Machine Learning

The world of machine learning and artificial intelligence can be enriching and intimidating at the same time with almost every 3rd article you read about these topics claiming how advancements in machine learning and AI would be conducive to the destruction of humanity-"These machines will steal our jobs and render us all unemployed, they will take over the world once they develop consciousness"- and so on. But is that really what machine learning is all about? What is machine learning though?

Ever wonder how Netflix is able to recommend you a movie you'll love or how Spotify is able to recommend you songs based on your daily tunes? Or how does Gmail differ spam mails from your regular mails? All these functions are a few of the basic tasks of machine learning. In simple words, machine learning means developing algorithms which help computers learn from past data and produce desired results based off of that data, without having to explicitly program them to do so- just like how we humans learn from our past mistakes and thrive to perform better in the future. So when Netflix or Spotify are able to accurately recommend you movies or songs, they do so based on your past streaming. They keep a record of the kind of movies or songs you stream and then give you new recommendations that you would like accordingly. This is called recommendations systems. And when Gmail seems to be able to filter out spam mails from the regular ones, it is called spam filtering. Other examples of machine learning include image/facial recognition, fraud detection and natural language processing which is how you are able to converse with ChatGPT and Google Gemini (previously Bard) as if you're chatting with another human being.

But how is machine learning able to perform these tasks? For the uninitiated, the kind of data an algorithm is designed for and trained on heavily influences the functions it can perform. Recommendations systems, spam filtering, image/facial recognition, fraud detection and natural language processing all work with labeled data being fed into the algorithm. This kind of learning is called supervised learning where the model is taught to predict the output for new data based on the learnings from past input-output relationships.

There are various other types of machine learning which I will touch upon in subsequent posts. Until then remember, machines and humans have a collaborative relationship than a competitive one.