Machine Learning and AI

If you come to this page you have essentially understood what AI is about. Now we need to understand the concept of machine learning behind the working of AI.

As the name suggests, the concept is based on machines learning from data. Machine learning is developed using data and algorithms. It focuses on making decisions based on the experiential data without actually being programmed to do so.
This branch is closely related to computational statistics, where predictions are based on statistical data. Consider a few examples:

Example 1: Face recognition
Law enforcement agencies used facial recognition technology to identify people wanted for crimes. There are online casinos which allow people to log in based on facial recognition technology. In 2019, Brit Awards used facial recognition to enhance security checks at the event.
Face recognition technology is based on machine learning algorithms that capture, analyze and compare the facial features with an existing database containing the photo of the person to be allowed entry.

If Pallavi had gone to Brit Awards event and did not have her details registered in the database, the face recognition technology would not allow her entry.

Example 2: Recommendations on Netflix

Pallavi likes to watch thrillers on Netflix show. The content that Pallavi watches on Netflix can be described in terms of Suspense and Violence. Now let’s see the input data categorized into the below graph. If a show falls in the green circle, ranging high in Violence and Suspense, Pallavi is most likely to watch it and the content in the red circle is least likely to be viewed. But what happens if the content intensity falls in the middle of two circles.

Here machine learning comes into picture. It would determine the proximity of the new content to the content most watched, and predict the desirability factor of the content as positive.

Example 3: Google Maps
A real time example of machine learning technique in AI is demonstrated by Google maps. People all over the world traverse almost 1 billion km using Google maps. The data related to all the routes is fed from the local government. AI technology learns from the past traffic patterns to predict busy routes with heavy traffic. It combines the past data with the current speeds of vehicles e.g. cars are traveling at a sped of 25 km/hour on a freeway allowing 80 km/hour. After predicting the busy route, AI determines the alternate route with lesser traffic and suggests the same. Thus the prediction of busy routes and suggestions of alternate route is based on authoritative data from local government and feedback from real time users.

Machine learning develops prediction model based on data to provide outcome. Extensive input data will lead to more refined prediction models and a more accurate result.

Machine learning can follow any of the three methods:

• Supervised Machine Learning: The machine is fed with labeled input data and the corresponding output data.
  e.g. Above example of Facial Recognition Technology can be an example of supervised machine learning where the facial features are compared to pre-existing database and the results are shown.

• Unsupervised Learning: In this type of learning, the AI interprets the pattern of the data from the unlabeled input. Netflix recommendations is a good example of unsupervised learning.

• Reinforcement Learning: This is a reward based or feedback based learning. Self driving cars work on the principle of reinforcement learning.

Applications of Machine Learning: