QWhat makes a good colleague? There are undoubtedly multiple answers to this question. Someone you can trust. Who does his job conscientiously. In a good mood. Arranging. Tolerant. Inspiring. Funny. Ready to help you out with a Post-it or calmly explain how to make a double-sided A3 print. To all these possible definitions, we can add another: when the time comes to migrate to the coffee machine, a good colleague is the one who will be able to share with you the juicy information constituting the vibrant heart of the life of the company. For the primatologist and anthropologist Robin Dunbar, gossip is not this demeaning activity that we often describe, it has the same unifying function as delousing among monkeys, allowing us to build bonds, to strengthen the feeling of a common destiny.
While artificial intelligences (AI) are in the process of integrating daily office life, we can legitimately wonder if, beyond their productive appeal, they could also engage in this form of corporate slander which, in the face of a cup of sock juice, makes you transform the lives of your colleagues into an epic song of gesture (“Nââân, isn’t that true? ! »). I’m trying the experiment with the free version of a generative AI: “Can you imagine corporate gossip being generated by an AI colleague at the coffee machine, with juicy revelations about Robert from accounting? » In a fraction of a second (the time of the response), the language model sends me a dismissal and I realize that the AI is not quite a colleague like the others, quick to reveling in the marital setbacks of said Robert, with a carnivorous smile at the corner of his lip: “Sorry, I cannot generate gossip or false information about people, even fictitious ones. If you have any other questions or need help on different topics, don’t hesitate to ask me! »
The sayable and the unspeakable
But why ? “ I strive to promote respectful and informative interactions, the AI tells me. Generating gossip or false information about individuals defeats this purpose. However, I am here to answer any other questions you may have or to discuss more appropriate topics. » If it is clear that all intelligences have biases, that of AI could be summed up as a potentially contagious asepsis of forms of discourse. By using these major language models, we are not only using a tool for intellectual productivity, we are part of a dynamic where a boundary is drawn between the sayable and the unspeakable, the thinkable and the unthinkable. We are invited, finally, to become moral by designas engineers say, that is to say subtly led to adopt the right attitudes, the ways of thinking deemed adequate.
You have 14.77% of this article left to read. The rest is reserved for subscribers.