As a newbie, you are most certainly correct that finding a way to stay focused and motivated is essential – however, it’s unlikely that privacy will be necessary.
In the process of learning, especially early on, you will most likely not be tackling novel problems with novel solutions – so google (or paperspace) being able to see how you tweaked the pipeline for something like a chapter (of the book) or tutorial in the docs is not a tantamount concern as you are trying to lay the foundation to have this capacity (for novel exploration) at some point later on.
As Jeremy (and others) have stated, the goal is to hit the ground running as fast as possible. In hindsight, this is the correct way to go about it. I spent weeks/months early on trying to get a GCP instance running when I first came to Fast.AI – while it was a useful exercise, it most certainly lead to headache and was detrimental to the momentum of remaining focused and motivated.
However, everyone is different and what works for most may not for all. What you want is to be able to engage with these tools, in an applied way, with as little time elapsed between the idea and implementation, even if it’s something as simple as tweaking a learning rate or batch size. Take a look here at the notebook servers section of the book/courses website: https://course.fast.ai → to find the best means of getting yourself setup in this fashion. If you decide you like it, you can invest more time (and money) into a more durable, persistent and secure node for your work.
You do not need to know too much math – if you are willing to learn it along the way, as concepts present themselves, you are well positioned to tackle the required math as it arises. As stated in the course, Khan academy is probably sufficient as a primer.
*I’ve only addressed the parts that I feel capable of commenting on 