While reading Dario Amodei's brilliant essay Machines of Loving Grace, I was struck by the originality of the ideas he puts forward. Even though I don't fully share his optimism—nor the radiant future he outlines—I must admit he offers a perspective on AI that is both scientifically ambitious and, in some ways, surprisingly grounded. He explores the conditions for a technological development that remains centered on the human factor—a refreshing contrast to those who present AGI as an inevitable step meant to overcome the limits of human intelligence.
Among the ideas that stood out to me: the true impact of AI doesn't lie in intelligence itself, but in the scale and speed at which it could support scientific research. Through an interesting analogy with the "factors of production" used in economics, Amodei introduces the idea of "marginal returns on intelligence" and confronts it with real-world frictions: complexity, data availability, reliability and, of course, physical laws.
Scientific progress often comes from a handful of breakthroughs that cut through this complexity. He acknowledges the mystery and brilliance of the human mind, arguing that it's not raw computational power that will trigger major leaps, but rather its alignment with the way we search for knowledge—possibly creating a virtuous cycle that could lead to the "compressed twenty-first century" he envisions.
These ideas echo the product vision of Anthropic, a company that doesn't just release a series of ever-larger models, but instead builds foundational tools aimed at enabling a broader shift toward agentic AI. A good example is their recent open-sourcing of the MCP protocol, whose elegance may well prove more impactful than many over-marketed iterations of existing models.
As I finished reading, I found myself wanting to believe in this thoughtful utopia, driven by a passion for science in its most beautiful form. Amodei's closing words describe AI as something of "transcendent beauty"—a striking sentence.
—
Just before reading his essay, I had gone through the long paper published by the Vatican on the relationship between artificial and human intelligence. Although the Catholic Church's voice has been relatively quiet in previous AI debates, this note delivers a thoughtful, well-documented and holistic reflection on AI—one that invites every reader, whatever their beliefs, to take a closer look "with a wisdom that is both ancient and new," as the title (Antiqua et Nova) suggests.
New, because recent advances in AI are reshaping the foundations of our societies—work, relationships, truth, personal responsibility—with unprecedented speed and convergence. At the same time, they give AI a human-like appearance, while keeping us further and further from how it actually works. We no longer see a large language model running, we just observe its output, unlike an engine. That said, Anthropic has done impressive work to "illuminate" emergent properties in its models, like a biologist examining living tissue to understand its structure. Amodei even suggests that AI might help us shed new light on the mysteries of our own brains.
Ancient, because the Vatican's text also places AI in the broader historical continuum of technological progress, calling attention to long-standing concerns: technocracy, Promethean utopia, functionalism, and digital reductionism. Each of these can lead to a loss of human dignity, especially for the most vulnerable. In contrast, the authors promote an "integral" vision of the human person: deeply relational, embodied and with a desire for truth and transcendence—a human depth no machine can access.
In the end, these two texts, though worlds apart, converge in surprising ways. Their dialogue seems to me both timely and fruitful.
In the tech world, the dominant narrative often leans toward a transhumanist vision, where even the reassuring image of the AI "co-pilot" gradually erodes human responsibility. As humans, we're left somewhere in a narrowing space, squeezed between the animal that claims his dignity, and the machine that claims his superiority. Amodei's essay offers a refreshing alternative: grounded, insightful, and genuinely inspiring.
In the spiritual sphere, religious thinking has long been dismissed as irrelevant to technological change. Shaped by centuries of uneasy dialogue between faith and reason, modern societies often distrust religion when it enters this field. Yet Antiqua et Nova offers one of the most complete and well-informed reflections I've seen on generative AI, evaluating its applications across many areas of life. It also acknowledges, without alarmism or idealism, the intrinsic beauty of this human achievement.
Perhaps that's what both texts share: the sense that AI marks a technological leap whose origins and implications may already lie beyond our full understanding.
From there, several paths open up.
Following the prophets of AGI, some will place their hopes in a bright technological future that promises to solve today's problems—as long as we step aside and let the co-pilot take over. As Pierre-Yves Gomez points out, this belief is a kind of modern-day obscurantism, not really supported by history, but fed by commercial narratives that push back the horizon of profitability. And perhaps this vision resonates so strongly because AI appears to fill a hypermaterialistic need for transcendence in secular societies.
Others will reject AI altogether. Etienne de Rocquigny describes this as a new kind of Luddism, echoing the nineteenth-century workers who rose up against the mechanization of their labor. Their resistance was brutally repressed. The fear is legitimate, and reality is already moving faster than fiction.
But the most dangerous response, and probably the most common, is indifference: the passive acceptance of AI that doesn't come from overestimating its power, but from underestimating our own. This becomes truly risky if we don't reinvest the time it saves into deeper thinking and more meaningful human relationships, both of which are already eroding in the digital age.
As for me, I believe this "technological moment" calls for even more reflection on what makes us truly human. Back in 1947, Georges Bernanos published La France contre les robots, a sharp and prophetic essay that remains remarkably relevant. One sentence, in particular, now shines with a new light:
"Each of these machines, in one way or another, adds to man's material power—that is, to his capacity for both good and evil. As he becomes stronger, more fearsome, it becomes all the more necessary that he also become better."
