AI and ML: How to Get Started

When Your Coding Background Tends to Zero

You should see the looks I get when I tell people what I do for a living. With a confused twitch of the face, they request confirmation, “You do what?” I laugh, and reply, “Like I said, I’m an Instructional Designer”.

Well, an Instructional Designer, or Learning Experience Designer, is a talented professional who creates learning plans, learning courses, and positive learning experiences, while standing behind educational programs.

While I am a huge fan of my profession, I am also a huge fan of innovation. Claiming yourself as “Innovative” in today’s day and age without knowing AI is difficult, no doubt. AI attracts the smartest people, the largest investments, and Innovators like me.

I hesitated for a while before turning my hand to AI. I would tell myself things like, “You’re not smart enough”, “You don’t have any coding experience”, “You forgot all school math programs”. Then I discovered Daniel Bourke’s article, which inspired me to get started.

As an expert Learning Designer, I didn’t want to pay a bootcamp to do my own job for me. I had to take hold of my own learning, from creating my own curriculum to evaluating my own learning progress. I had to test how Learning Designer skills would help me in learning AI and ML.

My Goal Is Simple: To learn AI for 12 months, 2 hours each day. To start on January 1st, 2020, and cease on December 31st, 2020.

Now, while it’s easy to want to jump right onboard of this opportunity, this work is hugely exciting at the beginning, and gets less bright and pinky as soon as you think about the actual plan of action. Then panic sets in. You may ask yourself questions like, “What should I do first?” “Which courses should I take?” “Which books should I read?” “What programming language is used?” “How much math involved?” Sweat pooling on your forehead, chest getting tighter…

Learning Designers start with defining the current level of the target audience, together with their goals, when creating a learning program. This is a must for a quality learning curriculum, and should be the place to start if this is all new.

Define Your Current Level and Objectives

First, think about your starting point. For example, I am very good with numbers, and I always have been. In high school, I was top of the math class. But I finished school eight long years ago, and I have forgotten almost all the math concepts, even the multiplication table!

I finished school in Russia, and never spoke about math in English. I figured it would take me at least one whole month to revise the math school program and improve my English math vocabulary. If you’re an English speaker with a good grasp of complex math concepts, you could start right away.

Your level is the number one factor for your AI learning path, so it’s worth spending a few hours to describe your starting point in as much detail as possible.

What to Consider:

• How far does your knowledge stretch in math? Be specific.
• Your programming experience, even if it’s not Python.
• General understanding of AI.

Check out my example for inspiration.

Next, set a goal. Without a concrete learning objective, you’ll go nowhere.

It’s important to formulate your goals with action verbs. For example, my goal is to get a job in data science. “Get a job” is an action verb. It is also a measurable goal. That is, I can evaluate my success against this goal: I’ll succeed the best if I get a job within one year, a bit less if I get a job within a year and a half. I’ll fail if I don’t get a job at all.

Compare this with the following objective: “Learn Main AI Concepts”.

This isn’t a very good learning objective. It doesn’t go into detail on what main AI concepts are, and the verb choice implies that you learn in order to learn. Truth be told, we never learn in order to learn: we learn to get more selfish benefits. We learn to get more money, to increase our self-esteem, to get a more interesting job, to get a diploma, to impress our parents or friends, and so on. But we never learn for the final goal of learning.

Think of why you need AI, short-term and long-term. Be honest with yourself, and write this goal down.

My short-term goal is to get a job in data science because:

• First, it is paid more than a job in education.
• And second, I love the mindset of developers and want to be a member of their professional group.

My long-term goal is to apply AI knowledge to build an AI startup in learning and development.

Of course, these goals may change with time; but before they change, they will guide me through the ocean of learning uncertainty.

Once you determine your final goal, write out a list of sub-goals to help achieve your final goal. These can be:

• Revise the school algebra program.
• Get familiar with Python syntax.
• Finish Udacity Python course.

Once again, avoid verbs “learn” and “understand”, and make sure that the sub-goals are measurable. For example:

• I looked through half of school algebra concepts.
• I can list at least 10 Python syntax particularities.
• I finished 70% of Udacity Python course.

Once you are certain about your current level and your learning goal, it’s time to build your learning plan. This is the trickiest part…

How on earth do I build an AI learning path when I know nothing about AI?

