6 Branches Of Artificial Intelligence You May Want To Explore

Artificial Intelligence (AI) and Machine Learning (ML) are the buzzwords of this decade and arouse a lot of interest and curiosity. It does appear that a large part of our lives are surrounded by multiple applications of AI and ML. The two terms are often used interchangeably, but machine learning is a subset of artificial intelligence which is a broader term and involves multiple disciplines working in conjunction. Let’s break down the terms and try to understand some basic terminologies.

What Is Artificial Intelligence (AI)?

Artificial intelligence, as the term suggests refers to non-natural (artificial) intelligence, or resembling an ability to think or understand. AI is not a system, but a field of study explaining how one would be able to build systems that can do things that humans do better, currently. It would mean enabling machines to have some capabilities which humans possess. The recommendations that you see on an OTT platform, e-commerce website, or a music streaming application are common use cases of the implementation of artificial intelligence through algorithms. Some other examples are Siri on iPhones or Alexa in Amazon home devices.

Let us now dive into some common problems that are currently being solved using artificial intelligence. This also gives us some insight into the different areas one can find their interest in and develop expertise in.

Also Read | Skills To Develop For Better Learning At Law College

Classification & Statistical Learning

This would involve pattern recognition or identifying which of a set of categories an observation belongs to. One example could be how certain emails end up in the spam folder without any manual rules set by the email user. Another example would be optical character recognition available on phones these days, where you can extract text out of an image. Some good starting points to understand this branch of artificial intelligence would be to grasp basic concepts of statistics and probability. Usually these are covered as parts of the mathematics curriculum in schools. One can also browse through the “Statistical Learning- Classification” course offered by University of Waterloo on their website to get an understanding of the nitty gritties of the field.

Also Read | ‘Programming Essential To Many CS Specialisations’: JK Lakshmipat University VC

Motion & Manipulation In Robotics

Artificial intelligence is becoming more and more important in the automation industry, especially in the field of motion control. You must have seen numerous videos by Google-acquired Boston Dynamics, where robots dance and it makes for great entertainment. There is a lot more that can be achieved in this space. As businesses migrate to software-based motion controllers, AI can help analyse a rotary bearing in a robotic machine so that it can forecast future abnormal behaviour. Using maths procedures, AI can predict when a rotary bearing may fail.

Artificial intelligence is a wide field with multiple subfields and many more are likely to get added in the future since the use cases and the areas of implementation of artificial intelligence continue to rise.

Speech Recognition

Speech recognition enables identifying and interpreting words and phrases in spoken language and converting them into texts. It is a subfield of computational linguistics that deals with technologies to allow spoken input into systems. Natural Language Processing can not be called an algorithm under the broader umbrella of speech recognition, but a subfield of AI in which machines are enabled to understand text and speech just the way humans do. It provides multiple ideas to approach the general use cases under speech recognition. Quite a few mobile devices use speech recognition in their voice-based searches - for instance, Siri or Cortana, which is an example of something referred to as virtual assistants.

Also Read | 6 Evolving Tech Careers You Must Look Out For

Fuzzy Logic

Fuzzy logic helps in developing systems which mimic how humans would think. In most of our daily lives, we do not have straight ‘yes’ or ‘no’ answers. The term ‘fuzzy’ would mean vague or unclear. Thus, such systems are smart enough to provide answers in scenarios where a true or false value won’t suffice. Situations, where a range of values can work as an acceptable answer, are a good use of fuzzy logic. For instance, you can be slightly thirsty, but not parched. Modern washing machines are a good example of fuzzy logic since they adjust the volume of water and temperature basis of the load in the machine. There are no set maximum or minimum values in this case. Similarly, air conditioners these days also adjust their energy consumption by sensing fluctuations.

Search & Optimisations

In a game, like DOTA or Call of Duty or Chess, there are multiple strategies to reach an end state and there is always a most optimum way of reaching that end state. Search algorithms exist to solve such problems. They explore all the possible paths to reach an end state from a start point and opt for the one with the lowest cost. The cost can be either the minimum number of moves to win , like in Chess or in the context of a strategy game, would be a path involving less damage. You can browse through the archives of Cornell University to go in-depth around how chess is programmed using search algorithms. Some examples of such algorithms would be Min-max Searching, A* etc. This in itself is a separate subfield of artificial intelligence.

Also Read | Career Opportunities For Engineers Of The Future

Neural Networks & Deep Learning

Neural networks are a series of algorithms that mimic the operations of a human brain to recognise relationships between vast amounts of data. Self-driving cars are a great example of neural networks. It is called a network because they are similar to how neurons in a human brain form a network of nerves, which enables our brain to form patterns and learn from the consequences of an action. These algorithms are also modelled to do something similar. The difference between neural networks and deep learning algorithms is that deep learning algorithms include more than 3 layers ( input layer + output layer + some hidden layer ). Deep learning models are used in computer vision algorithms and are the crux for the famous face swap trends one can see on social media.

Artificial intelligence is a wide field with multiple subfields and many more are likely to get added in the future since the use cases and the areas of implementation of artificial intelligence continue to rise. AI combined with data is a huge tool and asset in understanding and automating multiple problems and making human lives better. This article is meant to give a glimpse of the areas of application of artificial intelligence and the broader sub fields so that students can explore the area with more awareness.

Deboshree holds a BTech in Computer Science and Engineering from BIT Mesra. Backed with 6 years of experience working with Goldman Sachs and Walmart, she currently works with Cred as backend engineer.