Storms like Hurricane Polo in 2026, which rapidly intensified from a tropical storm to a Category 5 hurricane in just 24 hours, highlight the need for improved forecasting. These sudden changes can leave communities with little time to prepare or evacuate, emphasizing the importance of accurate predictions. Artificial intelligence has significantly advanced weather forecasting in recent years, with AI models now capable of producing forecasts that match or even exceed traditional physics-based systems. This progress has been driven by the availability of massive weather datasets, improvements in AI algorithms, and access to powerful computing resources. While much of the current focus on improving AI for weather forecasting centers on models and hardware, the data itself plays a crucial role. At the global scale, AI has access to decades of climate and weather records covering the entire Earth. These datasets, containing millions of examples of atmospheric changes over time, allow AI models to recognize patterns that were previously impossible to detect. However, when zooming in from the global to the regional scale, forecasting becomes more complex. This is particularly important for predicting hurricane intensity, as rapid intensification can transform a weak storm into a destructive force in a matter of hours. Regional weather forecasts, such as those for hurricanes, often focus on extreme events that develop quickly over short periods. These events require highly detailed data, which current global datasets often lack. Scientists training AI models to predict hurricane intensity typically use two types of data: direct observations from weather stations, radars, buoys, and satellites, and simulated data from weather models. Observations are detailed but limited, especially in the open ocean where most hurricane development occurs. Simulated data provides a more complete picture of the atmosphere but contains uncertainties due to approximations in computer models. Even with better data, predicting hurricane intensity remains a challenge due to the chaotic nature of these storms. Tiny differences in initial conditions can lead to vastly different outcomes, making long-term predictions difficult. Recent research suggests that hurricanes may have elements of chaos that prevent AI models from accurately predicting intensity over time. While scientists aim to minimize forecast errors, capturing this chaos means accepting some level of unpredictability. This challenge highlights that the future of hurricane forecasting may need to focus on predicting a range of possible intensities and their probabilities, rather than a single, precise outcome.