I wanted to get started with ML. So what are the basics that I need to cover? (I have no idea where to get started.)

First you need to figure out how deep you plan to go into ML. Do you just want enough to hack together a bunch of cool (and useful) projects, or really dive deep into the maths and inner workings of ML algorithms?

@RememberMe Deep dive into ML.

@donyvarkey start with maths. You need to have good understanding of integrals, differentiation, probability, matrix etc..
You're also gonna use a lot of statistics for data cleaning.
Make sure that you're learning from a good source like udacity, udemy etc..
Don't rely on YouTube tutorials if you're learning it for your career

@retard thank you very much.

@retard @donyvarkey um, YouTube actually has some of the best ML resources out there, like lecture videos for Andrew Ng's Stanford class. Udemy is quite bad for this my experience. Since you want a deep dive you want to go to places that are more academic-y - blogs, course lecture videos, lecture notes (Stanford CS229), Coursera, etc.

A decent knowledge of calculus, linear algebra, and stats is pretty much required, you should consult appropriate books etc. for that. Or just dive into some ML course, see what maths they actually need, then stop the course and go learn the maths first (or ask here for maths resources).

@RememberMe sure.

@donyvarkey how comfortable are you with calculus, linalg, and stats in general? I could recommend some sources for those if you want.

Also if you're learning on your own staring at formulae gets pretty tedious. I'd recommend you also take some decent ML library (like Python's scikit-learn) for a spin, try doing some basic projects with it. Great way to build that interest.

@RememberMe somewhat ok with maths.

Start with a problem you'd like to solve. Focusing on the tech is where all commercial attempts to deploy ML fail.

So, pick something like trying to spot porn in a list of images, say.

Benefit of thinking like this is that you find out that a lot of "AI/ML" is piss easy---eg define skin tone and count ratio of skin tone pixels.

Key to success as a data scientist is framing the problem. If you can do that well, you'll be way ahead of the curve.

But, you can implement a deep learning solution for it too. :)

Like @platypus said Problem first, and stats second. Seriously ML is a combination of stats with computer engineering on top