Artificial Neural Networks
Layered computational structures that learn by adjusting connection strengths against examples.
Why it matters now
They are now the fastest way to find structure in environmental data too large for people to read: every satellite pass, every acoustic recorder, every weather station. Forecast systems built on them run in seconds what once took supercomputer hours, which puts serious prediction within reach of countries that could never afford the hardware.
Where this is happening on Earth
- EuropeGraphCast and neural weather forecasting, developed in LondonMachine-learned forecasts matching or beating physics models at a fraction of the compute.
- North AmericaGeoffrey Hinton and Yann LeCun, Toronto and New YorkBuilt the backpropagation and convolutional methods the field rests on.
- AfricaCrop disease detection from phone photographs, Tanzania and KenyaCassava and maize disease diagnosed in the field without a laboratory.
- AsiaAir quality forecasting across Chinese and Indian citiesNeural models producing neighbourhood-level pollution forecasts.
- South AmericaDeforestation alert classification from satellite imagery, BrazilNear-real-time clearance detection that triggers enforcement within days.
The idea itself
The encyclopedia entry first, then live searches at journals and magazines that keep returning current work on this specific idea.
Current reporting from around the world
Standing feeds at reputable periodicals. These stay fresh on their own, which is why they are here instead of a list of articles that would be out of date within a season.