Drawing the Line Between Artificial Intelligence and Machine Learning

Can we really imagine AI without ML today? These days, AI and ML are used interchangeably so often that we can no longer tell them apart. In my previous article, I had talked about the misconceptions pertaining to AI and the dilemma faced by the industry as to whether AI is for the better or the worse. In this article, I’ll mainly focus on how different AI and ML are and how significant they can be if they go hand in hand rather than individually.

Contrary to popular belief, these terms are not synonymous- rather, they bring meaning to each other. AI can be interpreted as the concept of machines performing tasks in a way humans can categorize as smart. In AI we are trying to program the computer using rules, algorithms or specified paths such that it can perform or rather, mimic a human to an unimaginable extent. We wish to give it the human decision making and perhaps, thinking capability by providing it with instructions directly or indirectly.

Image Source- https://bit.ly/2kgI7xo

AI is a concept that was prevalent much before ML. AI can be implemented without ML as in the case of GOFAI(Good Old Fashioned AI), a term coined by John Haugeland in his 1985 book Artificial Intelligence. GOFAI can not only implement AI without ML but if used with ML it can also increase the efficiency to a great extent. Even today many companies continue to use GOFAI rather than ML.

Machine Learning can be understood as the process of literally a machine teaching itself or learning on its own. It is easier to understand this concept through an analogy perhaps wherein a program like a student is provided data (books in case of the student) and uses the given material to teach itself. The idea is for us to no longer need to define algorithms for everything rather the program to learn on its own what is the best result. In ML, instead of we writing algorithm to handle every scenario, in machine learning algorithms get trained over time to handle more than what they were initially programmed for. The benefit of this is that we no longer need to worry about the loopholes in our defined rules. Over time, programs takes care of such things by themselves.

Image Source- https://bit.ly/2GFUlID

To conclude my article, I present an attempt towards resolving this conflict as to when to trust ML and when to not. In my opinion, it would be best if we were to develop a confidence factor which could define a threshold such that if the confidence factor was above it, human intervention could be considered trivial and below it, human intervention could be employed. We are yet to see how long it takes for human intervention to become obsolete.

#IndiaStudents #StudentVoices