From Layman to Deep Learning Expert: | Natural Language Institute

Embracing Singularity

BLOG

From Layman to Deep Learning Expert:

1. Why?

By Victor - 30/aug/2019 #Science and Technology

When I read the astonishing news about AlphaZero having surpassed centuries of human and computer understanding about chess in the course of just four hours, I thought to myself: “This technology has the potential to change the world. I’m going to learn it.”

Never mind that I was working in international relations; that I had majored in Political Science and was getting a Law degree; that I had a language institute; that on the weekends I managed a farm; that I was 39 years old and had no relevant background in mathematics or computer science.

Truth be told, I think I had been hankering for an excuse to change gears in my professional and academic pursuits. I would occasionally reflect that, growing up, I had been equally inclined to numbers and language, but due to familial influence, I had never seriously considered studying math or the exact sciences deeply. As an adult, I had spent the past 17 years mostly managing projects and people and I now felt pulled toward the purity and power of numbers, physical laws, and algorithms. Sometimes I would comment to friends, “If I could start all over again, I would study physics or math, because they underlie how everything operates in the universe.”

Another alternate-reality fantasy I harbored was having become a computer programmer. For a few months when I was about 10 years old, I had delighted in teaching myself to program in BASIC[1] on the Commodore Amiga[2] I had convinced my mom to buy us kids. But with no one to help me, no Internet, and no concrete opportunities to take the hobby further, I gave it up like the chess that I was also passionate about at the time.

Subsequently, my experience with programming was sparse. Seven years after my short-lived foray into BASIC, I programmed a couple trivial games on the Texas Instruments calculators we were loaned in high school Calculus class. Another seven years after that, I dabbled in the most superficial Visual Basic code, SQL, and HTML to gain a little flexibility in the Word macros, Access databases, and primitive website that I had created to power my small enterprise—The Natural Language Institute ( “Natural” for short). I should note I was also answering phones, mopping floors, teaching English, and pulling all-nighters to turn in translations on time and pay back my student loans.

After turning over management of Natural to others when I entered public service, I tried to guide different directors in hiring IT staff or ordering custom IT solutions, but the results were consistently disappointing. Meanwhile, online classes and language apps became increasingly popular and Natural was missing golden opportunities for growth. Perhaps the only way to provide better guidance to managers so that they could find a path forward on the technological front would be to immerse myself personally in the field and in the milieu. But how?

At my day job, as head of international relations, my department was hampered by staff constraints, and I realized this problem would increasingly affect the entire audit court where I work. A constitutional amendment passed in Brazil at the end of 2016 to limit public expenditure had the concrete effect of severely restricting the hiring of new civil servants, which meant that the scores of auditors who were retiring could not be replaced. Yet civil society, outraged by recently exposed corruption scandals in the country, rightly expected our office to do more to ensure federal funds were being spent ethically and effectively. How could we do much more with a shrinking staff? How could I contribute to meeting this tremendous challenge?

The news about AlphaZero, like a flash of lightning, brightly illuminated the path I had been scoping out for some time. Advances in artificial intelligence were beginning to change the world; they might help solve our audit office’s pressing productivity problem and make language instruction ever more accessible, efficient, and customized.  

That was it! I myself would become a computer programmer. Doing so would provide me with a new avenue to help my audit office meet its most pressing challenges. At the same time, it would enable me to successfully guide managers in hiring professional programmers and developing software applications at Natural. Ultimately, it would allow me to comprehend and even help develop artificial intelligence applications that could have a tremendous positive impact on society.

AlphaZero had used neural networks in a deep learning model to master chess, and I immediately understood that deep learning would be at the forefront of many transformational technological breakthroughs. I had decided: I would ply my way from layman to deep learning expert. The only question was how, but that will be the topic of subsequent posts.



[1] A general-purpose programming language often used by beginners and amateurs, especially through the 1980s.

[2] A popular brand of personal computers in the late 1980s and early 1990s.

This publication is the first in the series From Layman to Deep Learning Expert.