MIT Technology Review is reporting that a new system developed by Amy Phung and Richard Camilli of the Woods Hole Oceanographic Institution (WHOI) allows remotely operated vehicles (ROVs) to navigate in cloudy underwater environments. The system uses sonar to quickly map surroundings, which works effectively in both clear and murky water, enabling ROVs to approach objects for closer camera inspection. To process mapping data in real time, researchers combined sonar with an image-matching algorithm from France that estimates pixel depth in a 2D scene. Camilli likened the technique to navigating a dark china shop without knocking things over, suggesting applications in scientific exploration, underwater construction, maintenance and handling unexploded undersea mines.
It also highlighted several popular stories, including a fundamental flaw that leaves large language models (LLMs) vulnerable to attacks, making them easy to trick into generating inappropriate content. The publication noted that Anthropic has developed a new technique to probe deeper into the workings of LLMs, revealing a "hidden space" where Claude processes concepts. Additionally, MIT Technology Review reported that a European fertility group is advocating for limits on sperm donors to prevent donor-conceived individuals from discovering hundreds of siblings. It also stated that AI is more prone than humans to develop biases in hiring, learning and even creating new stereotypes beyond its training data.
Full Article: Seeing through murky waters