Tiit Lepik’s mailing list

My duties at Katevara OÜ ended with September. The old e-mail tiit@katevara.ee no longer works. Letters sent there may not reach me.
My working mailbox is tiit.lepik.71@gmail.com. Letters sent to the technical mailbox tiit@lepik.co also arrive.

Today on Facebook I came across a post by my acquaintance Kalle Aleksius on the topics of an article in The Economist. The topics were exciting for me as an information technologist. I have always been amazed at people’s faith in the power of mathematics — serious people claim that an algorithm will take the Earth over from people. Today most people seem to find it fitting to believe such nonsense. It only gives the speakers more money and courage. The Economist’s journalist disagrees with the authors on the interpretation of the results of their work.

Below is the comment I wrote to Kalle, slightly polished. At the end, a link to my Google Drive, where there are a dozen PDF files of research papers that have seemed important to me in this context.

I cannot give an assessment of techno-optimism or techno-pessimism. Both are assessments rather than knowledge. I have thought that conservatism and the rejection of everything new is, in today’s society too, a self-defence mechanism of the species inherited from evolution, which is in us through the structure of the brain.

When I went to Võsu school in Soviet Estonia at the end of the 1970s, the role of today’s smart devices and AI as the culprit was filled by the ballpoint pen. All the teachers and the whole society of the time were convinced that one must write with a dip pen, or at most a fountain pen. Writing with a ballpoint was said to ruin one’s handwriting; it would produce a useless member of society, followed by the decline and ruin, if not of humankind, then at least of the Soviet Estonian person.

A few weeks ago, with the help of AI, I compiled a timeline from the data collected on my computer and saw that a similar accumulation of firm convictions also accompanied the invention of printing. It was thought that if book learning reached the common people, humankind would come to an end, because nobody would do the work needed for life any more.

This summer and autumn several large studies have been completed that make one ask whether the device is to blame, or what is done with it and how. In Scotland 3,786 children were followed until the ages of 5, 10 and 15 (McQuaid et al., JAMA Network Open, August 2026). When family functioning was taken into account, little lasting harm was found: in small children only very high screen time, five hours a day or more, was associated with greater difficulties, and by the age of 15 this association disappeared. Among adolescents only very long gaming hours were associated with difficulties. In the Finnish PANIC study (Jalanko et al., Pediatric Exercise Science, 2026), more screen time from childhood to adolescence was not associated with weaker cognitive ability; for some measures the association was even the opposite. The largest pooled analysis (Al-Jbouri et al., JAMA Pediatrics, September 2026: 23 meta-analyses, nearly 2.8 million children from 64 countries) found associations with lower general academic achievement, weaker language development and emotional difficulties, but not with mathematics, reading or science. There too the results depended on the type of screen activity, and these are associations, not proven causes. Logical: why should so-called smart time be limited if a child runs across a meadow with a smart device identifying flowers and butterflies?

The same nuance is in the study on the basis of which The Economist claimed that AI harms learning (Strömberg, Lei and Wu, 2026, nearly 27,000 Chinese pupils). With AI, homework grades rose and exam results fell. But the loss came from the four out of five users who let the AI do their homework. Those who still spent as much time on homework learned as well as before. According to the authors, teaching AI tools exist and are often free, but pupils do not choose them. In their view the problem lies in how learning is organised and what is assessed, not in the technology.

I have also thought that perhaps the strange times of today, when we live as if in the world of the drunkards’ talk behind the village shop, do not come from the development of information processing technology. The reason may instead be that humankind’s knowledge and experience are not in the machines’ information field. Most foundational texts and reference works, all the information that carries on humankind’s experience, are inaccessible to machines. Wikipedia replaces the encyclopedia, but it is more of a nice folklore. I noticed this in myself a few days ago. For my work I needed the position of an ancient Greek thinker. His book in Marju Lepajõe’s translation was on the shelf, but what I actually used was a machine translation found online by AI, probably via several languages.

I recommend reading these studies yourself: https://drive.google.com/drive/folders/1Wx3MCpeIFTEM3zENimPylbGjZaT31Wv0