What is a Robot?
In a previous blog post I delved a little into the book by Isaac Asimov “I, Robot” and his three laws of robotics. Although interesting, this book remains fiction and our washing machines/actual attempts at robots are not really programmed with these laws [that we are aware of]. Asimov, as fantastic as his work is, imagined robots working on Space Stations, autonomous working robots and even robots that could control other robots and sadly (or not?) we are just not there.
But what IS a robot? What is the definition of what is and is not a robot? Some may argue that it doesn’t matter, but as we continue to progress there will have to be definitions or we cannot control policy/laws in robotics. There is this idea that we are entering the second industrial revolution. In the first, man [muscle] power was replaced by heat engines. In the second, he will be replaced by programmed machines and automatic devices [an extremely interesting argument]. The only issue is … no one can agree on a definition of a robot, at the time of writing. Ask a group of roboticists and most will give you different answers. However, although it can vary wildly, there are a few things that appear again and again in peoples’ definition of what IS a robot, which I will discuss below.
The dictionary definition of a robot appears to be quite vague “a machine resembling a human being and able to replicate certain human movements and functions automatically.” What do others think?
An opinion piece by author M.W.Thring (Nature, 1966) immediately opens with the statement that a robot is “any self-regulating artefact which operates without human aid”. By this description, could we not argue that our kettles, washing machines and computers are robots? Although we push some buttons, the kettle is able to know when it is boiled without us telling it, our washing machine is able to go through cycles without us instructing it which cycle is next. M.W Thring was ahead of me on this one and continues on to explain that a robot is not an automatic machine or a computer. These machines are ruled by numbers and they are very limited to what they can achieve and can only carry out tasks as they are instructed. They cannot sense their external environment, reassess, reprogram themselves and continue on in [possibly] a completely different direction. In comparison, a true robot, though also restricted by human made programs and numbers, has the ability to sense its environment and [the important part] ADAPT to the new information. Thring goes on to discuss why robots are not, and never will be human. Despite being able to adapt to its environment, a robot is still restricted to the scope of its program. They lack emotion and [arguably as a direct impact] lack creativity, which Thring argues is what sets us apart from robots. A robot will never create a piece of art or think up a new hypothesis: nothing creative was made purely by logic and robots are logical beings.
Alan Winfield (Robotics: A very Short Introduction, 2012) delves slightly more into it, rather than defining a robot as something that has the ability to sense he lists five functions that all robots have:
- Energy (integrated into the body)
Further, a robot can be three things (as defined by Winfield)
- “An artificial device that can sense its environment and purposefully act on or in that environment”
- “An embodied artificial intelligence”
- “A machine that can autonomously carry out useful work”
I would agree with Winfield when he specifically mentions that it is embodied AI vs. a broad AI heading. To me, a robot is a physical being not a “cloud”. Central to Winfield’s definition, is this idea of autonomy [an ability to govern oneself and make decisions, literally translated “self-law”] and intelligence. A robot doesn’t need a high level of intelligence to be classed a robot, but it does need to have some semblance of autonomy. A remote controlled car, for example, would not be classed as a robot. But a toy car that can move itself, sense obstacles and make the decision to go around them has low intelligence but is classed as an autonomous robot.
Central to these definitions are the ideas of autonomy, the ability to sense and act and [to some degree] intelligence. Many relate the definition of a robot to the Sense-Think-Act paradigm. A robot can sense its environment, think about what it has sensed and then act upon this. There are a few add ons I have also noted, Mel Siegel (Professor of The Robotics Institute at Carnegie Mellon University ) adds to this “communicates” and some also add that a robot should be able to sense what a human wants and optimise it. For example, roboticist Anca Dragan of UC Berkeley (in Wired, 2018) gives the example of a robot arm. It is preprogrammed to always be at a certain height, but if you want your cup of tea lower it creates an issue. You can push down on the robot arm, but it has not been programmed to be this low, so we need another type of robot that is sensitive to a human beings prompts [like pushing down its arm to tell it to sweep lower with your cuppa].
The central themes remain though that it should be a physical being that can sense its environment, make decisions depending on this and act on them. Would you add anything to the mix? What would your definition of a robot be? Aside from definitions, I find the points Thring made resonated with me. When someone asks me if robots will take over the world, at the moment my view is no because they will only ever work within the scope of what we program them to do. They are without free will [at the time of writing] and creativity and I would argue that they would have to be pretty creative to take over the Earth. Maybe we’ll find out?