This article is narrated by Dr. Sridhar Mahadevan, Ph.D. in Computer Science from Rutgers University (1990), Professor at the University of Massachusetts and Director of the Data Science Lab in San Jose.
Table of Contents
Hofstadter’s first book
Let us answer this question first: Is machine learning really difficult? no
I wrote my first machine learning program in 1982, before the Internet, computer GPUs, laptops, Google, cell phones, digital cameras, and desktop PCs existed, and before almost anything in the world of technology that surrounds us today. You see there is.
So how did I discover the existence of such a thing?
At that time, they read books to learn. Of course, you had to go to the library, but in my case, a strange event called a book fair. At the time, I was studying electrical engineering at the Indian Institute of Technology (Kanpur).
In my opinion, this book is still the best, and it was very inspiring for me.
The book depicts a whole new fictional world in which there are deep connections between art, music, and abstract mathematics, represented by three main characters – Johann Sebastian Bach, Maurice Escher, Kurt Gödel – and computing that includes machine learning and neural intelligence. had been realized
The book features a series of visual puzzles from a Russian researcher named Bongard, in which the task is to discover an order that will separate the six figures on the left from those on the right. This is a basic problem in machine learning, which is called classification, and this task is like distinguishing email from spam or recognizing a face in an image.
As humans, we categorize sensory stimuli billions of times throughout our lives, and our survival depends on it. For example, when crossing the road, is the object approaching you a person or a truck? If you get the answer wrong your life is over and your mind solves such problems wonderfully.
Getting started with machine learning
I invite you to do this, I, without any training in this field, foolishly decided to make this the core of my master’s thesis. I got it, although it seems very naive.
Although I was familiar with Python programming, this experience made me realize that artificial intelligence and machine learning were my life goals, and I decided to immigrate to America in 1983. Collaborate with a talented person named Thomas Mitchell, who is now the Vice Chancellor of Computer Science at Carnegie Mellon University.
I learned an important lesson from Tom. That no book can teach you. Instead, the research is fun and at the same time very informative.
He researched harder than anyone I’ve ever met, and that lesson made a huge impression on me and has stayed with me ever since.
After getting my Ph.D., I joined IBM Watson Research in New York in late 1989, where they couldn’t figure out what a machine learning researcher would be good for so they moved me to a fledgling robotics group. I had no background in this field and had never programmed a robot.
However, it seemed that I was somewhat successful in this challenge, and started writing some articles based on my previous research, for example how robots can acquire new behaviors using reinforcement learning. Also, in 1993, I published the first book on learning robots, which included research from around the world in the field of artificial intelligence, and despite having no background in robotics, I managed to gain a relative reputation in this field.
Years later, I was elected as an elected member of the AAAI Association (the leading international professional society for artificial intelligence researchers), where a limited number of researchers are selected each year and the competition is fierce.
AAAI members this year and in 2020 include some of the founders of the deep learning revolution: Yoshua Bengio, and Yan Le Koon.
AAAI’s membership list has consistently included some of the greatest AI and machine learning researchers, and I am very fortunate to be part of such a distinguished community.
None of these things would have happened if, in 1982, I thought it was difficult and possible to do machine learning without formal training in this field, with basic calculations, or working in the field of robot learning at IBM without robotics training.
My recommendations
For those aspiring young researchers reading this, the best advice I can give is that nothing is “difficult” if you challenge yourself and get started. Above all, remember: that research is fun and engaging!
For many years, from 2001 to 2011, I was one of the pioneers of reinforcement learning in a lab at the University of Massachusetts. Andrew Barto and his Ph.D. student Rich Sutton were my colleagues and helped create the modern field of RL, the space that underpins Deep Mind and It was Alpha Go Zero, they helped.
Andy and Rich embodied the true spirit of researchers and enjoy doing research, and working with them was the best professional experience of my career. They hung a sign on the main door of the lab with a quote from one of the greatest scientists of all time. Albert Einstein was:
Ibn stated that “research does not require expertise” Einstein hated textbook knowledge. Above all, he valued the ability to dream and imagine. He advised parents of students: If you want to make your children smart, teach them fairy tales and witchcraft.
We are currently fighting the latest epidemic, the Wuhan coronavirus, and the greatest weapon we have is our ability to identify the genome sequence of the virus.
The greatest progress in biology in the 20th century was created by Watson and Crick, two unique biologists, and they advanced the world of biology with fun and entertainment!
Watson went on to write a very popular account of his discovery called the Double Helix. In that report, he told the story of how they were scandalized by prominent researchers, such as Oswald Avery of Columbia University when he found out that they did not even know basic biochemistry.
However, they cracked the secret of life by playing with 3D models stealing Rosalind Franklin’s collected data sets, and doing things they enjoyed and found fun!
So, in conclusion, my answer again is: No, machine learning is not difficult. it is fun! So, go ahead, explore, and have an open mind to learn.
Find some interesting projects that require machine learning skills to motivate yourself to learn and challenge yourself.
Useful resources to learn about machine learning
You can first use the following resources to get a proper understanding of machine learning and then start exploring machine learning applications and try to build your desired projects.
- Andrew Poker’s course in machine learning – machine learning Coursera
- Caltech Lectures on Machine Learning – YouTube Machine Learning Course – CS 156
- Nptel Video Lectures – Introduction to NPTEL Machine Learning Prof. S. Sarkar IIT – YouTube
- (Optional, if interested in Neural Networks) Deep Learning by Simon Haykins – https://cours.etsmtl.ca/sys843/R …