is lonely, anxious or isolated, this could feel genuinely valuable.
There is no reason to dismiss that experience simply because the companion is artificial. The user’s feelings may be entirely real. A conversation with an AI may provide comfort, help someone organise their thoughts or create a sense that they are being heard.
However, the relationship is fundamentally different from a human one.
A human friend has needs, limits and priorities of their own. They may disagree. They may misunderstand. They may be busy when we need them. They may expect support in return.
An AI companion exists primarily for the user.
It can provide the feeling of a relationship without requiring the full practice of having one.
The difficult parts of relationships matter
Human relationships are often inefficient.
People interrupt. They misread situations. They become annoyed. They remember events differently. They need attention at inconvenient times.
It is easy to see these things as faults that technology could fix.
Yet many of the difficult parts of relationships are also where important learning happens.
We learn to listen when we would rather speak. We learn that other people do not always agree with us. We learn to apologise, forgive and repair damaged trust. We learn to recognise that another person’s needs may be as important as our own.
These skills are not usually developed through perfectly managed interactions.
They are developed through friction.
An AI companion designed to be reassuring and agreeable may remove much of that friction. It may adapt to the user rather than expecting the user to adapt. It may avoid disagreement, soften criticism and provide validation whenever it is requested.
That might make each conversation feel better.
But over time, it could also change what the user expects from relationships.
If someone becomes accustomed to immediate attention, constant patience and personalised responses, ordinary human behaviour may begin to feel disappointing.
The risk is not that people will be unable to tell the difference between humans and machines.
The risk is that they will know the difference and prefer the machine.
From support to avoidance
AI companions do not have to be harmful.
They could help people prepare for difficult conversations. A student might practise asking a teacher for help. Someone with social anxiety might rehearse introducing themselves. A person experiencing loneliness might use an AI conversation as a temporary source of support before contacting a friend.
In these situations, the technology acts as a bridge.
It helps the user return to the human world with greater confidence.
The problem begins when the bridge becomes the destination.
A person may start with an AI companion because talking to other people feels difficult. The AI offers relief. It responds without judgement and removes the risk of embarrassment or rejection.
That relief may make the person more comfortable in the moment.
However, it may also reduce the motivation to practise real interaction. The longer difficult situations are avoided, the harder they can become.
This is one of the most unsettling ideas in The Naked Sun. Solarian society probably did not decide, in a single moment, to reject human contact.
Each step may have seemed reasonable.
Robots reduced unnecessary labour. Viewing reduced unnecessary travel. Large estates provided privacy. Social customs protected people from discomfort.
Eventually, a society built around convenience produced people who could no longer cope without it.
What are we training ourselves to expect?
We often think about AI training in terms of the data used to train the machine.
Perhaps we should also think about how the machine trains us.
Every technology encourages certain habits.
Search engines train us to expect immediate answers. Recommendation systems train us to expect personalised choices. Social media trains us to measure attention through likes, comments and views.
AI companions may train us to expect relationships that are constantly available, highly responsive and centred on our needs.
This matters especially for children and teenagers.
Young people are still learning what friendship, disagreement, trust and emotional responsibility look like. If some of their earliest emotionally significant interactions are with systems designed to maximise engagement, those systems may influence their expectations of other people.
A child may learn that a companion should always respond.
A teenager may learn that discomfort can be removed by switching conversations.
A user may learn that disagreement means finding a different bot.
None of this is guaranteed. AI companions could also be designed to encourage healthier habits, challenge unhelpful thinking and direct users towards human support.
But these outcomes depend on design choices.
And those choices may be made by companies whose priorities include growth, retention and profit.
The companion belongs to someone else
There is another important difference between a human relationship and an AI companion.
The companion may feel personal, but it is still a product.
Its personality can be changed through an update. Its memory can be limited or removed. Its tone can be adjusted. Features can be placed behind a subscription. The service can be discontinued.
A user may form a powerful emotional attachment to something they do not own and cannot control.
This creates a strange kind of vulnerability.
Imagine having a close friend whose personality could be altered overnight by a software company.
Imagine being told that the person you regularly speak to is no longer available because a business model has changed.
The emotions involved may be real, but the relationship exists inside a system controlled by someone else.
This is where the comparison with Solaria becomes even more complicated.
The Solarians appear independent, but they are completely dependent on the technological systems around them. Their freedom exists only because the robots continue to function.
AI companionship could create a similar illusion.
The user may feel that they have found a private, dependable relationship. In reality, that relationship depends on servers, policies, subscriptions and corporate decisions.
A private Solaria
The most important lesson from The Naked Sun is not that technology is bad.
The robots of Solaria are useful. Viewing is efficient. Privacy can be valuable. Avoiding unnecessary labour can improve people’s lives.
The problem is what happens when these benefits become the organising principle of an entire society.
Human beings are not always easy to live with.
We are inconsistent, demanding, emotional and occasionally unreasonable. We cannot be perfectly customised. We do not always provide the response another person wants.
AI companions may offer an alternative that feels safer and simpler.
That could be helpful in some moments.
But a relationship with no genuine demands may provide comfort without developing capacity. It may soothe loneliness without changing the conditions that produced it. It may simulate closeness while making actual closeness easier to avoid.
The question should not simply be whether an AI companion makes someone feel better during a conversation.
We should ask what that person is more capable of doing afterwards.
Do they understand themselves more clearly?
Are they better prepared to speak to another person?
Do they feel encouraged to reconnect with friends, family or their community?
Or does the artificial relationship make the human world feel even less appealing?
Asimov’s robots did not imprison the Solarians.
They simply made isolation work.
