The recent International AI Summit in Paris (February 6-12, 2025) was truly eye-opening, allowing me to witness both the excitement and the magnitude of what's at stake. The global race is now in full swing, with a "big four" leading the pack: the United States with its $500 billion "Stargate" plan, China with its DeepSeek and Qwen2.5 models, Europe through the EU AI Champions Initiative and its €200 billion investment, and France, whose €109 billion demonstrates its ambition to play a major role (see my recent article in Hérault Tribune).
We are at a civilizational tipping point. On one side, investment surges are flooding into AI - the global market is expected to exceed $1.5 trillion by 2030, according to Goldman Sachs. On the other, a nagging question remains: are we collectively capable of anticipating the socio-economic upheavals that lie ahead?
This paradox, it seems to me, stems from a tension: our genius at inventing disruptive technologies outpaces our ability to foresee their consequences. We must examine this dynamic with clear-eyed realism, or else the promises of a bright future could transform into a source of inequalities, social imbalances, and ultimately, a nightmare. This conviction drives me as I fly over the North Atlantic, heading to New York for a major AI Engineering conference. These constant travels, I believe, provide a broader perspective on these challenges.
The numbers are chilling. A recent Oxford study reveals that 82% of jobs could be automated by 2035. This is especially worrying when only 12% of companies seem to be preparing their employees for this transition. It's high time we questioned the consequences of such disparity. And while AI undeniably creates wealth, its distribution remains a burning issue. Brookings projects that 57% of the value created by this technological revolution could concentrate in the hands of 0.1% of the population. Such concentration, if confirmed, would pose serious problems for social cohesion, further increasing the Gini coefficient - a measure of economic inequality already at worrisome levels in many countries.
We have seen similar dynamics during previous industrial revolutions. Consider the first Industrial Revolution in England (1760-1840), when textile factory owners amassed colossal fortunes while workers' living conditions deteriorated, giving rise to what historians called the "social question" of the 19th century. Yet today, the scale and speed of transformation demand constant vigilance. The automation of intelligence itself, looming on the horizon, could profoundly reshape the professional landscape, threatening professions once thought untouchable.
That is why it is imperative that France, drawing on its history and values, takes the lead in developing a humanistic AI - placing ethics and inclusion at the heart of its development, as championed at the Paris AI Summit.
How do we stay on course in this storm? The Moravec paradox reminds us that while AI can excel in specialized tasks, it still struggles with tasks that are simple for a child. Human intelligence is far more than mere computation; it encompasses intuition, empathy, creativity, and judgment. We must focus on harmonious human-machine collaboration rather than succumbing to a naively technocentric vision. The real question is not only technological but also ethical and social: what kind of society do we want to build for future generations - a world where a minority takes all, or a more just and inclusive one? The answer, to me, is self-evident.
The lack of planning for "the after" is particularly alarming. Neither businesses nor public authorities seem to grasp the magnitude of the impending upheavals. Fewer than 15% of companies are developing reskilling strategies for their employees, opting for a risky wait-and-see approach. Meanwhile, public authorities remain strangely passive. Our educational systems, bogged down by cosmetic reforms, struggle to keep pace with a rapidly evolving job market. We continue training individuals for jobs that will soon be obsolete, perpetuating an outdated, stagnant model. It is urgent that both governments and businesses finally tackle these challenges and fundamentally rethink our approach to education - fostering creativity, adaptability, resilience, and critical thinking. After all, it is our humanity - more than our technical skills - that will ultimately make the difference.
It is time to stop burying our heads in the sand and learn from the past. History teaches us that technological innovations can have unexpected and sometimes devastating consequences. Social media, initially seen as tools for connection and sharing, have also contributed to the spread of misinformation and societal polarization. Algorithmic trading, designed to optimize financial markets, has sometimes led to increased instability.
Facing these challenges, a promising approach keeps me awake at night: participatory modeling and prospective simulations. Imagine virtual laboratories where we could simulate, using computers, the interactions among millions of individuals, businesses, institutions, and even autonomous artificial intelligences - those "agents" generating so much interest today. These simulations, far from being simple deterministic predictions, would allow us to explore multiple future scenarios, identify critical tipping points, and test the effectiveness of various public policies before implementation. Platforms such as Cormas enable the creation of complex models that integrate rigorous scientific data and valuable local knowledge, offering a richer and more nuanced view of the challenges. The goal is no longer merely to predict the future, but to actively co-create it by involving citizens in a collective reflection on the fundamental choices ahead. These tools, still in their infancy, could be the key to anticipatory social engineering, capable of guiding us toward a future that is not only viable but also deeply harmonious and humanistic.
Let us go even further. If AI is on the verge of revolutionizing medicine - perhaps even promising a cure for all cancers in the near future - why couldn't we use it, in close collaboration with human intelligence, to co-create the new social models we so desperately need? Today, we have all the necessary ingredients: brilliant minds, sophisticated algorithms and models, massive amounts of data, ever-increasing computing power, and, of course, AI itself. It is no longer about passively enduring technological transformations, but about consciously steering them toward a desirable future. We must act now to prepare our landing ground, or we risk being forced into an emergency water landing without ever having defined our trajectory.
The race to AGI reveals a deep contradiction in our relationship with progress. While AI systems achieve superhuman performance in specialized domains, they still fail at tasks that are trivial for a child. This cognitive asymmetry should prompt us to question the very nature of the intelligence we aim to replicate. At the AI Summit, I was struck by the divergence between the American vision - focused on opportunities and competition - and the European approach, which emphasizes ethics and regulatory frameworks. Finding balance is crucial: over-regulation could stifle innovation and widen the gap between large corporations - whose societal concerns are often secondary - and more virtuous players. It sometimes feels as if our political decision-makers, mired in lengthy and complex processes, are lagging behind this revolution. We need new agility, an ability to question ourselves and make bold choices. In short, we need governance that matches the scale of these challenges. This does not mean succumbing to the siren calls of Trump-style or Musk-style techno-libertarianism, whose simplistic solutions risk crashing into the harsh realities of society.
All these questions are, in my view, at least as important (and winnable) as the mad rush for datacenters and ever-larger models.
As Hannah Arendt wrote in The Human Condition: "What threatens us is not technological progress, but forgetting to question the meaning of this progress." Sixty years later, this warning resonates with renewed urgency. The alternative is clear: either we ride the waves of technological change, or we risk being overwhelmed as we strive to shape a future in which humanity and machines evolve in symbiosis.
The time to choose is now.