A recent breakthrough in computer vision has given robot the ability to differentiate objects.
MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has recently made progress in computer vision that lets machines identify objects and take actions without knowledge about the object.
The system “Dense Object Nets” (DON) developed by CSAIL now can create “visual roadmaps” of the object’s points, the robot will teach itself to understand and manipulate the items, enabling it to pick it up. This has been a breakthrough in this area of work, as a variety of advances couldn’t inspect the specific parts of an object according to Ph.D. student Lucas Manuelli. It is viewed as a potential use not only for manufacturing, but also for household application due to its self-supervised feature for some specific tasks such as putting plates onto a shelf while you are on vacation. In the future, the team hopes to make more improvements so that it can understand objects deeply like cleaning a desk.
Another issue should not be ignored is its ability to execute action without supervision. According to Layer 1 of the AIWS 7-Layer Model, machines and technology should be created with concern over human’s prosperity and safety. It should be carefully supervised, and programmed to avoid malfunction and further damage.
Watch video: