Despite all the advances that have been made in the areas of technology, AI, and machine learning, weather companies still cannot trust them with humans. While AI models, with the help of weather satellites such as GOES-16 and 17, are capable of predicting small weather changes and patterns with surprising accuracy, the effectiveness of they decrease sharply when major climate events occur.
These major weather changes, including natural disasters, often have warning signs that are too subtle for most AIs to detect. Even when they do, weather AI models can’t always give a correlation between initial indicators, such as a tornado, and its likelihood.
This often requires the observant eye of an industry veteran. Based on information from more than two decades collected by the US National Oceanic and Atmospheric Administration, humans outperform two of the most popular national weather forecasting models, the Global Forecast. System and North American Mesoscale Forecast System.
The data shows that humans outperform AI by 20 to 40%. That is no small distance, and can be the difference between life and death for those on the path of a dangerous tornado, tornado or blizzard. fast. Access to accurate weather information as soon as possible is what makes evacuation (if needed) or shelter-in-place recommendations possible.
Seasoned weather experts can look at the tiniest details, like slight changes in barometric pressure, wind speed or humidity, and draw conclusions with more confidence than a machine. count.
And frankly, this is no surprise. AIs are getting smarter, but they still lack what humans have always possessed: the ability to assess situations in a large context. Bringing machines closer to human perception would require an enormous amount of processing power that only a handful are currently being studied.
At this time, forecasters are still “safe” from the wave of robots replacing human jobs.