1. What problem is this trying to solve?
2. What are the strengths / weaknesses?
Usually you don't have to go very deep to get a feel for these questions. I figure that if I know these things, then in future if I encounter a problem where tech X is useful I can go learn the details. If I never do, I can at least prioritize within whatever free time/energy I have for additional side learning.
As an example, for quantum computing the problem it's solving is surprisingly hard to identify. Breaking (some) cryptographic algorithms is the classical example, but that's not a problem I plan to try and solve anytime soon. Other use cases tend to be vague - ML is often mentioned but actual players in the AI space don't seem to be investing into QC much. Quick eval result: pass for now until use cases become clearer.
What about ML? Many useful problems that can be solved with that - anything where you can't easily express an algorithm to solve a problem but one may be learnable by example. Strengths: capable of doing things no other technique can do. Weaknesses: requires lots of clean(ish) data, advanced ML may need specialists, training/inference can be expensive, finding ways to actually apply it to business problems is remarkably tricky due to its capacity for random failure and difficulty in coming up with fixes for that, which doesn't fit the needs of most automation projects ... but can be OK if you can predict fast enough to speed up a human who's already in the loop, or if mistakes are cheaply correctable. The exact details of how it works? I enjoy reading about it. There's a good article on the GPT-3 DNN architecture on the HN front page right now actually, but it's just some casual hobby learning. I don't feel stressed if I don't grok all the details right now because it's not necessary to do so.
In other words, focus on knowing the outlines of these topics and learning when/how to apply them, then forget the rest. That doesn't take as much time as you'd expect.