DEV Community

Hilal Eylul
Hilal Eylul

Posted on

How to Focus on the Right Content when Learning AI/ML

I began actively using Linkedin after I got my first job as a Machine Learning Engineer. Eager to learn about the field as much as possible, I began following more senior professionals who shared my enthusiasm. These people were working at a variety of companies, from FAANG companies to universities to growing startups to established companies that have been around for many decades.

Seeing their posts appear on my feed, I began compulsively liking them. And by that, I specifically mean clicking on the “thumbs up” button. That way, other people knew that I had liked a post, and they could go to my profile and see my history of liked posts on the platform.

It was like a curation of the type of content that I, Hilal Eylül, found to be interesting and worthwhile.

Even more importantly, I could go back to these posts later on. If any post seemed to contain information that I didn’t have the bandwidth to examine while scrolling, I could refer to it when I was in a more studious mood.

Not long after, I began to realize something.

Some of these posts were not actually very helpful.

This even includes the external sources that my fellow professionals on Linkedin mentioned and promoted in their posts. They were not the kind of sources or information that would in any way be vital on one’s path to becoming a more advanced professional in the field of AI and ML.

If anything, a lot of these posts were self-promotional. Either for the individual or for the company the person worked at.

That’s exactly why I decided to address this topic. These are the things you can do to filter out useless content on your journey to becoming a better AI/ML Engineer.

Learn how to spot covert ads

Let’s face it. A lot of things you see on the internet are secretly just ads.

This includes tweets, photos and videos posted on various social media platforms, newsletters, and even blog posts.

But in the field of ML/AI, it also extends to entire books and tutorials.

Lewis Tunstall’s book “Natural Language Processing with Transformers” very effectively promotes Hugging Face, the company where he works. The famous course “Practical Deep Learning for Coders” promotes fast.ai, which admittedly is a non-profit organization. Google has tons of “hands-on” courses for ML engineers where they essentially promote GCP.

This is not to say that none of these promotional materials are valuable. In addition to the course developed at fast.ai, Lewis Tunstall’s book is very informative and covers topics that are highly relevant in the industry.

But I would argue that the next time you come across educational material, ask yourself these two questions. First, is this a premier resource that can actually help me learn about topics that are frequently applied and sought after in the industry? Second, does the material lock you into being dependent on the advertised product such that you practically have to start from scratch if you want to move to an alternative?

The main idea is to identify from early on how you, and the people who are creating and promoting the content, will benefit.

Then there is content that is not necessarily sponsored or a native ad, but still promotional.

For example, it is not uncommon for someone to write a blog post or post a video on social media with a call to action. This is often innocuous.

It is not necessarily an attempt to instill false confidence or a feeling of productivity through a promotional post but just an attempt to build a stronger relationship with others. The call to action might simply be an invitation to join a newsletter. That way, it is easier to foster a relationship with the people who are interested in your content.

I also have a newsletter. There, I will be posting even more interesting stuff that I don’t post on this platform. The signup page is at mleresource.com.

Identify the right tech stack

Waylon Walker, a senior software engineer, wrote about the importance of focusing on the right topics and skills.

He references a social media post where another Linkedin user gives his own advice. That is, focus on Python and C++, avoid “dead” software such as Tensorflow, don’t spend much time with R because it won’t get you far, and also avoid languages that are “too academic” like Haskell and Julia.

I rarely use this word, but I’ll say it now. Baloney.

These various tools exist because there are people and projects and companies in the world that have a use for them. Tensorflow is still very much alive. R has a library that can help create beautiful data visualizations. If you’re really good at Haskell, you can potentially get a job at a major tech company and use it there.

Don’t chase after trends and buzzwords. Do identify a tech stack.

What are your interests? What types of projects do you want to work on? What projects have you seen and thought “hey, I wish I did that”? Hint: it doesn’t have to be something that everybody is talking about. It just has to be something that’s practical and something that you can see yourself doing.

Don’t think of your journey as something that involves learning everything that is relevant to AI or ML. That was the takeaway from Waylon’s blog post. But don’t feel pressured to learn something because it’s popular.

Identify what you want to do and what type of projects you want to work on. Figure out which skills you need to learn to fulfill that goal. Not the goal of optimizing your resume to match as many job descriptions as possible.

Next, focus on learning those skills.

Check before diving in

It’s quite simple.

Before you embark on a course or begin reading a book, don’t just make the decision to invest your time and money in that material based on reviews. Or even based on glowing reviews from your friends and colleagues.

Don’t just test the waters.

It is definitely worth looking over the material before you dive in. For example, consider skimming through the table of contents of that textbook or O’Reilley book that you’ve been considering for a long time.

Maybe it’s not as useful as you think. The material might have become popular because it is relevant to certain types of companies or projects. A book on, say, statistical methods for applications in finance will only be so helpful if you want to become an expert on computer vision or MLops.

Ask yourself: what is the end goal of studying this material? For example, it could be learning linear algebra concepts so you can become more comfortable with building deep learning models.

This can also apply for simple posts on apps like LinkedIn. What is the goal of this post, and how can it actually help me grow in my career?

Theory and concepts are more important than you think

In the field of software engineering, we love applying things. So much so that we usually disregard theory and concepts. And this is relevant to AI/ML because software engineering is a superset of ML engineering.

The internet is, thankfully, abundant with tutorials that can help us in various stages of the machine learning lifecycle. These tutorials can definitely be lifesavers when we need something figured out. Fellow software engineers, keep the tutorials coming!

The only issue is that many software engineers, and by extension ML engineers, often have a disregard for theory and concepts. So many software engineers scoff at “leetcode questions” and brag that they never had to learn how to solve them. These questions basically test for understanding of Data Structures and Algorithms.

Likewise, ML engineers focus primarily on the code. There seems to be more of an emphasis on getting the code to work than trying to learn the concepts. As a result, this compromises the quality of the algorithms and even the products being built.

It’s surprising how many self-identified software and ML engineers don’t even know what they’re doing.

What’s even more crucial is that once they get to a certain point, many people in this field begin to have a disregard for mathematics. Namely, topics such as statistics, linear algebra, and even calculus. If these topics were ever learned, they are quickly forgotten.

The advancements in technology, software engineering, and AI would not have been possible without math. Knowledge of this topic is still necessary if we want to contribute to this field.

Know when to get out of “tutorial purgatory”

Tutorial purgatory might as well be considered the bane of any junior software engineer’s career.

The sooner you get out, the better.

In essence, tutorials are great because they help get a beginner’s feet wet. Going through too many long form tutorials can result in fear. Eventually, it can get very difficult to get out of your comfort zone.

I would argue that tutorials are still helpful to me, a mid-level ML engineer. It’s just that what I find to be helpful when I am working on projects are short form tutorials and documentation. These only support the projects that I am working on.

Bonus tip: Don’t be afraid to waste time

You live and learn.

Ever heard of the 10,000 hour rule?

The more you focus on something, the better you get at it.

Now you might argue that the 10,000 hour rule is not based on science. Or you might point out that it is only helpful if it involves deep work and deliberate practice.

All of that is true. Deliberate practice certainly trumps more passive involvement.

The reality is that when you begin a journey, you will make mistakes. You will spend time focusing on things that you later wish were never priorities. There’s even a chance your areas of focus might shift.

And the latter might even happen multiple times.

By focusing on “the wrong things” you are still expanding your knowledge. And that knowledge might be helpful even after you become established in your career.

The End

That’s it for today! I really hope this post was helpful to everyone. Especially for those who were feeling a bit overwhelmed by this whole process and all the content out there.

Top comments (0)