About
There is one idea I cannot shake. A technology that thinks in our place does not free us from anything; it replaces us. The distinction may seem slight, but it changes everything. A calculator has never stopped me from understanding subtraction. A machine that sorts through my arguments, writes my answer and reaches a conclusion before I do is something else entirely. What I expect from it is that it sharpen my judgement, not spare me the need to exercise it.
My first computer was not mine. It stood in the school library, connected to a dot-matrix printer whose clatter drowned out every conversation. No one paid it any attention. There was the librarian and me, and that was about it. Forty years later, everyone is looking at the machine and almost no one knows what it is anymore. The gap has not disappeared; it has simply reversed direction.
The next one, I soldered together myself. A ZX80 kit that had to be assembled before I earned the right to use it. One kilobyte of RAM, shared between the program and the display, with programs saved onto audio cassettes. No one ever needed to explain to me that you do not delegate to something whose nature you do not understand. I had the soldering iron in my hand. At university, programs lived on punched cards. One mistake meant punching the card again. Drop a deck on the floor and you had to put it back in order by hand, one card at a time. Writing code was so costly that you thought before you began. The cost has since vanished, and the discipline went with it. Today, we have to rebuild that discipline deliberately.
I also spent a year writing algorithms on paper before I was allowed access to a machine. I hated it. I nearly gave up altogether, and I would have walked away had I not been able to program at home in the evenings. Looking back, nothing trained me better, because there is no trial and error on a sheet of paper. You have to see the whole before writing the first line. I do not tell this story to glorify hardship for its own sake. The same question now confronts everyone learning alongside a machine that already appears to know everything.
The rest followed. In 1996, I joined a fledgling telecoms operator to build an internet service provider from scratch, at a time when the entire country regarded the internet as a passing fad and the incumbent operator still swore by Minitel. The sceptics were not fools. That is precisely the problem. They reasoned from what they already had, and what they had worked perfectly well. The same mechanism is unfolding before us today, blinding the enthusiasts just as surely as the cautious. Then came California and cybersecurity, where twenty-five years ago I helped build a system that continuously explored environments on my behalf. So I am not observing from the outside the shift I write about here. There was also one unmitigated fiasco: a green-energy venture based on anaerobic digestion. The technology was ready; the market was not. Launched today, with the appetite of data centres, it would probably work. Being right too early amounts to exactly the same thing as being wrong, and I did not learn that from a book.
I hold PhDs in mathematics and artificial intelligence, and today I design systems that wield power, particularly in healthcare. It is a field in which imprecision carries consequences. At the end of the chain is a patient, a doctor who must decide, and a responsibility no algorithm will ever bear in their place. The system I am working on replaces no physician. It connects an isolated general practitioner with a specialist who, without it, might never have spoken to them. I ask little more of a machine than this ability to bring people together instead of making them disappear.
And then the same thing happened to me that happens to everyone. One evening, in a hurry, I asked the first version of ChatGPT for a chocolate cake recipe. I checked nothing; the stakes were nonexistent. I followed every instruction to the letter, all the way to the end, without once asking whether any of it made sense. The cake ended up in the bin. The text was clean, orderly and perfectly plausible, which is exactly why I never grew suspicious. Twenty years in the field immunises you against nothing. The fluency of an answer has never told us whether it is correct, and it cost me four eggs to learn that lesson.
That is why I write here. The public debate on AI swings between two forms of intellectual laziness. On one side, manufactured wonder, selling dreams by the kilo. On the other, panic, selling anxiety at the same price. Both offer the same convenience: they spare us the effort of looking. So I step through the digital looking glass and examine what these machines are actually doing to our lives, our schools, our businesses and our choices. I include my sources, my doubts and, when they occur, my mistakes. I am not trying to persuade you. I am trying to give you what you need to form your own view. The real break will not come from the machines. It will come from the clarity with which we choose to use them.
My column, “AI in All Its States!”, is available here. It is also available as a podcast for those who prefer to listen.
📎 For everything else, find me on LinkedIn: Philippe Buschini.