In his second report from the Science Beyond Fiction expo in Prague, Dan Simmons meets scientists trying to develop robots that behave like humans.
How bots are learning to collaborate with each other
Simple tasks such as walking and picking up objects are part of everyday life for people, but such mundane tasks still present major challenges for today's robots.
Researchers are using sensors, cameras and recognition to teach the machines to interact in a way that people will be comfortable with.
Rachid Alami, senior scientist at the National Centre for Scientific Research in France, said that actions by robots should be easily interpreted by humans.
Rachid Alami said robots needs to be optimized for dealing with people
"Generally the engineers optimize the energy, optimize the time but do not optimize the acceptability and legibility of the [robot's] task," he said.
Human movement is fluid and adaptable because people learn how to move unaided and without a stream of conscious instructions from the brain.
Few modern robots can replicate a smooth sequence of movements in the same way. But, say scientists, once they do master it machines could become useful in the fields such as neuroscience.
But the advantages do not lie wholly in human hands. Robots can perform some tasks that people find difficult. At the show one robot was shown to balance a ball successfully thanks to feedback from light sensors.
How bots could one day move like a person
The sensors tell it where to move next to keep the ball up - such precision and speed will be important if robots are ever to achieve good balance.
Some robots are learning to collaborate in a swarm-like way by communicating where they are and what they have found out.
"The system is very reliable because when a part of the system is out of order or destroyed, others can take this role and the whole system still works," said Serge Kernback from the University of Stuttgart.
He added that this new kind of self-organisation suggests "there is still frontiers in robotic research".
Machines are better at some tasks such as balancing a ball
One robot called Noa at the Science Beyond Fiction expo was able to pick up objects after being fed pictures wirelessly from a camera above it.
Noa knows that a ball will sit on a cube but a cube will not sit on a ball, but its intelligence is thin - based on pre-programmed learning sequences.
Researchers eventually want machines to be able to find the answers to their own questions about tasks.
To make this a reality, scientists have been studying the brainwaves of rats.
Stefan Schaffelhofer, researcher at medical and electrical engineering firm Guger Technologies, said it is possible to reconstruct the physical position of an animal from brainwaves by using plate cell neurons.
"It's like GPS but using brainwaves instead of satellites, explained Mr Schaffelhofer.
"Knowing how the brain deals with the space around us could for example help people with disabilities control robotic aides in future in a very natural way," he added.