The incident on Mount Shasta in California once again raises the question of the limits of trusting artificial intelligence in matters where the cost of a mistake is health and life. Three tourists, who prepared their climb using Google's Gemini chatbot, faced serious consequences from underestimating the route. As it turned out, the neural network recommended the group take significantly less food and water than was required for the actual ascent.
The planned climbing time was eight hours, but the actual ascent stretched to 16. The tourists began their descent already in the dark, which led to a knee injury for one of the participants. In the end, the group lost their way and was forced to spend the night in a canyon before rescuers could provide assistance and escort them back to the trailhead.
The Dangerous Overconfidence of Algorithms
This case is not an isolated example but a symptom of a broader problem. Modern language models, including Gemini, excel at generating text and analyzing data, but their "knowledge" of real physical exertion, weather conditions, and the specifics of a particular area is often averaged or outdated. The algorithm does not account for individual physical fitness, the current state of the trail, and other critical variables that an experienced ranger would assess on the spot.
The Siskiyou County Sheriff's Office, which conducted the rescue operation, emphasized the need to cross-check any information obtained from AI with local experts. This is an absolutely correct approach: AI should be a tool for preliminary analysis, not the final authority on safety matters.
My view as an analyst: The market for AI solutions for tourism and outdoor activities is growing, but developers should implement clear disclaimers and limitations in such scenarios. Until models learn to integrate real data from sensors, weather forecasts, and up-to-date community reports, relying on them in extreme conditions is a risk that could cost too much. The Shasta incident is a stark reminder that technology has not yet replaced human experience and common sense.