Cornell researchers are working on developing an algorithm through physical feedback to teach robots to be more careful with certain objects. For example knives! Simply because we don't want to die in the hands of a robot holding one!

The robot can perform movements as it does, but a human operator can intervene at any point with the push of a button and adjust the bot's trajectory and movements to something safer. The bot will grab a knife, but is unaware that the blade can cause damage, so the human operator moves the robot's arm into a safer position. Over time that algorithm will learn to be careful when handling sharp objects.

Of course, it's not just for our safety. If we can teach robots to be more careful with objects like that, we could teach them to be more careful with fragile items. Eggs or LCD screens or stuff that needs to be handled with more care than usual. Check out the video below demonstrating the algorithm: [Cornell University via TechCrunch]