Why our metal comrades still need a bit more practice

Technology is still our number-one hope when it comes to the great problems of humanity. Imagine all the things that might be possible with advanced technology like robotics? Sure, there are plenty of possibilities, but what is actually realistic in the near future? What can robots already do, and what’s still a long way off ? And what about in the social sector? Robots for development? And what influence do the various key players have on the use and evolution of robotics?

It’s summer, we’re in a beer garden. Students from the Machine Learning and Robotics research group from the University of Stuttgart are discussing how to improve the world over potato salad and wheat beer. With energy harvesting ships, maybe? it could work,
but not this decade. Because today’s robots, unfortunately, still can’t be deployed as universally as in the movies I, Robot or Star Wars. They are highly specialised machines which have so far been used mainly in manufacturing and in the military. There are also a few research robots, and some are getting close to being ready for consumers. In manufacturing, it's mostly large, high-precision robotic arms which are finding use. They are really good at repeating programmed sequences of movement with total precision. Robotic arms like the ones made by the German company KUKA have been used across the world since the 1970s, for jobs like assemling cars. But these robots can barely react at all to their environment. They have to be kept in cages, people have to be protected from them. 

Drones: saving lives instead of blowing people up

What’s new is that robotics companies are now trying to build robots that have more contact with their human counterparts, collaborating with their colleagues instead of putting them in danger. However, these robots still have to be confined to a clearly defined domain: their factory. Either they are not capable of moving, or they can only do so on smooth concrete floors. In any case, factories have the advantage that the environments can be adapted to the robots rather than the robots having to adapt to their surroundings. This way, you can simply avoid the most difficult situations. The military is the second major sector in which robots are used, primarily in the form of drones. These unmanned airplanes, helicopters, ships or vehicles are frequently used, for instance, to locate targets or to conduct airstrikes (in 2015, according to the US government, at least 2697 people were killed by US drone strikes, including 116 civilians). Technically, a drone is a more or less a highly advanced, remote-controlled vehicle. A pilot sits at ground control and steers it from a distance. Much of this has been automated so that pilots can control several drones at the same time. But the missions’ critical decisions are still made by a soldier.  Airborne drones in particular operate in a relatively clearly defined environment as well. Once they are in the air, they can move with considerable freedom, without running into too many obstacles. Things gets more confusing and unclear once you head into the research labs of the universities. They have been carrying out research on autonomous robots since the 1960s, as part of their research into artificial intelligence. The driving force here is the desire to understand human intelligence. As the physicist Richard Feynman said shortly before his death: “what I cannot create, I do not understand”. And that’s why they are still attempting to reproduce intelligent behaviour in the physical world as well. After the initial success and enthusiasm, however, the researchers came to the sobering realisation that it is not so easy to reproduce real intelligence. They can complete special, clearly defined tasks in controlled environments very well. But robots often fail miserably at solving problems that are simple for us. 

The complexity of our actions is immense

The psychologist Steven Pinker puts it this way: “The main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard. The mental abilities of a four-year-old that we take for granted – recognising a face, lifting a pencil, walking across a room, answering a question – in fact solve some of the hardest engineering problems ever conceived.” Therefore, researchers are currently focused on teaching robots to cope with ever-changing environmental conditions: at conferences, they present algorithms for the generation of movement, for grasping and walking. The field of computer vision, how robots see, has become a research field of its own, with its own conferences and journals. Machine learning is another field. And the higher levels of intelligent robot behaviour have only been brought back into focus in recent years. The results should not be underestimated. But the complexity of our own actions is immense, and it presents robots with equally colossal tasks. We are still miles away from general artificial intelligence that is able to operate sensibly beyond its pre-defined tasks.

The social aspects depend on the interests of the researchers

The current status of this technology can always be gauged from the DARPA Grand Challenge, a major robotics competition. There, in 2015, a robot had to open a large rotating valve. Using both hands, it could confidently make very precise movements. It was supposed to grasp the wheel and rotate it in order to close a valve. But the robot was tinkering away some 30 centimetres away from the wheel. The motion impulses of its extremities in thin air caused it to fall over. This example illustrates the central problems in today’s robotics: how can machines be made to engage meaningfully with their environment, to interact with it and react to new situations? If the answers to these questions are still going to take a while, how are today’s automatic idiots supposed to already help us to save the world? By putting them in the hands of well-intentioned foster parents. So let’s take a look at who is driving research forward in the field of robotics. To begin with, there are the national, researchoriented funding bodies: in Germany, the German Research Foundation and the Federal Ministry of Education and Research; in the USA, the National Science Foundation. In addition, there are independent research institutes: the Max Planck Institutes, NICTA in Australia, INRIA in France, and so on. They offer traditional research funding, with grants being awarded according to scientific rather than economic or social criteria. Social aspects in this type of funding are thus dependent on the personal interests and commitment of the researchers. 

Drones don’t need roads in order to transport blood samples
The robotics community is now focusing heavily on technical developments and less on the possible impact of the research. The possibilities for specific applications are of secondary importance at this level of fundamental research. In the US, the military is another major funder of robotics. Most of the money comes from the Defense Advanced Research Projects Agency (DARPA), which also finances the Grand Challenge mentioned above, and which early on invested a lot of money in the development of autonomous cars. DARPA not only finances research projects that can be used for military purposes, but also fundamental research. However, their interests are quite clear. However military-driven research can have unexpected effects, since the conditions of military and disaster-relief operations are similar. And that’s why the robots that are being used in projects with social and environmental relevance tend to be drones. For example, transporting blood samples from newborn children in Malawi to the hospital for UNICEF, for the early detection and treatment of HIV infections. 
We still know and understand very little

In 2014, the Drone Adventures project – which in the 2016 trend radar served as an example for our “drones for good” trend – was able to create important drone-generated maps of the disaster zone after Typhoon Haiyan. In the meantime, Drone Adventures has also contributed to saving sharks, again with maps, this time of the St. Joseph Atoll in the Seychelles. The WeRobotics initiative gives social projects access to the possibilities of robotics and encourages them to experiment – so far with a focus on drones for transport and cartographic tasks. Currently, there isn’t much besides drones in the robo-social sector. But that’s something. What the projects have in common is that they have emerged from the personal initiative of researchers who managed to win over sponsors with their ideas. Since applied robotics is still a young field, there are only a few initiatives aimed at social impact or applications. The research community is focused primarily on overall progress, and less on specific, realworld applications. 


Looking at the current state of affairs in robotics, we could summarise it as follows: we still know and understand very little. As such, it is difficult to predict what’s going to happen over the next few years. A wave of euphoria is now spreading among machine learning experts, for there have been many new successes to celebrate. And who knows, perhaps these new findings will also give a boost to robotics. But whether and how robotics will benefit social, environmental or development projects in the coming years is still uncertain. For the moment, our robots are too simple. And too arduous to grant them even an iota of understanding.