Researchers from Lawrence Livermore National Laboratory have developed a means to detect in real time whether a metal 3D printed part will be of good quality. Using machine learning algorithms (called convolutional neural networks) and a live video feed the team can spot defects in a piece as it's printing.