RoboEarth Released, but is it Safe?
After four years of European Union-funded work, scientist of the Dutch University of Eindhoven released today their project dubbed "RoboEarth", the first Internet designed specifically for robots. RoboEarth is a cloud based database which serves as knowledge base and allows connected robots to communicate with and to learn from each other. RoboEarth was presented to the public by four robots which worked collaboratively together to help patients in a hospital in a simulated enviroment. They can see maps that were created by other robots to self-navigate in an unknown environment. However, the information exchange goes much deeper; as soon as a robot learns a new object or a new task other robots that are connected to RoboEarth can view this information and learn from it without any further programming.
However such advancements, while full of great potential, always come with great risk. To the typical techie, the first thought that probably comes to mind is "Skynet". There is a great fear that when all robots are able to learn and grow independently of humans, and additionally when they can communicate from all across the world, they will become dominant. An expert on the topic, James Barrat, addressed this by stating:
In the short term, RoboEarth adds security by building in a single point of failure for all participating robots. In the longer term, watch out when any of the nodes can evolve or otherwise improve their own software. The consequences of sharing that capability with the central 'mind' should be explored before it happens.
This statement is extremely vague, but it is likely trying to state that this network would need to be very closely observed. It should become very limited in its capabilities, and have restrictions to what type of information can be shared in order to ensure any degree of security in this regard.
In addition to the "robot uprising" topic, there are more present-day security issues to address with this type of advancement as well. There will need to be heavy safeguards to prevent hackers from corrupting and stealing information contained within this network. If, for example, a cyber terrorist were to 'spoof' the knowledge of a newly-learned task regarding the preparation of food for household robots, and they were to include the instruction that potassium cyanide should be included as an ingredient, there would be quite a problem if all robots were to accept these instructions. In addition to that, this network could potentially contain extensive information about the day-to-day happenings within every household across the world, which would be extremely tempting for groups such as the NSA.
Additionally, it may be too soon to celebrate the collective hivemind. It's relatively likely that the robots currently testing this network are fairly similar in technology to each other, but in order for RoboEarth to be widely used, every robot would need to be able to very specifically describe what type of hardware it possesses and to be able to determine if new tasks that another robot learned would be compatible with its own technology. In addition, each robot would either need to select a subset of tasks and knowledge to download to its local memory (because RoboEarth would contain far too much data for each robot to store individually), or each robot would require a connection to RoboEarth in order to perform any tasks (meaning that when unable to establish a connection, the robot is useless).
RoboEarth is still an excellent idea and has a lot of potential. Finding a way for robots to learn independently is the natural course of events, and it will assuredly continue. For more information about how RoboEarth works, see their website, roboearth.org.