~/aixsci
weekly updates

learn/what the ai was for

Forecasting

Forecasting is using a machine to guess what a system will do next, based on what it is doing now and what it has done before, at a time when nobody can yet check the answer.

What it is

A forecast takes a record of how something changed over time, called a time series, and extends it forward. The machine is shown many stretches of past record, each paired with what actually came next, and it adjusts itself until its guesses match. Once trained, it is handed a fresh recent stretch and asked for the step ahead. Some models are built to carry a memory of earlier steps, so what happened last week can still shape the answer. Others compare every moment in the input with every other moment and weigh which ones matter. To reach further ahead, a model can be fed its own output back in as if it were real observation, and so step forward again and again.

Why scientists use it

Forecasting is used wherever a decision has to be made before the event. Where a storm will go, whether a river will run too warm, how much pollutant the air will hold tomorrow: all are questions that only matter answered in advance. Physics-based computer models can answer some of them, but they are slow and demand a full description of the system. A trained model instead learns the regularities straight from records, which is cheap once done.

A forecast is also a strong test of understanding. Describing a system that has already been measured is easier than naming what it will do, because the answer cannot be adjusted after the fact. Sometimes the question is reshaped: instead of predicting a value, the machine is asked simply whether an event is coming.

In the record

Limits

A model learns the patterns in its record, so conditions it has never seen can be forecast badly, and a changing world slowly invalidates it. Scores measured on held-back past data flatter a model, because that past is already known; genuine future performance is usually worse. Rare events are the hardest, having few examples to learn from. And a model that fits the data well may have latched on to a coincidence rather than a cause, so it offers no explanation of why the future will look as it says.

Written by AIxSci: the background from general knowledge, the examples from checked records.

By field

Every article

12articles
0structural-biology
0materials-chemistry
0astronomy
12climate-weather

aix-00283

Deep learning forecast of El Niño suggests Indian and Atlantic oceans extend warning time

forecastingsimulation surrogate

climate-weather · ai-result · Science Advances · 2026

aix-00285

Neural network forecasts freak waves minutes ahead from a single buoy's record

forecastingclassification

climate-weather · ai-result · Scientific Reports · 2024

aix-00286

AI fills gaps in river temperature records to track overlapping heatwaves

property predictionforecasting

climate-weather · ai-result · Nature Geoscience · 2026

aix-00282

Neural network trained on a decade of records forecasts typhoon strength and position

forecasting

climate-weather · ai-result · PLoS ONE · 2024

aix-00291

Statistical model estimates what Makkah's air would have held without lockdown

forecasting

climate-weather · ai-supporting · Toxics · 2022

aix-00292

Neural network forecasts daily sea surface temperature up to 90 days ahead

forecasting

climate-weather · ai-result · PLoS ONE · 2026

aix-00294

Neural network trained on simple equations spots tipping points in real data

classificationforecasting

climate-weather · ai-result · Royal Society Open Science · 2022

aix-00295

Diffusion model trained on past weather produces 15-day global ensemble forecasts

forecasting

climate-weather · ai-result · Nature · 2024

aix-00296

Neural networks trained on 39 years of weather data produce week-ahead global forecasts

forecasting

climate-weather · ai-result · Nature · 2023

aix-00297

Weather and climate model built from physics equations plus a trained neural network

forecastingsimulation surrogate

climate-weather · ai-result · Nature · 2024

aix-00298

Generative AI model forecasts UK rainfall 90 minutes ahead from radar images

forecasting

climate-weather · ai-result · Nature · 2021

aix-00299

Neural network forecasts weather straight from raw satellite and station observations

forecastingstructure determination

climate-weather · ai-result · Nature · 2025