A team from the University of California, in Berkeley, have manufactured a robot with the capability to pick objects from oatmeal to a small toy shark by training it rather than programming behavior.
Jeff Mahler, a researcher at the university, told the New York Times: "It figures out the best way to grab each object, right from the middle of the clutter."
The Berkeley Robot was trained by Jeff and his team by showing it hundreds of digital objects and following on from the training it was able to pick up items not in its digital data.
Robots in warehouses such as in Amazon and manufacturing companies including Foxcom are essential for moving containers in and around the warehouses or placing a chip on a circuit board and before this revolutionary training, they were unable to sort out a pile of items.
Professor Ken Goldberg said: "We're learning from simulated models and then applying that to real work."
Researchers at other establishments such as Northeastern University, Carnegie Mellon University, Google and OpenAI are also developing techniques and many believe robots will be able to master a wider variety of tasks including manufacturing.
Juan Aparicio, head of manufacturing at the German giant Siemens said: "This can extend to tasks of assembly and more complex operations. That is the road map."