Training for a long-term career in AI/ML

Recommendations for students (and professionals) on what and how to study if you're interested in a career in AI and machine learning

If you’re a high-school student, undergraduate, or a post-graduate interested in AI/ML, this article has advice on what you should study at what stage of your education/career to ensure a long career in this area. This is primarily based on a talk recently given by Dr Ashwin Rao, Vice President of AI at Target, and Adjunct Professor at Stanford (video). This article is my summary of the main points.

How do you learn AI/ML (Artificial Intelligence and Machine Learning)? There are blog posts that claim to teach it to you in half-an-hour. Coursera and other MOOCs have short and simple courses, as well as longer and more complex specializations. Universities offer entire undergraduate and postgraduate degrees in AI/ML, and it is not easy for a layperson to figure out which one is the right choice for them.

Ashwin has been mentoring students at the high school, undergraduate, and postgraduate levels as well as professionals for a long time. He now has a good understanding of what works well and what doesn’t.

I’ve captured the detailed recommendations later, but his most important advice is: before diving into AI/ML, focus on basic mathematics (statistics, linear algebra) and get comfortable with programming (Python). Otherwise, you end up thinking you understand machine learning, but you don’t and hence can’t really do it well.

What to focus on at different stages of your education. Having strong foundations in maths and CS/programming before diving into AI/ML algorithms is extremely important. Doing the latter before the former is a mistake.

Also, pay particular attention to the recommendations of what not to do. And if you’re not a high-school student, or you missed out on some of these subjects in high school, you can still use this list to understand what are the topics you need to catch up on now.

Focus in High School

• Most important at this stage is to have strong fundamentals in basic maths, probability, statistics, and basic calculus.

• You don’t have to be a star student in math. You don’t need high scores in your math exams. What you need though, is curiosity to understand math conceptually. For that, you must enjoy maths (the understanding part, not the exams). You have to be able to think through problems from first principles. Too much of high-school math is about applying formulae. But you really need is a genuine interest in finding out how the formulae are derived.

• You don’t have to do any advanced math (like advanced calculus) or even any AI/ML related maths like neural networks. Just do the basics but understand them inside-out. Get an intuitive feel for maths via websites like AOPS and YouTube channels like 3Blue1Brown.

• Basic programming at this stage would set a good foundation for all the programming and computer science (CS) background needed later in your AI/ML career. A simple language like Python would do.

• At this stage, you don’t have to learn or understand any machine learning at all. You can do robotics or other toy projects related to machine learning. But the focus should primarily be on having fun. You are not really expected to understand the AI/ML algorithms.

Undergraduate

• At this stage, your focus still should not be on AI/ML. Instead, strengthen your basics in CS and Mathematics.

• This screenshot gives you an idea of the kind of math and CS knowledge that you should focus on

  

  The maths and CS courses that really form the foundations on which the edifice of AI/ML is built. Focus on building these in your undergraduate days rather than any actual AI/ML algorithms

  This looks like a large list of courses that don’t seem directly connected to each other, but any substantial AI/ML project will sooner or later touch upon these topics. and most of these are covered by a decent CS undergraduate program these days.

Postgraduate

Now is the time for you to finally start getting into the details of AI/ML algorithms. This means learning not only the modern Deep Learning and Reinforcement Learning algorithms and their applications to computer vision, language, and other domains, but also getting exposed to classical/statistical AI algorithms. The best AI systems of the future are likely to be the ones that combine classical AI with learning AI.

Frequently Asked Questions

Q: I didn’t do a lot of maths in high school, or CS in undergraduate, can I still do AI/ML?

Certainly. Anyone who is not a high-school student now probably missed out on some of these courses/topics. But nothing prevents you from catching up now. Just make a list of what you’re missing and join online classes for them, preferably in order.

Q: Do all kids need to do AI/ML? Are you saying that the future is dim for students who don’t do this?

Absolutely not. The first sentence of this post starts with “If you are … interested in AI/ML”. It is not necessary for everyone to be interested. In fact, increasing adoption of AI/ML is likely to increase the demand for expertise in other fields, like say, psychology or humanities.

Q: I’m already 30+ and halfway through my career, and I have no background in AI/ML. Can I still switch?

Yes, certainly. But keep in mind that it is going to be a slow and long process. You can’t learn AI/ML from blog posts, or from downloading someone else’s code, someone else’s data and tweaking the configurations. But if you plan well, it can be done.

I will focus on this specific problem (mid-career switch to AI/ML/Data Science) in a future article. I will cover the journey of someone with no AI/ML/Data Science background who successfully completed the transition a few years back and is now a Principal Architect in AI/ML at a unicorn. Stay tuned.

Subscribe and Share

If you know someone who would benefit from this advice, please share this post with them:

If you haven’t already done so, please sign up to get the future posts by me via email.

Thanks to Ashwin Rao and Meeta Kabra for reading drafts of this article. Thanks to Meeta Kabra for editing.