How to Know What to Learn in AI

Things are moving so quickly in AI it can tough to know what topics to learn, what’s important, and what to focus on. Part of this newsletter is sharing that with you along with the resources for how to learn it, but I figured I would share with you how I go about finding important topics and identifying the learning areas to focus on.

While I can filter much of this for you, there is a huge benefit to understanding how to and doing this on your own to direct your learning not only toward what’s important but combining that with what’s interesting to you.

There are opportunities in AI across the entire stack and it’s very likely you can apply your previous experience and knowledge to one of the most important problems within AI.

Below is how I keep up with the most important skills in AI and find the best resources to learn them.

1. Where I find the important topics

Job listings are a gold mine for understanding what’s important to know and the skills that are in-demand. They’re also useful to get a more practical understanding of what you’d actually be doing when working with those skills.

Most job listings are organized in three parts:

Use the first to understand what the company is and what their goals are. You’ll always be better off working in the critical path of what that company is trying to solve. You’ll also get a better understanding of what’s important to that company, why they’re pursuing their problems, and their approach to solving them.

This information is great for keeping up with the industry as a whole and also ensuring you aren’t putting yourself in a terrible working situation. All engineers I know want to solve interesting problems. All engineers I know aren’t able to do so in toxic work situations.

Use the second to understand the day-to-day. A company can list a role as ‘Software Engineer, Infrastructure’ but that can mean many things. In AI, that can mean inference infrastructure, training infra, and more. Use this to understand what you’d actually be doing. When using this across multiple companies and roles, you’ll get a better idea of what the industry is doing.

This section is also important to ensure the role you’re applying for is indeed the role you’re looking for. I see horror stories on Reddit about engineers getting hired to do something that isn’t actual engineering. Again, all engineers want to solve cool problems but all engineers can only do so in an environment that enables them.

The third section is the most important and why I suggest you always track interesting opportunities even if you aren’t wanting to make a move. The third section tells you what exact experience companies are looking for to fill certain roles.

This is the section that tells you what is important to learn in AI right now. Too many people get caught up in the fads shown on social media and understanding them and ignore the more durable skillset that actually gets them paid.

My biggest suggestion here: Whip up some sort of monitor that tracks companies you find interesting or that are at the forefront of AI and tells you what the most in-demand skills are for the role profile you are most interested in.

2. How I find the resources to learn them

The best way to find high-quality resources is by following the people capable of identifying them. This is a core pillar of AI for Software Engineers and a service I try to provide for my followers across all socials. For AI topics I’m less familiar with (especially on the research and business sides of AI), I follow others with more experience and check out the resources they share.

The first way I go about this is via my RSS feed. I keep the sources slim and remove any that don’t provide merit over a long period of time. I also have fine-tuned an agent to go through my rss feed for me and filter according to what’s important and what’s noise. I get an email each day sharing the important items with me.

My suggestion is here is to spend time curating your feeds. It pays dividends in the long run. I use Readwise Reader to do so and highly recommend it. You can get a two-month free trial if you use my code to sign up (affiliate link—but I’ve been recommending Readwise Reader since long before I’ve had it).

The second is via social media. My preferred platforms are X and Substack. X is overwhelmingly the primary point of tech discourse online and is valuable because of this. Even if you dislike Elon, I still recommend being on there. Recent changes have made it a much less hostile environment. If you don’t how to get started on X, reach out to me there.

Substack is getting better each day and contains incredible in-depth content due to the focus on articles. I haven’t found Substack Notes to be super helpful for learning, but it’s possible I’m using it wrong. I find it still pushes pro-Substack content above user interests, but I guess that’ll be fixed over time.

The third and last way I find these resources is by chatting with people. Some of the best resources I’ve found were by DM’ing area experts on a certain topic to ask what they suggest. Even if a person isn’t sharing resources actively online, experts in those areas know where to find them.

Don’t worry about bothering people via DMs. The worst case is it gets lost in their inbox and they don’t get back to you. Just don’t spam them and this isn’t anything to worry about.

3. A note on learning

It’s incredibly easy to get overwhelmed when onboarding to this space. My advice is to take it easy and focus on the items that interest you first. Consider your previous experience and how to leverage that to get started.

There are many different ways to get involved. You can get involved at the network layer in data centers, at the hardware layer designing chips, at the research layer improving models, at the systems layer building out machine learning infrastructure, at the engineering layer putting it all together, and at the application layer bringing AI to people.

All of these can also be combined in some way with specific roles. Many researchers also do some engineering. Many engineers get involved in the research they’re bringing to production.

Find something you find genuinely interesting and go deeper into it. Don’t feel like you have to know it all. This is a trap I sometimes fall into, and I end up being less productive than when I really dig into the things I genuinely like.

If you have any questions, reach out.

Thanks for reading!

Always be (machine) learning,

Logan