Habr author warns: mindless use of AI is degrading professional expertise

A Habr author has raised the question of where mindless AI use may lead. In his view, the owners of large language models are chasing profit, and in doing so they risk degrading not only the quality of text but also human expertise.

Why generated text doesn't replace expertise

According to the author, generative transformers invent nothing. They simply use large language models to generate, that is, to combine, plausible-sounding text. People who use generated text without proper critical analysis unwittingly lose their expertise, if they ever had any. And people who were never qualified specialists to begin with fall into euphoria, imagining they can now solve problems without the necessary knowledge.

"This will work as long as genuinely knowledgeable people are running development.

But when they are replaced by people who have mastered working with AI yet lack the qualifications to soberly evaluate the results AI produces, that's when disaster will strike," the author warns.

What happens to models from version to version

The author notes that with each new version, models gain hidden deterministic mechanisms (what people often describe as "the model thinking") that are supposed to make answers more sensible. At the same time, though, more and more of the human factor gets baked into model outputs, and human bias can leave its mark on the final result. For the product's commercial promotion, the author wryly observes, this is certainly a good thing.

A separate problem is the data models are trained on: nothing guarantees its reliability. Code generation, the author observes, is in somewhat better shape, but even there AI fundamentally invents nothing new. The most that can happen if models are trained on the generations judged most successful is some improvement, or, conversely, degeneration of the final results.

Who is at risk

The author's central point is that the danger comes not from large language models themselves but from how people without proper expertise use them. As long as decisions at companies are made by people with real expertise who can critically evaluate AI output, the problem stays manageable. The risk grows when those people in leadership positions start being replaced by specialists who can only interact with AI and can't soberly judge the quality of the results. In essence, this is a generational shift within the profession: one generation still learned to solve problems with their own hands and heads, while the next risks learning only how to write prompts for a model.

What to do about this risk in practice

The author's reasoning leads to a practical takeaway: treat AI output as a draft, not a finished answer, and have it checked by someone who can judge the substance, not just the form. This applies to both text and code. In both cases the model combines what it has already seen rather than creating fundamentally new knowledge, and responsibility for the quality of the final result stays with the person who uses it.

What it means

The author urges readers not to confuse knowing how to use AI with professional expertise: the tool combines plausible-sounding text, but it doesn't replace the ability to critically check the result. As long as knowledgeable people stay at the helm, the risk is manageable, but it grows as they are replaced by people who have only learned to work with AI.