I am going to rehash something I discovered.

AI/ML are the buzz words of tech today. The hype is everywhere. One can find everything from bleeding-edge-tech-of-tomorrow to get-rich-next-year kind of posts from Devs, Journalists and companies. But most miss the nuance..

Data science (once called statistics) is different from ML/AI. They are a very vast field of work and it's easy to get lost as a beginner. Picking up a free introductory course online like Kaggle Academy or Coursera will provide a guided entry into the subject matter.

While it's easy to get started with the availability of libraries like Keras, PyTorch, Tensorflow, ScikitLearn..etc., and open datasets, any meaningful work would involve learning the mathematics surrounding the problem to solve. This is what will make or break ones ML foray

Here is a video by Dan Becker that explains things in detail

If you are purely from an engineering background and have solved problems by implementing components and gotten comfortable with creating data pipelines for users (webapps, backend, mobile..etc), getting into ML would be a complete shift in thinking.

After 3 coursera courses of learning the basics and processes in ML, I discovered it's a shift I can't handle. Without understanding the underlying mathematics, I can't engineer anything new. I also couldn't stand spending hours just collecting data cleaning it, feeding it ..

Into the learning library, tweaking it parameters and observing results. Entire thought process of engineers to predetermine the outcome of a process doesn't exist in ML. If you really love TDD, you will probably feel lost with ML.

Aside: TDD - Test Driven Development. A practice where you first write unit tests and then write code to pass those unit tests creating a well defined and tested codebase.

Now, about the lure of sky high incomes and get rich offers - Don't fall for it. Getting started with ML is quite easy but achieving success is hard. Eg., Microsoft's AI tweetbot turned into a racist, holocaust denier after 1 day. https://nymag.com/intelligencer/2016/03/microsofts-teen-bot-is-denying-the-holocaust.html

Success in ML is largely determined by the quality of data used for training. The other side of this coin is ML engineers become less important. A mediocre team with a good dataset will produce better ML models than a brilliant team with terrible data.

While in engineering, the success of systems built will for the major part depend on engineers involved. I prefer engineering for this reason.

So, if you are looking for a increase in income, it could be achieved by becoming better at what you are already doing.