When remotely operated vehicles settle on the seafloor or dig through a sand bed, they can kick up clouds of sediment that onboard cameras struggle to see through. Often, the only thing to do is wait until the dust settles. But a new system developed by Amy Phung, SM ’23, PhD ’26, and her advisor Richard Camilli, SM ’00, PhD ’03, of the Woods Hole Oceanographic Institution (WHOI), offers a solution.
In this technique, the vehicle first uses sonar to quickly map its surroundings. That technology doesn’t provide detailed resolution, but it works equally well in cloudy and clear water. And it allows a vehicle to safely get close enough to specific objects for cameras to visualize them more carefully.
To speed up the processing so the mapping can be done in real time, the researchers combined the sonar tech with an image-matching algorithm developed by researchers in France that quickly estimates the relative depth of each pixel in a 2D scene. It makes it easy to trick them into doing things they shouldn’t, such as telling you how to sabotage an aircraft’s navigation system. A new technique has let the company probe deeper than ever into the weird workings of an LLM.
Some donor-conceived people are finding hundreds of siblings. An international cap on donations could help prevent that. AI doesn’t just learn stereotypes from its training.
It can cook up new ones, too. Discover special offers, top stories, upcoming events, and more.
Extract — continue reading at the source.
