What is Artificial Intelligence?
Nowadays, almost everyone has encountered the term ‘Artificial Intelligence’ but there are various aspects of AI that can confuse people when they hear of it for the first time. Artificial Intelligence is, in fact a conglomeration of multiple fields, including Machine Learning, Deep Learning, Neural Networks and Machine Intelligence. The aim of this article is simply to try and explain these terms in layman’s language.
Artificial Intelligence – Overview.
Artificial Intelligence is a term coined in the 1950s to refer to the field of trying to make machines learn, think, act and process like humans. Machine Intelligence and Artificial Intelligence are inter-changeable names for the field, which includes language processing, computer vision, problem solving, robotics, pattern recognition etc. It also includes everyday apps like Google Now, Apple’s Siri, Amazon’s Alexa, your voice recognition doorbell and Google translate.
Machine Learning and its approach.
Machine Learning is a tool in AI with which we attempt to give a computer the ability to learn based on repeated input provided over time. This can be compared to a little child. If the child is shown a dog and told “This is a dog” over and over again, eventually, the child will be able to recognize a dog on its own. In the same way, a machine learns by storing all the data it receives, then comparing that with new data to arrive at a logical conclusion.
Machine Learning can be achieved through various means including Neural Networks and Deep Learning algorithms.
Neural Networks are algorithms that are structured similar to the neurons in the human brain. These networks find patterns in the data by comparing the input stream of data to ‘past experiences’. Neural Networks work by trying out approaches and finding the most optimum solution to a problem. If a new approach produces better results or a higher score, it is ‘mapped’ as being better and used the next time. This way, the computer learns to do things in a better way every time it does something, just like a human.
For example, a computer can learn to play Mario by itself using these networks if, say, it gets a higher score when it collects a mushroom, but a lower score when it fails to jump over a pit. The computer will then try various iterations of the game, each time, learning what to do to get a higher score. What it learns are mapped in the memory, much the same way as humans memorize stuff.
This is the next level in neural networks. Deep learning is an algorithm that breaks down the input data into various small data sets to analyze it more in depth and identify the underlying patterns. Using this, it tries to predict what should happen and take preemptive action. For example, in a driverless car, it is analyzing an image of the road ahead. Deep learning algorithms will then do a multi-layer analysis of the image by breaking it into multiple pixel sets of say 10 x 10 pixels. It will analyze each of these small sets of pixels to see if an object is present in in a straight line on the highway. If there is an object that is also in motion relative to the camera, it should appear to move from one such pixel square to another. So, the computer analyses the next pixel square to see if the object moved to it. The algorithm then learns the trends of motion of vehicles on the road and can predict what might happen based on multi-layer analysis of the image.
I hope you now have a better idea of what goes into making artificial intelligence. If you would like to learn more on the subject, the following articles and videos may help you. Happy learning!
Parakh M Gupta