Let me start by apologising to those who seriously study and practise Zen. I am a computer scientist, and I have put the word “Zen” in the title of a book about language models. I understand that this might raise an eyebrow. If reading it raises the other one too, write to me: I will have something to learn.
The book grew out of a feeling I experienced while working with these models. Between a question and an answer, I noticed that the most interesting work also involved the way I sat in front of the screen. What was I asking? What had I forgotten to explain? And why could a well-written answer reassure me before I had even checked it?
It is very easy to pick up speed with a model. I write, receive, correct and try again. The result always seems one prompt away. Then I pause and discover that the question was unclear, that an important condition had stayed in my head, or that I was seeking confirmation of an idea I had already grown attached to. The machine had produced plenty of words. I still had to understand the problem.
In that pause, I found something that, with due caution, I called zen. Stopping before asking. Reading without rushing to a conclusion. Letting an answer reveal the gaps in the question too. Appreciating a suggestion while keeping the freedom to reject it. My philosophical expertise remains what it is; the need for attention, however, is something I recognise very well.
An answer that misses the point can be surprisingly useful. It forces me to clarify what I meant. The error remains an error, but trying to explain it also makes my own thinking clearer. Sometimes the decisive step happens right there, when I stop expecting the machine to understand everything I never told it.
Of course, I still do the work of a computer scientist. I open and read a reference, redo a calculation and put a solution to the test. Even the most elegant answer must account for what it claims. If a sentence sounds wonderful but the data is wrong, the beautiful prose can wait: first I fix the data.
I care especially about what remains with me after the interaction. Can I explain the result in my own words? Can I say why a suggestion convinces me? Do I still have room to decide, or have I allowed the volume of text to occupy the place of my judgement too? These are concrete questions, and they stayed with me while writing ZEN and the Art of Artificial Intelligence.
The essay starts with how models work and moves towards our relationship with their answers: attention, culture, delegation and responsibility. The words we read on a screen can end up in documents, software and decisions. It is worth pausing to see how they get there, and which part of the journey we want to keep doing ourselves.
So, to the real experts in Zen philosophy, I apologise for the excursion and welcome any pointers that might help me find my bearings. This book comes from curiosity and a working experience, with all the limitations of the person telling it. I have given it a slightly different place in the series: “A step aside”. It felt like the right name for a computer scientist who follows a question and ends up looking at himself too.
The book is available in Italian and English, with both editions under review. If you are curious, you will find it below. And if a passage makes you pause, think or disagree, write to me. The conversation can continue without a GPU.