(Photo Credit: Dr. Mandana Sassanfar from MIT).
Gil Dekel will never be lost at sea. His rudder has always instinctively steered him towards his many passions. Durtti wants to find out about those passions and what Gil’s learnt on his journey so far.
You are a very competent bass player and composer and you had a successful career as a musician for 10 years before embarking on a career in computer science. Was the transition a sudden one or was the ‘coder’ always there inside you waiting in the wings to take center stage, Gil?!
In a way he was! The transition came out of necessity. It was the realization that while I was (and still am) very passionate about bass, music, and live performance, I was not happy doing it professionally.
Back in high-school I took some advanced classes in computer science. Although I was completely focused on music during those years, coding resonated with me. Specifically, the problem-solving aspect and how our teacher was able to optimize our solutions to elegant, one-liner pieces of code.
So when I was exploring my options in college I took an introductory course to object-oriented programming in Java and immediately fell in love. It felt as if my brain was wired for this stuff. I declared my major in computer science the following semester and never looked back.
Now a Computer Science Tutor, what is the one thing that never ceases to amaze you about the students that you teach?
It is the stark cutoff between true problem-solvers and those who are are just trying to get into the promised land of the tech industry.
This is not to say that non-problem-solvers lack intelligence or slack off.
To become a successful coder/engineer/developer/computer scientist you must be, first and foremost, a devoted problem solver. Everything else will come with time. It is just a matter of passion; the light is either on or off.
Who has had the biggest influence in your life to date?
I am a strong believer of the mentor-mentee system. To clarify, this is not what you usually find in schools and colleges. It is a one-to-one relationship between a professional of your field and those she takes under her wings. And so those who had the biggest influence over me were my mentors.
Naturally, the first two are my parents who guide me through life and support all my shenanigans.
Then came my music mentors who, amongst other things, turned me into an autodidact. Musicians must be able to self-educate themselves on a daily-basis. Learning new songs, figure out new exercise routines to overcome technical (performance) difficulties, learning new techniques and/or instruments, and more. This is an invaluable skill I am very thankful for.
The person who has sparked my interest in artificial intelligence and taken me as her mentee is Professor Susan L. Epstein who runs Hunter College’s Problem Solving and Machine Learning laboratory. I currently take an active part in her lab as an undergraduate research assistant and look into the problem of autonomous robot navigation.
Finally, Professor Epstein introduced me to MIT’s Diversity and Outreach Coordinator, Dr. Mandana Sassanfar, who informed me about different opportunities at MIT and guides me through the application process.
Part of that process has lead to you working towards admittance to MIT’s Computer Science and Artificial Intelligence PhD program. Which industries do you believe will be most commercially successful at adopting artificial intelligence solutions in the next 3-5 years?
I would like to believe that within 3-5 years we will see self-driving cars completely disrupting the car market.
Similarly, I hope that drones will be utilized for long-distance delivery in companies like Amazon to significantly reduce delivery times. So that’s automation for you!
In addition, Virtual and Augmented Realities (VR/AR) will find their way into houses and hospitals in the forms of gaming/media and therapeutic methods (e.g. for the mentally ill and/or lonely elder.) Although VR/AR are more tech than artificial intelligence per se, their software must deal with human day-to-day challenges, which requires intelligent software.
Finally, there is a lot of research done in human-robot interaction. Hopefully, intelligent communicative systems will become common practice in anything from fridges, to smart TVs, your personal computers, and spacecraft control systems.
As humans, we all have good days and bad days! How do you go about turning things around when they are not going your way?
When things go bad I get in my own head. And when I do, I panic.
However, I try to look at it as a different source of energy, and what is all this mental fuss if not some force pushing things in the wrong direction? It may sound like a bit of a cliché, but I channel the negativity and worry towards productivity. I immediately spring into action when things go awry, even if it’s 3:47am. That way I am too busy to get stuck in my head.
Getting things done is a much better feeling than brooding over my troubles and misfortune.
Tell us about the neural net implementation for C++ you have created.
Towards the end of summer 2016 I got a copy of a book called Make Your Own Neural Network by Tariq Rashid.
This book is a great introduction to neural nets and I highly recommend it. However, because it uses Python as the programming language, a lot of low-level challenges are abstracted away from the reader (like matrix manipulations and algebra).
So my implementation was a learning experience of neural nets and how to implement one in C++.
I read the book cover-to-cover to understand how neural nets work, implemented one, and packaged it in a C++ library. It requires no installation! Just download and #Include. The library, which I named N++, includes the neural net implementation and a Matrix class for basic linear algebra operations. The hyper parameters of NN++ include (1) number of input nodes, (2) number of nodes per hidden layer, (3) number of output nodes, (4) number of hidden layers, and (5) the learning rate. Performance of the net was tested over the MNIST dataset (where the model is asked to classify handwritten digits from 0 to 9) and showed accuracy of [96.0533% ± 0.139499%] with 95% confidence.
Can you see yourself ever creating and wearing a pair of 3D printed shoes one day?!
Absolutely. I mean, I see myself living in a 3D printed house (those exist, look it up) with 3D printed furniture and eating 3D printed food!
Heck, I will even accept 3D printed organs if I ever need that done (really smart people are working on that as we speak).
Finally Gil, what’s the most valuable piece of business advice you have ever been given?
It was my mother who told me that if I spend so much time thinking over and doing something, I might as well do it to the best of my capabilities.
Go big or go home. I followed that advice in music, and I follow it in now in my new career.
You can see Gil’s simple self contained neural network for C++ here.
Gil is a member of The Artificial Intelligence Group on LinkedIn.