Geoffrey Hinton, Nobel Prize winner in Physics in 2024 and one of the fathers of deep learning or deep learningHe called for a greater scientific effort to understand the risks of systems capable of surpassing certain human abilities. His warning about artificial intelligence (AI) is part of a tradition of those who break new ground and are often the first to anticipate its dangers.
Alexander Fleming offered an example in 1945 when he warned that the misuse of penicillin could lead to the emergence of resistant bacteria. Three decades later, the pioneers of recombinant DNA proposed biosafety measures and containment levels based on the risk of each experiment, starting from the premise that "technology is too valuable to abandon and too powerful to develop without safeguards." In both cases, caution arose from science itself. Progress continued, accompanied by protocols and responsibility, teaching us that scientific caution is an integral part of progress.
In technologies much closer to our daily lives, this discussion has another precedent. In 2011, Betsy Sparrow, Jenny Liu, and Daniel Wegner published four studies that gave rise to the so-called "Google effect," characterized by the fact that people who expect future access to information remember little of it, but without forgetting how to access it, thus turning the internet into a kind of external memory where information is stored collectively outside of ourselves. This finding evokes a reflection attributed to Albert Einstein in 1921, when, asked about a piece of information he didn't know, he explained that it was pointless to retain information readily available in books and added that the value of a university education lay less in learning a lot of facts than in training the mind to think. A century later, this observation acquires surprising relevance, since the real key may be deciding what is worth remembering, what we can delegate to our tools, and what intellectual capacities we should preserve.
Imagine a student consulting an equation with AI. If they use it to compare procedures, detect errors, and explore solutions, the technology allows them to expand their learning. But if they simply copy the result, they reduce the opportunity to exercise their reasoning. The difference lies less in the machine itself than in the cognitive and emotional relationship established with it.
In this way, Hinton draws attention to future risks; the "Google effect" reminds us that some transformations begin silently. One suggestion is that educating for and with AI requires questioning the origin of a claim, reconstructing procedures, comparing sources, acknowledging uncertainty, and preserving fundamental skills.
Therefore, the challenge lies in knowing how to use these tools to think better, never to think less. The history of science shows that responsible caution can preserve what makes an innovation valuable. Scientific prudence, far from hindering progress, helps to give it direction and meaning.
@betancourt_phd
