Article Preview
TopIntroduction
There was once a time in history when people could not have imagined sitting idly on their couch and getting information about almost anything that came across their minds with a few simple touches on a little device in their hand. Even the idea of a machine that could do all of our computational work in a matter of seconds was unfathomable.
But we are now living in a century where these technological dreams can be brought to life. Nothing seems impossible now. From health care to education, science, agriculture, and almost every possible discipline out there, there are innovations and newer technologies coming up each day to make people’s lives easier and more efficient. The study of robotics is something very complex, but at the same time, equally intriguing and fruitful. The entire world could change exponentially if a day comes when people have successfully developed robots that are not just meant for jobs like cleaning and cooking, but ones that look, learn, and behave just like humans. Imagine the amount of time people would save if these machines could do all of the tedious, manual work that people typically do. Honestly, it would be a little frightening too, since most people have watched or heard of at least one movie where machines have turned against humans and taken over the world. It would also be unfair to acknowledge the various attempts that have been made over the years at creating humanoid robots that are similar to humans in some of their attributes. Modern robots are mostly capable of handling repetitive jobs that do not really require intelligence or decision making. Also, the majority of robots are limited to labs and manufacturing units, and they are not really competitive enough to be able to work in the real world. Working in a natural environment with several factors that affect a person’s decisions and requires thinking spontaneously; these characteristics are difficult for robots because they are working based on a set of algorithms and can easily get confused in real-world situations (Rivlin, 2019).
There have been some fascinating discoveries in this field, but the one common feature that lacked in all of them was that of judgment or what has been famously coined as “theory of mind” (Scassellati, 2002). Another term that could explain this is self-awareness. The machine’s source of information recognition and sensing relies on human input (Krening & Feigh, 2018), and they require datasets for reference. For example, for a robot to identify a ball, humans provide it with several pictures of balls so that it can relate those images to identify one when it sees it. In the end, it comes down to a series of 0s and 1s, which is the only thing robots are programmed to comprehend (Lauckner & Lintner, n.d.). How do people teach robots to do complex actions? Robots are used to 'supervised learning', which means that they are aware of the output they should expect during any given task for specific input. This is where reinforcement learning comes in. Reinforcement learning is associated with deep learning that teaches systems to solve problems using trial and error methods. It is used when complex decision making is required and is therefore iterative in nature due to the multiple training algorithms it uses to come to a solution. This is because the environment it is dealing with is dynamic or unstructured in nature. The environment that the robot is adapted to work in is a structured environment where everything is ordered and in place (Govers, 2018). Thus, this technique is more time consuming because, for every given task, all possible scenarios are checked to obtain a result. But, for the same reason, it has a higher accuracy rate. Therefore, it is being heavily researched and implemented in robotics. Next, the authors provide a deeper understanding of this discipline.