The short answer: you do not need advanced math to start, but a few core topics make things much easier as you progress.
What do you need to start?
To train a model with ready-made libraries, high-school math and logical thinking are enough. The libraries do the calculation for you.
Math comes in when you need to understand why a model is not behaving as you expected. At that point what helps is not memorized formulas but knowing what the concept means.
Topics to prioritize
- Statistics: Mean, distribution, correlation and probability; the basis of interpreting data
- Linear algebra: The idea of vectors and matrices; to understand how data enters a model
- Derivatives: The basic idea is enough to grasp how a model reduces its error step by step
How should you study?
Rather than studying math front to back as a separate course, it is more efficient to learn the concept behind a topic when you run into it. Math learned alongside code sticks better, and you see right away what it is for.