Comment by jasode | Hacker News Reader

Comment by jasode | Hacker News Reader

You don't state your programming background but you mentioned C# in previous comments.

I would recommend machine learning as a very useful skill to acquire.

I have to be more specific because "machine learning", "artificial intelligence", and "data science" are really large topics that can encompass difficult math and PhD research.

For application programmers, I think a very realistic subset is using an existing "machine learning toolkit or API" (e.g. Keras in Python) to analyze data and solve problems. You use the machine learning algorithms but don't necessarily write them from scratch. This level of proficiency only requires high-school math. The analogy would be knowing SQL even though one doesn't write b-tree algorithms from scratch or using the "=PMT()" function in MS Excel without deriving the annuity equation from first principles.

I think machine learning concepts (classification, clustering, etc) will become an expected baseline of programmer knowledge just like SQL was 25 years ago. Its usage will become more pervasive and will be utilized by more people that don't have "data scientist" in their job title.

I don't mean to imply it's impossible, or throw out some appeal to authority, but dabbling in ML is not going to open doors for you. I worked in and around the field for more than a decade. Companies who are serious about this stuff can sniff a beginner out and they tend to want credentials. It's a deep subject, so that's not terribly uncalled for, and knowing how to throw a classifier together does not in any way mean you're competent and able to solve hard problems.

Not trying to be a downer, just my $0.02. For fun? Sure, go for. For career development? You had better be serious and ready to devote the next 5-10 years. I do not at all agree that ML is going to become a baseline skill.

I don't mean to imply it's impossible, or throw out some appeal to authority, but dabbling in ML is not going to open doors for you. I worked in and around the field for more than a decade. Companies who are serious about this stuff can sniff a beginner out and they tend to want credentials.

What I think you're missing is that ML is becomeing accessible to organizations WAY outside the sphere you're referring to. In the not-too-distant future, "Bob's Small Engine Repair and Screen Doors" will want to take advantage of ML / Data Science / AI. And it may well be the case that all they need is somebody that can implement Linear Regression or K-Means using SK-Learn, or using a cloud ML API.

And larger organizations will always have room for people with varying levels of skill, especially as demand for the most highly skilled people outstrips the available supply.

and knowing how to throw a classifier together does not in any way mean you're competent and able to solve hard problems.

Not everybody is solving a "hard problem". By way of analogy... if your car has a broken connecting rod, you need to take it to a mechanic who knows how to pull an engine, tear it down, rebuild it from scratch and put everything back together. If you have a bad O2 sensor, that work can be done by somebody much less knowledgeable. If you need the oil changed, that can be done by somebody even less knowledgeable still. Or to use a different analogy, you don't need a neurosurgeon to cut out an ingrown toenail. Use the right person for the job at hand.

> Companies who are serious about this stuff

What about companies who are not serious about this stuff? There are tons of companies who are not actually trying to solve a hard ML problem. They just want their own recommendation or prediction engines. That doesn't require any deep understanding or 5-10 years in this field.

A ML puritan might say that is not real or serious stuff and this is all fad. That might be true. But I think that is one of the ways someone can start off their 5-10 years journey - working on simpler problems and getting paid.

> ...just like SQL was 25 years ago.

Oh boy have I got news for you :)

What’s the news?

That SQL is still relevant after 25 years.

Original comment never mentioned that SQL wasn't a baseline for programmers anymore.

There appears to be a “10” situation here, some see through the hype, others don’t.

I would say the baseline is a little higher now, maybe that is what he meant? But yeah, SQL is required knowledge for any programmer.

Reply on news.ycombinator.com