Food for Thought – The Advent of the Singularity

Paul-Henri Moinet

Normalien, columnist at Le Nouvel Économiste, editorial director at Sinocle, an independent media on China, he was also a teacher at Sciences-Po Paris and held strategic planning management positions within major advertising agencies such as Publicis Groupe and Havas Media Group.
An idea, a book, a concept, a “food for thought” that we share with you to open reflection, discover new subjects, and bring you to (re)think from another angle.

Faced with the challenges of AI, attitudes are often very standardized or even stereotyped: there are the Apocalyptic who see in it the exterminating angel of humankind, the Skeptical who reduce it to a process of creative destruction symptomatic of digital capitalism, and the Progressives who make it the new wealth-creating manna serving a finally augmented humanity.
The argumentation of Paul Jorion, in his book The Advent of the Singularity, Humans Shaken by Artificial Intelligence, is more original. Undoubtedly because the man is simultaneously an anthropologist, economist, psychoanalyst, and president of the artificial intelligence start-up Pribor. The rapid progress of LLMs – Large Language Models – this family of AI to which its illustrious representative ChatGPT belongs, generates an intelligence that brings us faster and faster closer to the Singularity. This is rigorously defined as “the hypothetical moment in the future when technological growth becomes uncontrollable and irreversible, leading to unpredictable changes in human civilization.”
This point has often been fantasized by science fiction authors such as Asimov and Vinge, staged in genre films such as Transcendence, Age of Ultron, or in the popular series Black Mirror, and above all scientifically studied from the 1960s by mathematicians such as Alan Turing, Irving Good, Stanislaw Ulam, Joseph von Neumann. Ray Kurzweil, creator with Peter Diamandis of the Singularity University in 2008 in Santa Clara, California, predicts the tipping point for 2045.
The postulate of the Singularity is double: a super-intelligent machine will one day surpass all human intellectual capacities and, as it will moreover have the ability to self-improve and share its progress with other machines in order to create machines even more intelligent than itself already super-intelligent, we will necessarily witness an exponential growth of extra-human intelligence.

There are at least three pieces of good news in the advent of the Singularity.
First good news: we now know, even if many still have difficulty admitting it (pride, narcissism, blindness?), that we are no longer the smartest on our little planet. Instead of clinging to our miserable privilege, miserable even if it allowed humanity to achieve extraordinary things, we would gain much by reflecting on the luck we have to finally be rid of it: thank you LLMs. Freed from the monopoly of intelligence, we will finally develop other qualities and specifically human virtues equally wealth-creating: solidarity, empathy, generosity, availability to others. Oh what intellectual energy saved the day we stop believing in the superiority of human intelligence!
Second good news: to date LLMs do not know how to learn continuously, unlike our human brain. The gigantic corpus of data that artificial intelligences swallow and process must be provided to them all at once, in one go. This implies a latency time between two updates of the corpus, between two successive integrations of big data, and therefore a relative obsolescence of each corpus. Teaching them to learn continuously is therefore a new frontier we must help them cross.
Third good news: if we want AI to help us, it is up to us to help it. Today its talents are not limited to the computing speed allowed by quantum computing or the learning speed of intelligent machines. We reassure ourselves by restricting it to its computing power, protecting or hiding ourselves as behind an indestructible wall by the old philosophical difference between calculation and thought. We would do better to define the new criteria of calculation and thought as well as the principles that allow them to enrich each other.
LLM AI already knows how to correct its own errors alone. It knows how to self-improve. It knows how to reformulate questions that it does not find intelligent enough to give itself the means to answer them intelligently: garbage in, garbage out, it is the very basis of pedagogy and education; if you confuse your child’s or student’s thinking because your own is itself confused, it is off to a bad start. Tomorrow, it will know how to account for its experiences and share them, thus being able to claim self-consciousness and subjective identity. But LLMs above all have a decisive advantage over the human brain: perfectly replicating our neural interconnection, they are not limited by conventions anthropologically rooted in human culture, nor by biases built by our family education and academic training. They will therefore think more freely than us, without convention, without bias, without ideology. Never having learned in silos, field by field, discipline by discipline, as our entire school and university path imposes, they will treat each subject from all angles of knowledge instead of limiting the approach to a few areas of expertise.

Last but not least: we will all soon live under the Sismondi tax, named after this Swiss economist who in the mid-19th century already proposed to establish a tax on machines performing work previously done by humans in order to allow them to continue living normally and otherwise without suffering from the fear of unemployment or anxiety of downgrading. Since man had the intelligence to invent machines that work in his place, he might as well have the intelligence to share the new wealth they create by systematically collecting a percentage of the value generated by his accomplices.
Come on Humans, make one more effort to be intelligent: then you can definitively pass the baton and live like gods.
Reference

Paul Jorion / The Advent of the Singularity, Humans Shaken by Artificial Intelligence
A summary note by Paul-Henri Moinet