I've had no programming knowledge prior to GML. I started by following a simple game tutorial using D&D actions, then expanded on that and generally fooled around and played with it. It was surprisingly effective as a teaching tool because all i did was play but at the same time things were falling into place and i gained a better understanding of how a complex program comes together.
Pretty quickly i became annoyed at the D&D actions and wished there was a better way to combine multiple ones, or edit them more easily or ... oh hello, an Execute code action. I fell in love with that one, and with a very helpful manual began using nothing but GML.
Years later, i would say i have a good grasp on a lot of GML though there are some things/functions i have never touched. When i bother to design, i mostly use paper and pencil then with the aid of the manual, convert that into code or at least pseudo code. Sometimes GML and what i have on paper disagree, other times it works like a charm. I also tend to write pages of comments for each line of code, because i'm the poor soul that will have to revisit that code and most of the time i have no clue what the hell any of it does. Or that i'm the person who wrote it...
Not an expert by any means, but good enough to try and tackle problems using it. And it makes me happy. AND it makes other languages more accessible! It may take a long time, but it's not about time or retention. In my opinion it's about experience. Everything done with GML or D&D or other languages, stays with you and leads to better understanding, even if most of the time this isn't immediately obvious. It helps a lot if you love it.
The manual is allways open. Most of the time it's indispensable and some times it outright gives ideas on how to do something. It isn't analy specific or detailed enough, but the function descriptions are excellent for beginners. The code examples, not so much. And there's been a few times i despaired at not havining enough information but the gripes are few and far between, and not enough to knock what has been arguably GameMaker's greatest asset.