articles

Neural networks, which learn to perform computational tasks by analyzing large sets of training data, are responsible for today’s best-performing artificial intelligence systems, from speech recognition systems, to automatic translators, to self-driving cars.

But neural nets are black boxes. Once they’ve been trained, even their designers rarely have any idea what they’re doing — what data elements they’re processing and how.

Two years ago, a team of computer-vision researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) described a method for peering into the black box of a neural net trained to identify visual scenes. The method provided some interesting insights, but it required data to be sent to human reviewers recruited through Amazon’s Mechanical Turk crowdsourcing service.

A venerable U.S./European oceanography satellite mission with NASA participation that has expanded our knowledge of global sea level change, ocean currents and climate phenomena like El Niño and La Niña will take on an additional role next month: improving maps of Earth's sea floor.

The Ocean Surface Topography Mission (OSTM)/Jason-2 satellite, a partnership among NASA, the National Oceanic and Atmospheric Administration (NOAA), the French Space Agency Centre National d'Etudes Spatiales (CNES) and the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), marked its ninth year in orbit on June 20. Designed to fly three to five years, OSTM/Jason-2 has now completed more than 42,000 trips around our planet, contributing to a database of satellite altimetry that dates back to the launch of the U.S./French Topex/Poseidon satellite in 1992.

Imagine going to the supermarket to stock up on groceries but coming home empty-handed because you just couldn't figure out how to work the shopping cart or figure out how to get to the ice cream tubs in the freezer aisle.

Welcome to the life of a bumblebee. 

Gathering sweet nectar from flowers, it turns out, is much more difficult than one might think, and it requires a lengthy learning process. By the time a bee has figured out how to efficiently pry open the lips of a snapdragon flower, for example, most likely it has made dozens, if not hundreds, of floral visits. 

A new CU Boulder study comparing dissolved black carbon deposition on ice and snow in ecosystems around the world (including Antarctica, the Arctic and alpine regions of the Himalayas, Rockies, Andes and Alps) shows that while concentrations vary widely, significant amounts can persist in both pristine and non-pristine areas of snow.

Black carbon is the soot-like byproduct of wildfires and fossil fuel consumption, able to be carried long distances via atmospheric transport. Because these black particles absorb more heat than white snow, the study of black carbon concentrations in glaciers is important for predicting future melt rates.

A study of 60 million Americans over the age of 65 estimates that thousands of people are still dying prematurely each year because they are breathing polluted air.