Application of Machine Learning in Artificial Intelligence
“AI is the new Electricity”
says one of the most famous and well-renowned computer scientists Andrew N.G. From our younger days, we have been hearing about this so-called AI and didn’t have a clear picture of what it is. Still, after so many years passed by, the first thing that pops in my head when I hear the term AI is about a robot who is so smart and had eventually conquered mankind. Yes of course this vivid picture in my imagination is all thanks to the movie IRobot, which reminds me of a famous quote from the movie itself.
“You are just a machine.
An imitation of life
Can a robot write a symphony?
Can a robot turn a canvas into a beautiful masterpiece?”
Technically speaking this would have been true at the time of the making which was way back in 2004. Fast forward to 2021, AI is making symphonies[1], AI is painting masterpieces[2], and many more inconceivable tasks. It is true why Andrew N.G claims AI is the new electricity. In the same manner, as how electricity transformed industries, AI is going to do the same. Ever wonder how AI suddenly got this superpower? The obvious answer is Machine Learning.
Simply put an AI is an intelligent machine/system that works without human intervention. The very idea is that it mimics human intelligence. But how a machine goes about being intelligent? Unless there is an approach, a machine will not be intelligent. The most proven and widely used approach used to achieve AI is Machine Learning. Before diving deep into how machine learning contributes to AI we should look into how AI was achieved before the coming of age in machine learning.
Known as the fight of the 20th century, the chess match between the world champion Garry Kasparov and IBMs Deep Blue[4] was the first time an AI competed against a human being. Deep Blue was able to comprehensively beat Kasparov as it was able to evaluate over 200 million moves in a second. This supercomputer used a tree search algorithm to evaluate the best possible move.
Another instance of AI being used without Machine Learning was when an advertising company named “Digital Equipment” used to build a so-called “expert system”[4]. Rather than focusing on general intelligence, this system had a narrow task. So it was programmed with very few rules.
The list goes on and on. But if we look a little deeper at all of these early applications of AI, they had one thing in common. It is that they were built using the top-down approach. Simply put they were programmed based on rules. AI scientist Rodney Brooks thought the better of this and argued for a bottom-up approach for making intelligent systems inspired by the advances in neuroscience[4]. This bottom-up approach marks a crucial juncture in the advancement not only in AI but in computer science as well as it mimics how human intuition works. This paved the way for a new programming paradigm. Since then, AI is being followed throughout the world using this new programming paradigm with the help of Machine Learning.
There are a lot of applications of Machine Learning in recent times. It is used so intensively that we don’t know where it is being used. Machine Learning has influenced so many modern discoveries in the world and would go on to do so in the future without a doubt. Before going into describing few applications of Machine Learning, I would like to describe a little about the learning approaches used in Machine Learning. Although there are many classifications of Machine Learning, human supervision-based classification is the most common. According to this, Machine Learning is classified as supervised and unsupervised learning. The data used in these two types of learning methods are different. Supervised learning uses labeled data and unsupervised learning uses data with no labels. Hence, the applications of these two types are very different. Therefore, let's discuss a few of them from each category.
Ever wondered how auto-generated captioning works on Youtube. Google’s AI team is doing many unimaginable tasks these days, and this is one of the basic applications of supervised learning. What happens here is the speech-to-text translation. Training data comprised of tons of short audio clips which are labeled and fed into the machine learning model to train. In this instance, it is a deep neural network that is behind all these wonders. The accuracy of this system is staggering and hardly gets a word wrong.
One of my favorite and most inspiring applications is the real-time sign language conversion by a Brazilian startup named "Hand Talk". Here what is exciting is that this affects millions of people all over the world. Isn’t that is just great that a small piece of software could make the life of millions of people better? This is all thanks to Machine Learning.
Top E-Commerce sites like Amazon, eBay gives us recommendations of products. This is a great example of unsupervised learning. The uses are clustered according to the data collected by the platform and it is fed into the recommender system to show us recommendations. Though this is quite annoying sometimes it still a very advanced feature.
Very recent advancements of Machine Learning are unimaginable such as a very recently developed app named Deep Nostalgia can bring life to old pictures[5]. It makes sense to use that name as it is indeed nostalgic. What more Machine Learning could do? I think the better question is What more Machine Learning cannot do? I think it will be very few. In the coming decades, the world will be fully transitioned to an AI-powered world without a doubt. As Andrew N.G claims AI will be the new electricity, and one might wonder what made AI be the new electricity. It is indeed Machine Learning.
References:
[1] https://www.nbcnews.com/.../ai-can-now-compose-pop-music...
[2] https://medium.com/.../generating-art-with-artificial...
[3] https://pub.towardsai.net/differences-between-ai-and...
[4] https://aiartists.org/ai-timeline-art
[5] https://zeenews.india.com/.../ai-based-deep-nostalgia-app....
0 Comments