I can offer my perspective as an auditory researcher that is currently Up To No Good in machine learning most days of the week. My first languages were matlab and c#. These days, i write almost exclusively in python.
Python is the lingua franca of machine learning, and given the dominance of the library ecosystem, that is unlikely to change. If you have prior experience with matlab or another vectorized programming language, that probably helps, as a large portion of python's numerical computing ecosystem is strongly influenced by matlab's choices over the years (or, strictly speaking, both owe a great debt to fortran/linpack/lapack/blas).
My primary advice to anyone beginning to program is to know what problems you want to solve first; it's just a tool, not an end unto itself (unless you're simon peyton jones, and i'm sure not). I think you're going to have more immediate perspective on this than many new programmers would and for similar reasons that I had as an undergraduate. My motivation was "I have this large heap of experimental data in this folder and i need plots to put into this report"; data analysis pipelines are the bread and butter of scientific computing.
If i were you, i'd grab a copy of
https://thonny.org/, which is a lightweight tool for python beginners that is at the same time a full tool, and start hacking around on your own data. Python's great for exploratory coding, which is good to become accustomed to the language, though you'll want to move to more formal software methods at some point.
For a good free text for python the language, I recommend
https://greenteapress.com/wp/think-python-2e/ -- professor downey writes well. His books on DSP, probability, stats, bayesian analysis, etc. are also all very good references. Read a little, code a little, read more, code more.
Once you're comfortable with the language syntax, you'll want to know about these libraries for machine learning: scikit-learn, numpy, matplotlib/seaborn, and pytorch. Warning: each of these is effectively a mini language which happens to be written in python, but these are feature-FULL libraries, and they move pretty fast. I don't know of any great books, as a result. A good overview of numpy (and DNN-based computer vision) is this course material at stanford
https://cs231n.github.io/python-numpy-tutorial/, though.
Finally: there's a great amount of interactive help that you can get for this. On IRC, the freenode #python channel is very active. There's a discord channel as well that is also full of folks who are eager to help.
sorry for the firehose of information!
-graham