I always had the same feeling. I'm not bad at math in general (I'm not well versed in high level maths either), but as a developer, trying to jump in the ML field seems really impossible. One would think that he could teach himself ML algorithms, but you ALWAYS end up reading math notation instead of pseudo code.

To be honest the 3blue1brown videos seems really wonderful at explaining what is going on without going too deep, as the math in ML lecture seems to be trying to prove everything, and is always trying to teach using math notation all the time.

I guess this is happening because most of ML is mostly coming from research since it's all new, so it's being taught mostly by people who can grok the math, meaning mathematicians, it's not taught by people who are programmers. This really shows how much math should keep being math, and not leak into fields where practice matters more. Programming languages and pseudo code are not for nothing. Computers don't talk math.

So as years go by, ML will be taught more as a practice subject rather than a theory one, and things will get better. I think it's just a matter of how it's being taught, because reading code will always make more sense than reading high level math. Videos and oral explanations also help a lot.