MIT scientists have developed a training method meant to make climate emulators — AI models that quickly simulate how human activity can affect the future climate — faster to build and more adaptable. The team reports that an emulator trained on a smaller, carefully chosen set of synthetic data could still predict outcomes of scenarios it had not seen during training.
Published News » Tech
New research could lead to faster, trustworthy AI models for climate prediction
Posted by budbl.com (#1) 17 minutes ago (https://news.mit.edu)NOAA weather modeling hits key milestone with Rapid Refresh Forecast System debut
Posted by budbl.com (#1) 17 minutes ago (https://www.noaa.gov)
NOAA says its National Weather Service has started running the Rapid Refresh Forecast System, a next-generation model that provides detailed, hourly-updating short-term forecasts over North America. The new system is intended to replace a patchwork of older regional models and offer longer forecast lead times than those earlier tools.
New AI "copilot" helps engineers design better structures
Posted by budbl.com (#1) 17 minutes ago (https://news.mit.edu)
MIT researchers built an AI copilot that asks engineers for feedback while it is generating designs for structures such as buildings and bridges. In tests described in Communications Engineering, the human-in-the-loop approach produced designs that were easier to manufacture and better at resisting buckling, while adding only about 15 seconds compared with a fully automatic method.
« previous1 next »

RSS