The ability of computers to replace humans for a variety of tasks has aided several domains and industries. However, computers are still far from being able to function as humans because of the huge difference in the way the two entities learn and acquire knowledge. An article published in a journal named ‘Science’ reveals that a number of fundamental changes to the learning algorithm of computers need to effectuated in order to help them ‘learn’ like humans. This would essentially be done by accustoming the computers to trial and error methods and instinctive learning. The article was written by a freelance writer named Matthew Hutson who points to the commendable ability of deep learning networks to outdo humans.
Limitations of Robots
The biggest limitation of robots is the absence of common sense which often leads them to function anomalously when there is a slight variation in command. To exemplify, a chess playing robot may beat an intelligent human but the same robot would lose to a child in a game of checkers.
Learning like Humans
Hutson expounds that the essence of helping robots function like humans is embedded in training the robots to learn the way animals or humans do. With the help of research done by Gary Marcus, Hutson further explains that the design of the brain needs to be studied to understand cognitive patterns in humans. These human abilities, instincts, and behaviours need to be fed into the computers to help them acquire learning abilities similar to humans.