Big tech may have finally solved a problem the world has been trying to solve for years.
For decades, prediction systems have tried to accurately predict weather patterns but have failed to forecast snowstorms and rain showers accurately.
Now, one company has harnessed machine learning to create a prediction model for natural elements that outperforms the Ensemble Forecast (ENS). Until now, that system has been considered the best available weather prediction technology.
🚨Don’t miss this amazing Cyber Week Move! Get 60% off TheStreet Pro. Act now before it’s gone. 😲
Tech companies have been working for years to build a model that can accurately predict weather patterns. Recently, Nvidia (NVDA) and Huawei both released AI models to forecast weather elements.
But now, a different company seems to have surpassed them in creating an AI model that is delivering more accurate predictions than the ENS system 97% of the time.
Weather prediction methods forecasted to change
The winner of the race to address weather prediction with AI may be Google (GOOGL) . The big tech leader is trending today on news of positive results from Google DeepMind’s new AI-based prediction model, GenCast.
A subsidiary of Google, DeepMind is responsible for building Gemini, which is considered the company’s most advanced AI model. Now, its latest offering is revolutionizing weather forecasting.
Related: Google analysts sent scrambling by startling break-up news
GenCast isn’t even the first weather prediction model that Google has released this year. In July, a research team laid out the details of NeuralGCM, a model that combines physics with machine learning to forecast weather patterns.
However, GenCast is different, as it makes predictions based entirely on AI.
The MIT Technology Review compares it to tools like ChatGPT, noting that it predicts the next weather pattern based on previous ones, much like some writing tools that predict the words or phrases that a writer may type.
“It starts with random parameters, or weights, and compares that prediction with real weather data,” the outlet notes. “Over the course of training, GenCast’s parameters begin to align with the actual weather.”
Clearly, this training has gone well so far. According to the Review, after being trained on weather data that spanned 40 years between 1979 and 2018, it outperformed the ENS system at predicting both wind and extreme weather conditions, such as tornado paths.
Related: Apple unveils AI decision that is a major blow to Nvidia
DeepMind researcher Ilan Price, one of the creators of GenCast, notes that the weather prediction model will likely improve as it is trained on new data that is constantly becoming available. “A model that was trained up to 2018 will do worse in 2024 than a model trained up to 2023 will do in 2024,” he said.
GenCast’s forecasts are probabilistic, meaning they predict varied weather outcome likelihoods instead of precise predictions. Huawei’s Pangu-Weather, by contrast, generates deterministic forecasts with a single number, not a probabilistic range.
How does the GenCast model fit into mainstream weather predictions?
Price claims it could be incorporated into operational weather forecasting systems, providing new insights to help people prepare for important weather events and plan accordingly. He describes the model’s creation as an “inflection point” for AI advancement as a tool for weather forecasting.
More Tech Stocks:
- Analysts update Pure Storage stock price target after earnings
- Nvidia, AMD risks rise as U.S.-China chip war heats up
- Tesla replaced by much smaller rival in EV charging project
A new market for Google
Like every tech sector leader, Google has been doubling down on AI production lately, rushing to corner new facets of the market before its competitors. This development suggests that it has successfully beaten rivals to the front of the weather forecasting race, further establishing itself as a leader within the AI space.
Most companies don’t have the resources to compete with Google.
Related: Fed's Powell may have just reset bets for interest rate cuts in 2025
The company seems to have presented us with a new way of predicting the weather, which will likely be instituted into operational weather forecasting, as Price suggests. As new data becomes available and the model’s training continues, its predictions will likely increase in accuracy.
A probabilistic system makes the most sense for weather predictions. It is based on the likelihood of results, not absolute certainty, and weather patterns involve atmospheric elements that cannot be directly observed. For that reason, meteorologists often default to using physics equations to estimate their likelihood of impacting weather patterns. Using AI models to predict weather is a logical step, and Google has figured out how to achieve it faster than its competitors.
Related: Veteran fund manager sees world of pain coming for stocks
