The COVID-19 lockdowns have brought in a dry spell for students all across the world. For the seven-year-old Kautilya Katariya, however, this became an opportunity to brush up his knowledge of computing and coding.
“Computing is a bit like solving a puzzle because the operations are like pieces to make the puzzle and to solve the problem,” Kautilya tells me during our candid chat over a video call. The Indian boy residing in the UK has been declared the ‘youngest computer programmer’ in the world by the Guinness Book of Records. He has been invited to speak at multiple international AI conferences, too, and is slated to speak at the World Artificial Intelligence Summit in Amsterdam in October 2021.
His parents, father Ishwari and mother Trupti, proudly relate their son’s many achievements before I get into a one-to-one conversation with Kautilya. “What is artificial intelligence?” I ask, and he promptly replies, “AI is basically making a machine have intelligence, so it almost represents a human brain.”
I ask him if all machines are intelligent. “No,” he says, “first a person needs to program a machine, then it can be intelligent.” Playfully, I nudge him to give examples of ‘stupid machines’ then. Amused at first, he later raises his computer mouse and exclaims, “A mouse is stupid!” The little boy does know his AI.
Early beginnings
“How I started learning to code was when my Dad gave me this book about computing and I loved it so much that I finished it in only one day, built some basic computing programs and also installed Python,” says Kautilya.
His parents concur that most of his AI knowledge comes from reading books on the subject as coding hasn’t been taught to him at school nor was he given any specialised coaching.
“When I started coding, I used to do HTML, CSS, Python and JavaScript but now I only do Python,” Kautilya reveals.
He nonchalantly adds, “It’s very easy to learn AI. I could have started coding even when I was 4.5 years old, which is just one year before I actually started learning it.”
He even has a word of advice to other young kids his age who are desirous of learning how to code: “I suggest that they start with a block-based language like Scratch and then move to Python.”
Getting the grip on AI
The boy is as excited to talk to me about his AI learnings as most kids his age would be when speaking of cartoons or cars or chocolates. “One of my favourite things that I’ve learnt about is IBM Watson. Do you know that it actually beat Ken Jennings in Jeopardy?” he says.
And you know that he isn’t name-dropping when he can explain with conspicuous confidence the fundamentals of AI too.
“Data,” he says, “is not only numbers; it can also represent images, morse code, text.” He adds, “In fact, there are 30tn pages on google and there’s 1bn GB of data on Amazon, I think.”
“And what do they do with all this data?” I question.
“Netflix keeps its data in a model to see what you most like so they can give you a recommendation. I think it’s either machine learning or deep learning. And if it is machine learning, it’s probably supervised learning,” he responds.
He goes on. “Supervised learning is a type of machine learning, and there are two other types: unsupervised learning and reinforcement learning. In supervised learning, you give the data with the labels, whereas in unsupervised learning you try to get the labels with unlabelled data. And reinforcement learning is trial and error just like a human.”
He describes some other concepts to me, such as the Big O notation, Natural Language Processing and self-driving cars, with equal enthusiasm.
Playing around with algorithms like toys
“K-Nearest Neighbour is an algorithm that uses artificial intelligence and it finds the KNN to a point on a graph. I don’t mean a graph as in a graph of a line, I mean a scatter plot with dots,” says the young boy whose eyes light up as soon as I bring up the topic of AI algorithms.
Binary Search and Tim Sort are among his favourite algorithms, along with Interpolation Search and Smoothsort, he proclaims. He goes on to explain each of these in detail.
“Binary Search is my favourite because it’s the fastest searching algorithm and Interpolation Search is the fastest algorithm in the average case. Tim Sort is one of the fastest sorting algorithms and it is a hybrid stable sorting algorithm. I don’t really know what Smoothsort is but it was my favourite once I knew its time complexity and space complexity.”
After he narrates the mathematical formulae for these algorithms, I question to check if he’s aware of any real-world usage of algorithms.
“Amazon uses search algorithms to look for what you type in the search bar and so does Google.” That’s his instant answer.
Building with AI
He has used IBM Watson to build a program that can detect whether there’s smoke or fire. “I think it’s image recognition. I also think it can be used in places where there’s a high risk of fire so that the camera can be connected to my AI model and then it can check if there’s fire or smoke. And if there is, then it can call a firefighter.”
How did he train it? “I provide lots of data to my model. Some data to describe fire, some data to describe smoke and some to describe neutral,” he says. Kautilya’s now waiting to be a teenager to apply this model to a real-life situation.
He very proudly demonstrates for me the chatbot that he has deployed on his website. I quiz him about its working, and he says, “If you ask it something related to what I’ve already taught it, then it will be able to answer. That’s what AI does.”
“Does it keep getting better with time?” I ask. “Reinforcement learning does but IBM Watson uses supervised learning which doesn’t,” he answers confidently.
Kautilya’s dream project is to build a robot that can turn into almost anything. “I think I have an idea what to make it of: it’s either aluminium, or carbon nanotubes when scientists figure out how to make it longer without destroying the strength,” he says.
Source: indiaai.gov.in
