Artificial intelligence can now analyze the root causes of traffic accidents and forecast future collisions.
Researchers at Johns Hopkins University (USA) have developed an artificial intelligence (AI) tool designed to forecast road accidents. Their pioneering findings were published in the esteemed journal Nature Communications.
The scientific team created a sophisticated program named SafeTraffic Copilot. This AI is not only adept at analyzing the underlying reasons for traffic incidents but also boasts the capability to accurately predict future accidents. Experts anticipate that this innovative AI will significantly contribute to reducing both the frequency of road accidents and the number of casualties on American roadways.
The researchers leveraged large language models (LLMs), training them on an extensive dataset comprising text, numerical values, and satellite data. The SafeTraffic Copilot model is uniquely equipped to evaluate both individual and synergistic risk factors. Furthermore, the program can discern connections between various events and the circumstances that precede accidents.
“With AI, we can transition from relying on broad, aggregated statistics to gaining a precise understanding of the root causes behind specific collisions,” stated lead researcher Hao Yang. SafeTraffic Copilot’s continuous learning capability allows it to integrate new information and adapt to evolving conditions. The scientists reported an impressive 70 percent accuracy rate for the AI’s predictions. Yang underscored, “Despite various countermeasures, the incidence of automobile accidents in the U.S. continues to climb. These are complex occurrences, influenced by a multitude of elements such as weather conditions, traffic patterns, and driver behavior.”
