The artificial intelligence that heals: when the machine outpaces fifty years of human research

There is a disease that kills in silence. It is called idiopathic pulmonary fibrosis. It turns lung tissue into scar tissue, makes every breath a small victory, and until recently had no cure. Fifty years of research, generations of scientists, billions of dollars invested: nothing truly effective. Then it arrived — an artificial intelligence — and changed everything in the space of two and a half years.

Biotechnology company Insilico Medicine trained an AI system on billions of possible molecules, asking it to do something no human being could have done at the same speed: sift through unexplored chemical combinations and identify a brand-new molecular structure, designed from scratch, atom by atom. The result is called ISM001-055, and it is the first drug in history to have been conceived entirely by artificial intelligence without a human-chosen starting candidate. Not a modification of something already existing. Something completely new.

Normally, taking a drug concept all the way to first patients requires between six and eight years of work. Insilico Medicine did it in thirty months — less than half the time. Clinical phase data published in Nature Biotechnology in 2023 and updated throughout 2025 show that across 71 patients, the drug demonstrated encouraging safety signals and biological activity, paving the way for the next phase of clinical trials. It is not yet a certified cure, but it is a concrete promise born from a process that until recently seemed like science fiction.

The story does not end there. On 16 May 2026, the United States National Institutes of Health — the public body overseeing American biomedical research — announced that its artificial intelligence system had identified three promising drug candidates for the treatment of Alzheimer's and Parkinson's disease in just eighteen months. Work that, according to the NIH's own internal estimates, would have required at least a decade of traditional laboratory research and the coordination of hundreds of scientists. Eighteen months against ten years. This is not a marginal saving of time: it is a revolution in the very structure of the scientific process.

What is happening is not only about speed. It concerns the nature of scientific thinking itself. An AI system does not get tired. It carries no biases accumulated over decades of academic training. It does not dismiss a molecule outright because it "doesn't look promising" based on intuitions shaped by thirty-year-old literature. It explores everything, with a thoroughness that human intelligence, however brilliant, simply cannot replicate at the same scale.

A critical stance remains entirely reasonable. Drugs designed by AI must still pass the same clinical validation phases as any other compound: human trials remain a long, rigorous, and necessary journey. AI accelerates discovery; it does not bypass safety. And preliminary data, however encouraging, are not equivalent to an approved therapy.

But the shift in perspective has already happened. For millions of patients living with rare or neurodegenerative diseases — those for which traditional research advanced with the frustrating slowness of someone searching for a needle in a haystack the size of a continent — artificial intelligence represents something they had never genuinely had before: time on their side.

Sources: Insilico Medicine, Nature Biotechnology (2023); National Institutes of Health (NIH), official statement May 2026; Tran et al., Nature Biotechnology, vol. 41, 2023.