How I taught myself to get cutting edge in AI

How to teach yourself AI- for technical folk

A lot of people ask me how they can learn Machine Learning. If there is an optimal set of courses, certifications, books, degrees, projects, etc, that would get them fully ready to tackle cutting-edge AI.

I don’t like the question because there are two important factors to consider-

1. Different goals require different actions. Trying to follow one path will lead to inefficient results (engineering vs research requires different skills, managers/investors often benefit by not going as deep into the details as they’d think…). I can’t just give you one set of recommendations.

2. The question treats AI knowledge as a fixed state with clear objectives and metrics. Imo, success in AI (and tech in general) is less about knowing the right things at any given moment, and more having the ability to filter through constantly shifting (and very often very polarized) knowledge sources to pick what works for you in a specific context. This is a mentality/skill that has to be cultivated, and reliance on courses and other prepacked knowledge sources tends to work against this.

In this article, I will talk about the approach that I used to teach myself difficult Machine Learning ideas and concepts on my own, without a Master’s Degree, paying for expensive boot camps/courses. I did it using the free resources available on the internet.

In this article, we will cover the following points-

1. Why the standard advice for learning Machine Learning (do some foundational math + Computer Science (whether by yourself or through a college), learn about basic neural networks, do basic projects/Kaggle Competitions) is a bad base to build your knowledge around.

2. What you should do instead (reading cutting edge work solving real problems, even if you understand very little).

3. Why that works much better.

and more.

Since this is a paywalled article, I want to make a things clear before you get into it:

Who is this for-

1. Technical AI Folk- either AI Engineers or Researchers (or people that want to be technical). While the ideas will be useful to anyone, we’ll do a separate piece for non-technical people.

2. People who are willing to commit sometime regularly. This approach will require at least 4-5 hours weekly (don’t burn yourself trying to add too many hours), done over 8-12 months to really click (this assumes you have no baseline knowledge- prior knowledge will speed your learning up a lot). You will start to see improvements in the first 2 months, but the real benefits come later. In my experience, there’s no real way to build lasting knowledge without deep exposure, which comes with time. You can’t speed-run excellence.

3. People who can handle unstructured learning. If you’re someone who absolutely requires structure with graded assignments, learning paths, and clear instructions- you probably will have a hard time with this approach.

Assuming you meet the criteria, this approach will likely be as helpful to you as it will be for me. I know it’s helped a lot of other people (there was a time I used to work with people as a tutor/mentor and it was very helpful to them)-

For access to this article and all future articles, get a premium subscription below.

Many companies have a learning budget that you can expense this newsletter to. You can use the following for an email template to request reimbursement for your subscription. You can also pay what you want here.

I provide various consulting and advisory services. If you‘d like to explore how we can work together, reach out to me through any of my socials over here or reply to this email.

Executive Highlights (TL;DR of the article)