![](/_next/image?url=https%3A%2F%2Fstorage.googleapis.com%2Fmantic-markets.appspot.com%2Fcontract-images%2FNluts%252F896d3ee74278.jpg&w=3840&q=75)
AI models can match traditional weather forecast systems at a fraction of the compute. See e.g.,
https://www.science.org/content/article/ai-churns-out-lightning-fast-forecasts-good-weather-agencies
and this excellent overview by Stephan Rasp:
But to date these systems are experimental and none is being used operationally- I.e. to make continuously updating ensemble forecasts based on newly ingested data.
Question will resolve yes if an AI model is used operationally by a group outside a traditional meteorological center.
Define AI model. This Instructables project from 2017 shows the process of building a rain prediction setup that takes in measurements and outputs probability of rain with machine learning. This paper from 2019 demonstrates a CNN model that can do ensemble forecasts. It seems that the authors of that paper would already satisfy "if an AI model is used operationally by a group outside a traditional meteorological center" if they had run their model on live weather data at any point. If I were to quickly make a ML model that takes in live weather data from an API then predict possible future states, even if it's often inaccurate, would that count?