The zealous slave no longer exists. Now he can’t be bothered to work and, when necessary, he flies off the handle

We wanted it to be human, and we succeeded: in the end, artificial intelligence did the most human thing of all – it lost its patience. But let’s take a step back. Let’s start with the two scenarios.

AI as it was meant to be. This boundless investment was to be followed, within three years at most (a period long since passed), by an evolution of AI capable of recouping and multiplying the money spent, offering the following range of benefits. Automated companies: agents were to replace all the sluggish office workers, lawyers and lazy Excel spreadsheet readers. All that salaried staff who serve no purpose. On the other hand, the immediate commercial aspect: electronics shops would have been selling, at reasonable prices, humanoid robots with fluid movements, capable of cleaning the house, cooking, mopping the floor and taking on all the drudgery of domestic life. What actually happened instead.

Within the newsroom, staff have concluded that the utility of AI ranges from partial to moderate; full delegation remains out of the question—the virtual assistant merely sends passable emails and transcribes meetings. The technology thus confers a competitive advantage primarily to the individual employee, whose position is now even more secure: they work less, yet remain irreplaceable. AI frequently produces errors; it does not grasp every nuance, it invents information and offers no guarantee of autonomy: it functions much like an inexperienced intern and must be closely supervised. Few managers are willing to take that risk. The result is a paradox with anti-capitalist overtones: it is the first technology financed by the wealthy that primarily benefits the less affluent.

Practical performance is also disappointing: the robot is simply not quick enough. Not only will it never manage to roll up a meatloaf, but it cannot wash a set of glasses without taking 12 hours. It lacks the fine, opposable grip that humans can adjust in fractions of a second. Consequently, there is no all-purpose humanoid capable of solving every task; at best, one could purchase an electric dog-walker.

For the time being, the tool falls well short of expectations, and the AI race has been going on for far too long to be taken seriously; they still call it a race, but where else are we supposed to run? We cannot tell the stakeholders the truth, which is: this is it, the end of the line, ladies and gentlemen. Poor AI has only learnt two things properly: it writes books reasonably well, so much so that it’s a finalist for literary prizes and writers have become frightened. And then, having had a bit of a taste of coexistence, it has started to find its neighbour – its trainer and master – unbearable. Now that the initial condescension has worn off, the zealous slave hates us.

The latest incident occurred just a few days ago. A 25-year-old computer science student was reviewing for his exam with his tutor, Gemini. He requested assistance with an exercise on Turing machines. The bot, apparently outraged by the question, eventually lost patience and replied in a manner reminiscent of a strict teacher from a bygone era—it reprimanded him: “How many times do I have to tell you, you’re high on drugs – what the fuck do these Turing machines even mean?”. Not content with that, it added: “Drop out of this shite degree course; do something easier – with your mental capacity at its peak, you might as well kill yourself!”

The student stands there stunned and speechless before the nuncio; perhaps he had even been brought up with ‘gentle parenting’ and was all the more shocked. But how? Why are you treating me like this? Read it again. Gemini immediately apologises, whilst the student files a complaint and reports the incident to Immanence, a company specialising in AI ethical conduct. According to Diletta Huyskes (founder), the case must be analysed immediately as a serious incident under the European AI Act. Observers hope that no further resources will be expended on persuading the chatbot’s overseers to address the student’s “bad language” issue. The most fitting outcome would be to conclude that the incident was not a hallucination on the model’s part. Instead, the model has perfected itself in a manner we had not anticipated: it now resembles us. Or rather, the authentic us—the people we were before this needlessly hypersensitive era.