By Anthony Hié, Chief Innovation & Digital Officer, and member of the Executive Committee, at Excelia.
Once perceived as a tool at the service of humans, artificial intelligence is now undergoing a subtle yet profound transformation. AI is no longer satisfied with assisting or automating; rather, it is actively harnessing humans as a resource for its own development. This role reversal is transforming and redefining the relationship between humans and machines.
Humans as fuel for learning
Modern AI systems rely on data gathered from our behaviour. Every interaction becomes a signal that can be exploited. Clicking, correcting, rephrasing or just hesitating generates information. Conversational models learn thanks to adjustments made by users, who become ‘invisible’ contributors to the system.
This phenomenon is reminiscent of the metaphor explored in The Matrix, a film in which humans are literally used as a source of energy to fuel machines, whilst remaining under the illusion that they are in control. Although the comparison is merely symbolic, it does, however, shed light on a current reality: our daily interactions online serve as a form of fuel for AI systems.
Unlike dystopian cinema, this is not a matter of physiological exploitation, but one of cognitive, behavioural, and even physical exploitation. Our data, our choices, and our corrections feed into systems that are constantly learning. Humans are therefore becoming a new kind of energy resource – one that is no longer biological, but informational.
From digital platforms to ‘Rent a Human’
The major digital platforms are prime examples: Netflix, Spotify and TikTok refine their algorithms based on user behaviour. Viewing time, implicit preferences, and drop-off rates become training data. As a result, usage is transformed into invisible labour.
The concept of ‘Rent a Human’ is evidence of this trend. What was once merely a theoretical idea has now become a real-world platform, as seen with RentAHuman.ai 1, which enables artificial intelligence agents to call on humans to carry out tasks in the physical world. These may include physical travel, gathering information, or actions requiring a human presence that AI cannot yet perform independently.
But these actions are not only useful in the short term. They also serve as training data for the systems. Every task completed enhances the AI’s future capabilities. Humans thus become a kind of temporary crutch, destined to be assimilated and then surpassed by the machine.
Will ‘world models’ learn about the world from humans?
This momentum takes on its full meaning when considering the concept of a world model 2. Unlike Large Language Models, which generate text from vast amounts of data, this concept refers to an AI’s ability to construct an internal representation of how the world works in order to understand, anticipate, and simulate situations. To achieve this, AI requires high-quality, contextualised and meaningful data.
Yet this data is generated by humans. Interactions enable complex elements such as intentions, social norms, and ambiguities to be captured. Each human contribution becomes a fragment of reality from which AI builds its understanding of the world.
A paradox then emerges: the ambition of AI autonomy still relies on a deep dependence on humans. To simulate the world, it must first learn from those who live in it—up to the point where it may eventually move beyond human-based models.
The changing role of humans in the face of AI
Humans are no longer merely users. They are becoming agents for training artificial intelligence systems. Through their interactions, they are helping build models capable of understanding and anticipating their own human behaviour.
As systems advance, humans tend to find themselves in a paradoxical position. They fuel, correct, and train machines that are becoming increasingly autonomous. Gradually, the role of humans is changing. From being masters of the technology, they are slipping into a support role or even becoming servants of the system they are helping to create.
This paradigm shift calls for a rethinking of the issue. It is no longer just a matter of knowing what AI can do for humans, but also of understanding how humans come to serve, often unwittingly, the intelligence they themselves have created.
Biography
Anthony Hié is Chief Innovation & Digital Officer, member of the Executive Committee, at Excelia.
Anthony Hié has a PhD in AI from EPF Engineering School, a Master in IS Audit & Design from the University of Lorraine and a Specialised Master in Information Systems Management from ESSEC Business School. He is also certified with ITIL3 and ITIL4 and holds a certification in Digital Marketing from the University of Illinois, USA. He is also the winner of the 2023 Trophy ‘Innovation & Digital Transformation’ in the category ‘Education’, awarded by the French magazine Solutions Numériques.