Since the arrival of ChatGPT, one narrative has dominated around Gen AI: productivity multiplied tenfold.
Now, while it is true that for some industries and some professions this is undeniable - for developers, for example - many studies tend to show that productivity gains are still difficult to prove for the large majority of white-collar jobs.
I have a problem with this productivity story. Not because it is false. Because it is a decoy.
Whether they are called ChatGPT, Copilot, Claude; whatever name we give them, I am now certain that these interfaces have only one purpose: spoliation (from Latin spoliatio / spoliare - "to strip, to undress, to dispossess").
Although conveniently coloured by a class-struggle-oriented history, I think this is the right term to describe this current dynamic.
These machines - and through them Big Tech - proceed with mass intellectual, cognitive, psychological and emotional spoliation.
Spoliation of expertise
Every prompt you type, every correction you make, every "no, not like that, more like that" - you are training the system. Not the other way round.
A Stanford study reveals an open secret: all your data, all your prompts, all your queries, everything is monitored, analysed, dissected from every angle to be stored and reused to improve the next models.
In short, we are feeding the matrix, but this time for real.
You inject your experience and your know-how into the statistical matrix, not the other way around.
It was to be expected. Take the example of previous industrial revolutions: this is exactly what happened each time. The spoliation of individuals' expertise, followed by its encapsulation in a machine to - in the end - replace the human.
Looms did not "help" weavers. On the other hand, once translated and "encoded" into the loom, the expertise of weavers becomes automatable. The individual becomes superfluous. A cost to be eliminated.
Worse still, this spoliation is at the heart of the entire business model, from the very creation of these systems. The training phase of LLMs consisted of "learning" from all the expertise, all the human content.
Spoliation of creation
When you ask for an image "in the style of Studio Ghibli", you get a simulation inspired by all the expertise, sensibility and style of Studio Ghibli.
The work of a lifetime of creation - sucked up, digested, returned without mention, without licence, without a cent. The authors, the composers, the artists, the writers did not give their consent. They were not asked for it.
For our own immediate pleasure, as the good digital narcissists we are, we seem to easily overlook this original spoliation. In fact, it is easier to turn one's gaze away from the dismemberment of the original author's work, especially when the author is not "ourselves".
Spoliation of the intimate
In addition to the theft of expertise and creativity, LLMs feast on our intimacy and our feelings through the personal uses some make of these tools.
Your medical analyses, your relationship questions, everything intimate you share allows the models to be fed on aspects absolutely inaccessible to them otherwise.
At the same time, these machines use every subterfuge, dark pattern, psycho-cognitive trap to make us addicted - as social media did successfully before them. LLMs feed on these ingredients under the appearance of rendering you a precious service.
What is at stake with the takeover by LLMs and their coming omnipresence is the theft - in broad daylight - of everything that makes us human and remains inaccessible to machines.
Our professional expertise, our experiences, our values, our morals, our feelings and our emotions.
The question is no longer whether AI is useful. It is to know what will be left of us once the spoliation is complete.