I think more people should be studying statistics, machine learning, and data science, especially Bayesian methods and causal inference. Not only do these skills offer a chance to contribute to AI safety, they're also critical for evaluating scientific papers (important for any field given the replication crisis), doing predictive modeling, and generally thinking in a data-driven and evidence-based way. Math is apparently 80k's #1 recommendation, but when I was a student, I went to an event where math majors talked about their experiences in industry. Most of them said they didn't use the math they learned much and they wish they had studied more statistics. So I would suggest applied math with a statistics emphasis.

I'm not sure I can answer from an EA perspective, but I'll try to personally reflect on the most useful university courses I've taken:

-I took two postgraduate research methods courses (Quantitative Research Methods and Evaluating Public Policy). I am now much better at evaluating the claims made in social sciences papers, which I'm really happy about.

-My Christian theology courses helped develop my thinking about my faith, which I've found personally valuable.

-Sociology / Psychology courses have influenced some of my core beliefs (nature vs nurture; rewards vs punishment; etc), so in a sense I use this knowledge on a daily basis.

-Drama classes can be great for developing your communication skills, as well as your ability to put yourself in someone else's shoes.

-I trained as a teacher, and my classroom management class was surprisingly transferable to an office setting. It really helped me think about how I can influence other people's behaviour by choosing a meeting's environment and agenda and through my own verbal and nonverbal communication.

Taking one programming class to test your aptitude (or spending the equivalent time learning online) might also be worthwhile. If your one programming class goes well, you might decide that programming would be a good back-up option in case you don't end up working at MIRI/GPI/CEA/wherever else you really want to work.

I think people don't stress enough how important it is to be able to pay your student loan and keep a roof over your head after your graduation. Learning a programming language and, if that goes well, interning or working part time or taking a summer job working in that language, is an excellent way to keep yourself out of poverty.

I think it also depends on what you already have and how much you think you can improve.

From Social Psychology I know that personality rarely changes. Skills do.

So if you are an inrovert (which is by the way, the most stable characteristic), not agreeable, not consciouscious (meaning not hard working), not open to experience (uncurious) or neurotic, sadly there is a little to be done about it.

You can improve little "islands" in your personality, like setting clear goals as a leader (I guess), but a lot of soft skills are ranging between obvious to easier said than done.

Is the soft skill too easy for you? Is it too hard? I guess the best fit will be a challenging skill - a bit too hard for you to step out of your comfort zone, but not to be overwhelmed.