We Will Become More Stupid and Lazy According to Two Studies: And It's the Fault of Artificial Intelligence

Artificial intelligence has shown to undermine professional capabilities, specifically reducing the skills of certain specialists. An article in Nature highlighted how excessive reliance on these tools can lead to a deterioration of human competencies, analyzing two studies conducted on healthcare operators and computer engineers.

The first study involved a group of Polish doctors specialized in endoscopy, all with experience of at least 2,000 colonoscopies. These professionals were provided with an AI system capable of analyzing colonoscopy images in real-time, indicating the presence of adenomas and precancerous intestinal lesions. The tool was available only on certain days, allowing a direct comparison between performances with and without AI support.

The results, published last year in The Lancet Gastroenterology and Hepatology, revealed a significant decline in doctors' performances when the AI system was not available, once they had begun to use it. In the three months prior to the introduction of AI, specialists detected at least one adenoma in 28.4% of colonoscopies. In the three months following, the detection rate of adenomas in colonoscopies performed without AI assistance dropped to 22.4%.

Artificial Intelligence and the Loss of Computer Knowledge and Skills
Robert Wachter, a physician cited by the source from the University of California, San Francisco, and author of a book on the impact of AI in healthcare, suggested that even highly qualified professionals may suffer a decline in their tasks due to a growing dependency on artificial intelligence tools. The authors of the study concluded that continuous exposure to such tools can make doctors "less motivated, less focused, and less responsible in making cognitive decisions without the assistance of AI."

The second study mentioned in the Nature article focused on the loss of skills in the computer industry. Researchers from Anthropic conducted a randomized controlled clinical study involving 52 software engineers. All participants were assigned a basic programming task, with access to the web and detailed instructions. Half of the participants received an explicit suggestion to use an AI-based assistant. Subsequently, all engineers were subjected to a quiz to evaluate what they had learned from the activity. The results were clear: the group that had used the AI assistant scored an average of 50%, compared to 67% for the group that did not use it. The performance of the AI-assisted group was particularly poor on questions requiring error diagnosis in the code, suggesting a gap in learning the underlying concepts of the produced code. This study was published on the preprint server arXiv, pending peer review.

Tapani Rinta-Kahila, an information systems researcher at the University of Queensland in Brisbane, Australia, noted that other technologies have already made some skills obsolete in the past. For example, GPS navigation systems have gradually eroded people’s orientation skills, and this—apparently—will happen with other types of abilities through artificial intelligence.