A Translator That Follows the Meaning: How a Neural Network Reads the Ocean Two Years Ahead
A translator who gets distracted by every rustle in the room turns a one-hour meeting into two. That is how old statistical El Niño models worked: they picked out individual numbers from ocean maps but missed how a warm patch off Peru connects to a cool tongue near Australia. The new neural network behaves differently: it follows not every word but the meaning of the paragraph.