Knowing only AI means not really knowing AI

Knowing only AI means not really knowing AI

“Those who only know Medicine, don’t even know Medicine". 

This sentence attributed to Abel Salazar (Portuguese physician, scientist, and artist who became significant for modernizing medical science in Portugal and linking science with humanist culture) this sentence that I found engraved on a wall of University of Porto, sounds paradoxical. 

Yet its meaning is disarmingly simple. No discipline can be understood in isolation. When knowledge turns inward and becomes self-referential, it loses contact with reality.

This idea extends of course beyond medicine. It applies to any field that mistakes technical mastery for full understanding. Expertise that ignores context, social, ethical, historical, human, becomes brittle and insignificant. It may be precise, but it is not wise.

AI is a clear example.

To know AI is not only to know models, architectures, benchmarks, or optimization techniques. Someone who knows only these things may build systems that are impressive yet blind to reality. Systems that function perfectly while failing the people they affect. AI systems do not exist in a vacuum: they shape decisions, distribute power, and redefine what counts as knowledge, authority, and even responsibility.

The quote’s second line (“I have no political ambitions, nor will I ever have them, as I never did; but I do have social duties to fulfill, which I will fulfill in accordance with the dictates of a scientific outlook.”)  sharpens the point. It stress social obligation as part of the scientific role, drawing a careful distinction between seeking power and accepting responsibility. One can, and even must, work from a social perspective without turning science into ideology.

For AI researchers and practitioners, this means acknowledging that neutrality is not an option. Every dataset, design choice, and deployment context reflects assumptions about the world. Ignoring this does not make AI apolitical; it simply makes its politics unexamined.

Working “according to the dictates of a scientific outlook” today requires more than technical rigor. It requires reflexivity: awareness of impact, openness to critique, and engagement with disciplines that study society, values, and governance. Ethics is not an add-on. Social responsibility is not a distraction. They are part of what it means to do the work well.

Knowing only AI is not enough.
To truly know AI, one must also know the world it enters, and the world it helps to shape.