Let me ask this directly: are large language models making us less intelligent? The honest answer is more nuanced than a simple yes or no.

On one hand, the shift is undeniable. We are increasingly operating within systems we do not fully understand. The world has become a deeper black box than it used to be. I often find myself interacting more with AI agents than with other humans. They correct my writing, debug my code, perform calculations, and even suggest how to handle emotionally charged situations. In many ways, they are not just assisting thought. They are beginning to shape it.

This raises a legitimate concern. Thinking is not effortless. It requires focus, time, and discomfort. Much of what we truly learn comes from struggling through problems, not from receiving polished answers. When AI removes that struggle, it can also remove an important layer of cognitive development. Generative AI makes it easy to consume conclusions without engaging deeply with the reasoning behind them. Over time, this can reduce our tolerance for effortful thinking.

A more subtle risk is the gradual erosion of cognitive stamina. When answers are always a prompt away, we may lose the habit of holding complex ideas in working memory or following long chains of reasoning. This is like how GPS affected our spatial navigation skills. Convenience replaced practice. I can’t imagine driving to a new place without the maps on. With LLMs, the stakes are higher because the outsourced functions include reasoning, synthesis, and articulation. If we consistently default to AI for first drafts of thinking, we may become better editors than originators, relying on external systems not just for efficiency, but for the very structure of our thoughts.

However, there is a strong counterargument. Tools have always changed how humans think. Calculators did not make mathematicians obsolete. They allowed them to focus on higher-level problems. Search engines did not destroy knowledge. They expanded access to it. In the same way, LLMs can be seen as cognitive amplifiers rather than replacements. They allow us to prototype faster, explore more ideas, and offload repetitive tasks so we can spend more time on creativity and decision-making.

From this perspective, the issue is not that AI makes us less intelligent, but that it changes what intelligence looks like. The ability to ask the right questions, evaluate outputs critically, and integrate insights across domains may become more important than raw memorisation or manual problem-solving. Writing a good prompt or identifying flaws in an AI-generated solution is itself a form of thinking.

The real tension lies in how we choose to use these systems. If we rely on AI as a substitute for thinking, we risk cognitive atrophy. If we use it as a partner in thinking, we can extend our capabilities. The same tool can either weaken or strengthen us, depending on how it is integrated into our habits.

Looking ahead, there is also a broader societal concern. If future generations grow up relying heavily on AI for reasoning, will they develop the same depth of understanding? Or will innovation increasingly emerge from machines trained on past human knowledge? At the same time, these tools could democratize expertise, allowing more people to contribute to complex fields without years of specialised training.

I am not against progress. AI is one of the most powerful tools we have built, and its benefits are real. But there is something worth preserving: the human ability to sit with a problem, to struggle, to think deeply without immediate assistance. Perhaps the better question is not whether LLMs are making us less intelligent, but whether we are using them in ways that make us less willing to think. The answer to that is still being written.

END OF ARTICLE