Ask HN: Do you regret being a generalist?

I don't think an area such as AI can be called a specialist area. There is simply too much breadth once you start getting into the nitty gritty of building an AI product. If your goal is to be a well versed product builder in AI, then you need to master software engineering, building deep learning models, training on large clusters, deploying on the edge - man it's a lot.

A data scientist or a deep learning engineer - yes that's somewhat of a specialist role. It again depends on what you want to accomplish. If your goal is to research deeply in one area such as NLP then yes that's a specialist role.

OR,

If your focus is purely on infra, like building and enhancing Pytorch or Tensorflow, yes that's specialist knowledge in fine tuning a loss function etc or building a new loss function altogether.

As far as I am concerned, yes I do sometimes wish that I had stuck to one breadth area such as AI or mobile app development or AR/VR instead of somewhat dabbling across all these areas. No regrets though.

Fair enough :)

It takes a long time to get truly skilled in something. What we call an expert is really what a reasonably intelligent person can be expected to learn about a single topic within a lifetime. The gains in one area compound on themself. Given that these gains compound, you can't learn 2 things half as well. So the T-shape represents the constraint that you only have a shot at becoming an expert at 1 thing. Spend 80% of your time on that and 20% on the rest.

The Pareto principle (80/20 rule) also implies that you only need to read a handful of books on a single subject to know more about it than most people. The knowledge gains become marginal unless you focus most of your time on it. So the things that you find interesting that aren't your main thing, just "learn the ropes" and move on back to your area of expertise. That's the most efficient traversal of the Sum of Human Knowledge: Familiarize yourself with the lay of the land, then focus most of your energy on some place that hasn't been mapped yet.

This is a complex topic!

Having exposure to a wide variety of environments makes you far more valuable. Making gears gave me a whole new perspective on the world of manufacturing. If there's something new to learn, why not go for it?

Why not both? You've got a long career in front of you. I started out a straight software dev, stumbled into being a "DevOps" expert before the term existed, ran a team for while, and am now sliding into "architecture."

Being a "generalist" means that you are going to pick up a ton of varied skills. Eventually, you'll find out that one of them is something that no one else has and your current team/project is in desperate need of. Congratulations, you are now the "expert!" If it is something you like (or, in my case, was simply angry didn't exist), double down and keep pursuing it. Now you are the "specialist." If you get bored, step sideways into something else.

There are places where you really do need to stop and specialize- for example, there are parts of AI are in fact hard to get into without lots of specialized experience, but much of that seems to come from Academia which at least has a clear path to gaining the experience and credentials. Even then, there are ways to side-step into it- my current team is a data science team, where I am the voice of software and infrastructure engineering reason. There is a fairly clear path picking up more and more data science or AI tasks if I wanted to, eventually doing that full time instead of what I'm doing now.

I've found a lot of success and happiness by focusing on problems that I'm convinced need solving rather than technologies or categories. Sometimes that means I go months without coding (that does make me a little sad), but it is very fulfilling to see my positive change for the better come to light. It also has driven me to pick up all sorts of skills and technologies that 23-year-old me never would have thought I would, let alone enjoy!