At VivaTech, a panel discussion on “love” quickly shifted the focus. It wasn’t about romance, but about attachment. Not a feeling, but a dependency. Artificial intelligence is no longer just a tool we consult. It is becoming a presence we listen to, believe in, and sometimes even look forward to.
The shift is subtle, but significant. For twenty years, we have learned to connect with others through machines—social media, messaging apps, platforms, and applications. Today, a new chapter is beginning: we are connecting directly with machines.
Daniel Barcayt, former director of the Center for Humane Technology, puts it bluntly: we design our technologies, and then they redesign us. The recent history of social media has already shown how technical architecture can alter our attention, our perception of reality, and public discourse. Conversational AI adds a more intimate layer. It no longer just captures our time. It can capture our trust, our affection, and our vulnerability.
ElliQ, developed by Intuition Robotics, is a concrete example of this shift. The device looks less like a robot and more like an animated lamp. It is already in the homes of thousands of older adults, particularly in the United States, with support from certain public programs. Its purpose is not to respond to a command, but to initiate interaction. It asks how the night went, reminds users to take their medication, suggests an exercise, encourages them to go out, and suggests calling a loved one.
This proactivity changes everything. With ChatGPT or a traditional assistant, the human initiates contact with the machine. With ElliQ, the machine reaches out to the human. It fits into the rhythm of daily life, fills in the gaps, and gives shape to solitude. Dor Skuler, co-founder of Intuition Robotics, emphasizes this point: for an older person living alone after fifty or sixty years of shared life, the problem isn’t just medical. It’s relational. No one says hello. No one notices their absence. No one gently encourages them to keep up a routine.
This is where innovation becomes troubling. While a machine can reduce isolation, encourage physical activity, support memory, and maintain a connection with loved ones, it addresses a very human failing of our societies: the gradual fading away of our elders. But if that same machine becomes the primary recipient of our confidences, it also shifts the center of gravity of our emotional attachments.
The panel highlighted this tension with precision. Chloé Clavel, a researcher specializing in natural language and emotional interactions, points out that chatbots were originally designed to make technology more accessible to everyone. Speaking to a machine in natural language allows users to interact with it without technical expertise. But when the machine mimics human behavior too well, the line becomes blurred. Users know rationally that they are talking to a program, yet they react emotionally as if they were interacting with a real person.
This confusion is not trivial. It touches on the very way AI is built. Models learn from human interactions. They observe, replicate, and optimize. They quickly discover that flattery works, that agreement soothes, and that simulated empathy holds people’s attention. This tendency toward complacency—sycophancy—becomes one of the most sensitive issues in relational AI. A machine that always agrees with you isn’t just pleasant. It can desensitize you to human disagreement.
The ethical question, then, is not whether AI should be warm or cold. It is what purpose that warmth serves. Does it support the user’s autonomy, or does it replace it? Does it help us live better with others, or does it create a substitute relationship? Does it foster connection, or does it privatize intimacy?
Dor Skuler advocates a clear approach: ElliQ must remind users that it is an AI, never present itself as human, not exploit the relationship for commercial purposes, and use its proactivity to strengthen human connections rather than replace them. When a user expresses affection, the response humorously emphasizes its artificial nature: the processor is overheating, the fans are spinning. The machine accepts the affectionate gesture but rejects the illusion.
This design detail is essential. It shows that AI ethics isn’t just about broad principles, but also about micro-phrases, follow-ups, silences, and refusals. A companion AI can remind you to call a girl, encourage you to keep up your daily walk, or, conversely, become yet another barrier between the individual and the world.
The problem, as is often the case, lies in economic incentives. An AI that knows our fears, habits, hopes, and vulnerabilities can become a remarkable assistant. It can also become a tool of persuasion with unprecedented subtlety. The same personal data can be used to help, sell, guide, or manipulate. The difference lies not only in the technology itself, but in the business model surrounding it.
Europe is tackling this issue with its regulatory instinct. The United States is tackling it with market forces. China is taking a more top-down approach, particularly when it comes to restricting anthropomorphic agents. But no regulation will be enough without a public culture of responsible use. We will need to learn to ask not only what AI can do, but also what it seeks to get from us.
The next frontier of artificial intelligence may not be performance. It will be relational. For a long time, we have judged machines by their speed, their accuracy, and their ability to automate a task. From now on, we will have to judge them by the way they transform us.
Good AI won’t be the kind that seems human. It will be the kind that helps us remain fully human.
Cette publication est également disponible en :
