I’m a bit puzzled by the negativity of the current answers. From what I know about ML applied to neuroscience (computational neuroscience), and at least some U.S. universities, this isn’t even close to being the case.
It is true that you might help reviewing papers as well as work on grants - both your own, and others to help your PI. At this stage of your career though, it won’t make up anywhere near as much time as postulated in other answers; helping your PI with their grants might not happen at all because it's more common for postdocs instead. Assuming you consider an academic career, it’s also useful training as especially grant writing will make up more of your time in later stages.
As to classes, if there even are any, they will only happen in your first - possibly second - year. Instead, you rotate through labs at the beginning to help find out what you’d like to do (you’re not necessarily locked into computational work upon entering, and might decide to do experimental work instead). If you work interdisciplinary, it’s important to be good not only on the computational side, but also understand the experimental work and related biology, so, again, in any case it’s time well-spent.
When you hit the research stage, if you’re lucky you can hop on a project your adviser has handy for you. If not, you’ll spend some time reading and trying to find your own niche, hopefully working closely with your boss.
When the research begins, you’ll cycle through talking to the experimental groups you work with to decide next steps, and setting up models and running them. Running your models will make up the vast majority of your time then. As needed, you research alternative approaches at the side. When it finally comes to writing your paper, you and your co-authors likely iterate through a fair number of rounds to get it done so everyone is happy.
If you’re the academic type, it’s a rather fascinating life.
