The workshop and the machine

There is an image that comes back to me when I think of artificial intelligence: the craftsman’s workshop. The worn bench, the neatly arranged tools, hands that know the material better than any technical drawing. For centuries work was chiefly this: knowledge that lived in gestures.

Then came machines. First those directly controlled by people, then increasingly autonomous tools, able to perform precisely what had previously required years of apprenticeship. The craftsman did not disappear. His role changed: from the person doing the work to the person thinking it through, setting up the machine, checking the result and correcting mistakes.

Something similar is happening with computing.

From craftsman to programmer

For decades programmers have been the craftsmen of our time. They wrote code line by line, knew languages, architectures and logic almost physically.

Today a growing share of that work is passing to artificial intelligence, which can generate code, correct it and suggest solutions. The programmer does not disappear: increasingly he is less the physical author of every line and more the person who defines the problem, guides the system, checks the result and takes responsibility for it.

It will be objected that industrial machines carried out precise instructions, whereas artificial intelligence interprets and can make mistakes. True. But the first machines, too, produced waste, jammed and required constant supervision. Every technology initially seems too imperfect to replace what came before it.

The industrial revolution was not completed in a few years. Factories continue to evolve even today. It would therefore be naive to judge present-day artificial intelligence as a finished work. We are at the beginning, and beginnings are imperfect.

The machine that helps build the machine

Here, however, the analogy with the past breaks down.

The lathe never designed a better lathe. The loom never rewrote its own operation. Artificial intelligence does not yet improve itself entirely on its own, but it can already take part in the process by which better systems are built: it generates code, suggests algorithms, runs tests, compares results and develops the most effective solutions.

For the first time, the machine does not merely execute a human design. It is beginning to collaborate in designing the machines that will follow.

This is not yet full autonomy. But it is no longer merely automation.

This also changes the nature of knowledge. If work will no longer consist solely in knowing how to do things, but in knowing how to question, direct and judge a machine, what should we teach our children? And what value will a work made entirely by human beings have tomorrow?

Learn it before it happens to you

One thing, however, is already clear: artificial intelligence is not a skill reserved for computer scientists. It will enter almost every trade: surgery and agriculture, workshops and schools, professional practices, courts, shops and public offices.

It will not replace everyone or make experience useless. On the contrary, it will be more useful the more it is entrusted to people who can recognise an error, frame the right question and distinguish a plausible result from a reliable one.

Learning it is therefore no technological affectation. It is a new form of professional literacy. Everyone will have to understand how it can change their work, which tasks it can speed up, which risks it introduces and which decisions must not be delegated to it.

Those who do not learn to use it will not necessarily be replaced by a machine. They are more likely to be overtaken by someone in the same trade who has learnt to use it better.

The craftsman who came through the machine revolution was neither the one who rejected machines nor the one who surrendered to them. It was the one who learnt to control them, preserving what no automation could guarantee: a sense of work well done.

That is the task before us. Not defending the gesture, but preserving the purpose. Not worshipping the machine, but learning to direct it.

The workshop is not closing. It is evolving and updating its ATECO code [Italy’s classification of economic activities].