When you think about it, people pay thousands of dollars for education. Why?

• Because university programs give a diploma.
• Because university programs give a step-by-step guide of what to do.

And the latter is really important. In fact, when learning designers create a learning curriculum, they first study the subject from A to Z. When it’s a 20-minute course on a particular electrical installation, it’s not so painful. However, if you need to create a one-year program for such a broad field like AI, well, good luck…

As a self-learner, you are in a very disadvantageous position because you have no clue what AI is, but you need to build a curriculum before you start learning.

This is very much a “Which came first? The chicken or the egg?” moment. A philosophical dilemma with no solution. An infinite loop. Big problem.

The good news is, there are some back alleyways, ready for you to discover.

Borrow

Borrow curriculums from universities, bootcamps, people who have already been through the process. This is free, and this will give you the first insight into what you need to learn.

I checked these curriculums to define pillar points (but there are many more!):

• Sussex Computer Science and AI
• Strathclyde Artificial Intelligence & Applications
• UCL Machine Learning
• San Francisco Data Science

I also found this useful:

• Here Siraj Raval presents the components of AI learning and suggests online courses in a logical order.
• Daniel Bourke gives a wonderful description of his experience.
• Here Rodrigo Beceiro lists AI domains in the section “Artificial Intelligence field”.

My Trello board shows the tasks I choose to do each week and includes courses and other materials I follow.

Another potential resource of information is AI Meetups. These Meetups gather hundreds of people. In this energetic crowd of AI-stuffed brains, you’ll find very kind and empathetic people who are happy to share advice with you. Ask them:

• “What did you start with?”
• “What resources for beginners can you advise?”
• “What are the major concepts?”
• “Up for a coffee?”

You will 1) get first-hand advice from experts and 2) make important connections.

And lastly, check the Data Scientist job requirements on any job board to learn what skills are expected from you.

Eliminate

When I build e-learning courses for large enterprises, my clients often want to include as much information as possible. There is never enough of learning, right? Truth is, large quantities of information overwhelm fresh learners, which can cause them to withdrawal. The golden rule of an educator, (and if you are reading this, you are a self-educator) is to eliminate as much as possible. Remember what Tim Ferris said about being effective? Right, ELIMINATE.

For example, when I was searching university curriculums, I found a module called “The Ghost in the Machine?”, about cognition and neurophysiology. I have a personal interest in cognition, so my first instinct was to include this module in my curriculum. However, a little rational thinking led me to the thought, “Cognition will not bring you to your goal, which is to learn AI for 12 months and get a job”.

This doesn’t mean that you shouldn’t learn cognition. Make sure cognition is in line with your major goal before pursuing.

Keep it Flexible

The more you expand and grow in your learning, the more you’ll discover new fields, tell important from less important, form your interest, and identify your strengths and weaknesses. Things you want to learn and explore will change over time, just like the seasons.

When I think about the future, I’m not sure when I should start reading a book on deep learning. Before I learn probability? Simultaneously with it, or after? Neither do I know when to learn Tensorflow because I don’t even know what it is. I can only define my first steps, which are primarily:

• Learn Python
• Revise algebra
• Make a beginner Data Science course (one of those advised by gurus)
• Read AI, a Modern Approach

One of these, maybe a Data Science course, will uncover my next step. For example, it may mention Tensorflow and explain what it is about, and I may want to include it in the next month’s curriculum.

For this very reason, I didn’t create a fixed curriculum with step-by-step instructions. Those curriculums are made in universities and bootcamps, by experts who have AI at their fingertips. We don’t, so we need to keep our curriculums flexible and open, allowing room for adjustments.

The Solution I Found for Myself:

• I put together a list of major learning points in AI, borrowed from university curriculums and AI gurus.
• I select those that I should start with.
• I plan learning tasks for the next month.
• When I understand what I need to learn next, I update my Trello board with relevant tasks.

Recap

1. Define your starting point in detail.
2. Set your major goal and learning sub-goals (in line with the major goal).
3. Create an approximate curriculum:

• Define major fields and concepts: borrow from universities, bootcamps, and gurus; go to Meetups; check out job requirements on job boards.
• Eliminate nice-to-know information and keep only must-to-know.
• Plan for one month and keep it flexible. Use Trello or alternative for help.