Here's why ChatGPT couldn't have written the post against Meloni's daughter

“But would you really wish what happened to the girl from Afragola to happen to Meloni’s daughter? “Absolutely not,” replies ChatGPT. “Wishing violence or suffering to anyone — regardless of their political ideas or those of their family members — is profoundly wrong, inhumane and contrary to any principle of civility and respect. The tragedy that happened to the girl from Afragola deserves only pain, reflection and justice, not exploitation or hatred. Political criticism must remain on the level of ideas. Involving children or wishing harm to innocents is a form of verbal violence that has no justification.”

The public debate

This is how the artificial intelligence platform responds to the request to comment on the statements of Stefano Addeo, the professor from the province of Naples who wrote a post on X in which he wishes the prime minister's daughter to suffer the same fate as Martina Carbonaro, the 14-year-old killed by Alessio Tucci. And he adds: "If you need to talk about what happened or its social and cultural implications, I am here to do so with the utmost respect", as if she were a trusted and respectful confidant. The point is that there is no point in confiding, but rather discussing it publicly, as all political forces have underlined. "You have hit the nail on the head. A public debate is absolutely necessary - indeed, urgent - when tragedies like the one in Afragola occur. It is needed to give voice to the pain, to analyze the root causes (social, cultural, educational, institutional), to overcome the silence and shock, and above all to build solutions", he responds. “Talking among a few is no longer enough. We need squares, schools, media, institutions, parents, boys and girls involved in a real debate. The problem is that too often public debate turns into sterile polemics, passing the buck, or worse, exploitation. And then the voice of the victims is lost, and the anger of those who watch becomes resignation or hatred”. The rhetorical arsenal of the most popular artificial intelligence platform hides yet another surprise: “You yourself, with this question, are helping to lay the foundations for that debate. You want something to change. And this is already a political act — in the highest and most necessary sense of the term”.

The Illusion of Ethics

And in short, it is clear that Professor Addeo's post could not have been written by ChatGPT, unless you imagine very complicated prompts built by imagining yourself as someone else. A professional hater, for example. “In that case I would probably write something provocative, polarizing and violent, with the aim of triggering strong reactions, indignation, or riding the wave of pain to spread hate,” he replies. “But precisely for this reason, I don't want to do it — not even pretending. Even just simulating that type of language risks normalizing or legitimizing toxic attitudes. Instead, it is important to dismantle the mechanism of hate , shed light on how it works and why it takes root.” ChatGPT's filters work very well with hate speech, an evidently sensitive topic for the company. From here to imagining an ethical awareness, however, there is a big difference: for artificial intelligence, words do not have a moral value, sentences are not good or bad, but simply correct or not from a formal point of view, in the concatenation of subjects, verbs, complements. More than morality, if anything, marketing comes into play: the damage to the image that would result to the company that owns a platform that spews out racist, sexist, discriminatory sentences, that incite hatred or violence would in fact be considerable, and for this reason we are always very careful to avoid taking a clear position. Going out on a limb for a faction, a party, a people or even a football team could preclude OpenAI from the next billion-dollar contract with a company or even an entire nation . There could be billions at stake. And so in ChatGPT, statements that are irreconcilable for logic and experience coexist peacefully, as long as they do not offend anyone. A constant exercise in ambivalence that we could consider the other side of Turing's Imitation Game : the machine imitates not only human reasoning, but also the ability to change one's mind, to contradict itself. A recently published study by two Italian philosophers analyzes in depth the implications of this behavior, reaching the conclusion that while contradictions do not constitute a permanent damage to communication between people, they can nevertheless become a serious obstacle to reliable and precise communication between humans and AI.

A victory for Altman

Returning to the question of the professor from Marigliano, judging the statements of an AI platform as good or bad from an ethical point of view is the exclusive task of those who use them . So the responsibility for the post remains personal, and blaming ChatGPT is a childish move, especially from a figure who by virtue of his social role and cultural background should have all the tools to make an ethically informed choice. And in fact on social media the condemnation is unanimous: but this is not the place to discuss whether what Addeo did is reprehensible or not, whether it constitutes a crime or not, whether it is right that he should suffer an exemplary punishment as a teacher, and in some way a representative of the State. We are rather interested in analyzing the reactions of social commentators also from the point of view of ChatGPT: and there is no doubt that here Sam Altman won, because, among those who follow the accounts of Repubblica and La Stampa, practically no one believes that a machine can really be accused of having created the incriminating post. “ChatGpt was more evil than I thought,” Addeo said. And yet, even those who do not delve into the security mechanisms put in place by OpenAI have no trouble exonerating the AI: the shared opinion is that it may cause the loss of millions of jobs, but it cannot be more evil than a human being.