AI and Open Access Weather Data Will Provide Groundbreaking Insights
Humans have experienced a unique relationship with weather. In addition to living with the environment around us, people have discussed the weather as a classic small talk tactic, predicted weather using old knee injuries, and have been caught off guard when the weather was completely different than we expected.
With technology, modern society has taken away the mystery from more and more things. Case in point, you can watch your food delivery person on your phone in real-time, thanks to the smart computer that is your phone and a network of interconnected satellites orbiting far out in space. Arguably this isn’t why all this technology was created, but it is certainly an interesting side benefit.
That said, the weather continues to elude us in terms of having high accuracy. We can predict the shipping of an item we bought halfway around the world, and most of the time, that estimate is correct with perhaps a day variance on either side. But with the weather, we are not surprised when the forecast calls for sun, but later that afternoon, you are very much wishing you had brought an umbrella.
Artificial Intelligence or AI ML News: Iterate.ai Adds Generative AI Features to Interplay’s Low-code Environment
So, why is this happening? Why have we made massive advancements in so many things but not in the weather forecast industry? Well, we actually have made quite a bit of progress. Still, the critical point is this: thanks to advances in artificial intelligence (AI), more ways to acquire huge amounts of weather data, and the ability to compile and gain insights quickly, we are on the verge of gaining major leaps in understanding how our weather works, how climate change is affecting it, and how to predict it better.
Artificial Intelligence or AI’s Natural Strengths
If you are to believe some of the current headlines, AI seems to be everywhere, accomplishing unheard-of achievements and seemingly capable of anything. While there have been major breakthroughs that continue to build on each other, AI is still complex, difficult, and in nearly all cases, frustratingly narrow-focused. It is able to perform tasks and solve problems extremely well, assuming it has the right data, enough of it, and the boundaries of the problem are small enough. We are a long way away from “general AI,” the type of artificial being that can perform wildly different tasks.
However, we don’t need general AI to solve many previously impossible problems.
Artificial Intelligence or AI tends to be extremely good at a handful of things, which are then applied creatively to countless use cases. AI can detect patterns (e.g., image classification, object detection), use evolution-like behavior combined with structured reinforcement to learn new tasks (e.g., teaching a robot how to walk), and it can predict the near future (e.g., autocomplete, translation, regression). This last element is critical to the problem of predicting the weather. With enough data, even something as complex as the weather can be understood and predicted through the proper application of AI.
Weather’s Incredible Complexity
Weather is notoriously difficult to predict because it is an incredibly complex, interconnected, fluid system. The term “the butterfly effect” is appropriate when discussing the weather because although large trends like average temperatures in a given location for a given month don’t change a great deal each year (excluding the effects of climate change), the day-to-day changes can vary significantly in terms of temperature, precipitation, pressure, wind, and other variables. Because these variables are not consistent over even a small space and are constantly being mixed around in the fluid we call the atmosphere, change is constant, and prediction accuracy past a day or two drops significantly.
A key issue is that there are many constantly changing variables. While we can see correlations between them, it is difficult to assign causality between variables in a system this complex. Even advanced statistical analysis can’t handle this many variables affecting a system that is only a “closed system” if you include the entire globe.
Breakthroughs On The Horizon
However, this is where AI comes into play. It is exceedingly good at taking data sets with hundreds or thousands of features, all of which could be interacting with each other, and with the proper training, develop an understanding of how the variables influence each other to create a certain result. In addition to the advances in AI algorithms, our ability to capture, store, and process vast amounts of data is also critical.
AiThority.com AI ML Insights: AiThority Interview with Pete Wurman, Director at Sony AI
For example, Lockheed Martin and NVIDIA are teaming to generate a “digital twin” of the current global weather, funded by the National Oceanic and Atmospheric Administration (NOAA). This twin (a digital representation that uses data to recreate and mimic the original essentially) will have terabytes of data and will use AI to help the process, understand, and display it so that researchers can better understand on a global scale our entire weather system.
What is unique about our current state of the art is that it doesn’t take a Lockheed Martin to create insight, innovation, and breakthroughs for better weather prediction. The two biggest hurdles—data and the proper code—are both available to anyone who wants them. For Artificial Intelligence, while it takes training to understand the algorithms and the process, nearly all the code is open source, and there are vast amounts of training material online. For the weather data, there are now platforms like Tomorrow.io that offer a historical weather data API where users can connect and access vast amounts of detailed weather data in order to train AI models, then utilize the API to get current weather data. This wide-open access will undoubtedly attract many talented developers, and the effort is very likely to result in new weather prediction breakthroughs. We have the data, we have the AI, and we have the talent.
Comments are closed.