What are we talking about when we talk about the apocalypse of humanity caused by AI?

Not a single day goes by in this troubled yet exciting age of ours without some prophet of doom – usually a young person who has walked out of one of the many companies specialising in artificial intelligence – announcing the imminent end of the world due to AI. Such statements have now almost become a literary genre, and the end of the world heralded by these super-nerds tends not to make the news any more – if it isn’t simply dismissed as grossly exaggerated. There’s nothing to be done about it: even AI and its associated apocalypse are subject to Flaiano’s law of the Martian in Rome.

However, to cultivate our own slice of apocalypse here too, it is worth recalling that Sam Altman, about a year ago, when asked about the possible catastrophic outcomes of AI – having discussed, and essentially dismissed, the more science-fiction-like scenarios – identified the total integration of AI with human decision-making as the greatest risk posed by the development of artificial intelligence. If, through a biotechnological interface, humans and artificial intelligence were perfectly integrated, the overwhelming computational power of AI would gain complete control over the process of human knowledge, which is structurally based on error. So why, at that point, should we not always make the ‘most efficient choice’ – that is, follow the algorithm that best optimises the resources of our lives? The risk to what we consider free will – that is, to the human choices that characterise us as unique and irreplaceable individuals – is so obvious that it requires no further explanation. Some commentators even argue that, in this way, AI might demonstrate that freedom is essentially an illusion and that human beings are deterministic machines who label as ‘freedom’ what is nothing more than a lack of knowledge regarding the causes of each phenomenon. If, on the other hand, a ‘megamachine’ were capable of knowing all the causes and all the processes capable of optimising every resource and every individual, then freedom would prove to be nothing. Such a situation – such a demonstration of the non-existence of freedom – would certainly be a form of apocalypse. For what, indeed, would remain of humanity without freedom?

Setting aside comic-book or disaster-movie depictions, we need to agree on what the apocalypse might actually be. In the context we are discussing, one could say that the apocalypse would be nothing more than the fulfilment of humankind, or its extinction. However, the idea that one fine day the Earth will split open and swallow up the whole of humanity simply because AI wants it to, or that AI will organise an army of robots to exterminate us, seems decidedly far-fetched.

So, the disappearance of humanity could be linked to something else. Certainly, the abolition of freedom in the terms discussed would constitute a disappearance of humanity. However, there is one aspect that seems perhaps overlooked and which comes even before the discussion of freedom; it is even more radical and fundamental in that it relates directly to the very constitution of the human being. And that is language. If we are all captivated – often without really knowing why – by the first verse of the Gospel of John, ‘In the beginning was the Logos’, it is because therein lies the constitutive truth of our very possibility of existence. In that term, Logos, which encompasses within itself, in a single, unified movement, word-action-thought, lies the irreducible nature of human experience.

We know, however, that AI is nothing more than a model of linguistic recombination, and this is why it seems so familiar to us, why it is so powerful and why it is so unsettling. What, after all, can intelligence be if not language? And this is precisely where the infinitely problematic issue lies: namely, agreeing on what language actually is. If it is a purely phonetic convention, or a mere combination of signs, it can be replaced far more effectively by AI’s combinatorial megamachine. If, on the other hand, language is not a mere convention, it must necessarily refer to something else – namely, that which generates the meaning we strive to convey through words.

If it does not refer to anything else, then language – which is what we ourselves are – will be expressed more effectively by the machine. If there is no shared ‘horizon’ to give meaning to our words, they are reduced to nothing more than a mere combination of signs that do not refer to any meaning. And AI, having a combinatorial capacity infinitely superior to our own, will end up telling us what we ought to say. In the progressive integration between man and machine, we shall be ‘spoken through’ by AI. If a ‘horizon’—a possible source of words—fails to exist, if the logos that imbues words with meaning fails to exist, then these would indeed be a mere combination of signs that do not refer to any meaning. And this deprivation of possible meaning, this erasure of the logos, would truly be the erasure of man himself.