Any great research work ever succeeded only because people were able to understand it. In other words, if something makes sense to a lot of people, it probably has value. A case in point is Transformers. The intrinsic architecture is quite complicated, but essentially natural language is converted into tokens, and all tokens are compared with each other, which is then used to predict the next best continuation for any sentence. Beautiful. Although that work is groundbreaking, if we look at a more humble day-to-day research paper, it worked because people were able to use it to solve their particular problem.
The first step in getting to do valuable work is to find the right problem statement. And the most important fundamentals for any worthwhile PS come from figuring out the existing work that has been done, and finding a gap in it. Ideally, this should take up 80% of the effort being put into solving anything. The other 20% is to figure out the right experiments to back up any proposed findings. To reiterate, it is easy to find solutions for tough problems, once we know the problem and the existing effort put into it. But, why bother with all this?
If no one bothered, then we probably wouldn’t have put a man on moon. Someone in some grumpy-ass lab did all the grunt work to set the stage for everything else that was built on top of it. This is a thankless job, but that is just work like any other field. I do not mean to comment on the moral implications of it, although it might sound like it. Research is a noble pursuit. I have heard that enough times, but no one bothers to think of all the failure and hitting brick walls that accompanies it.
I started writing this just to re-emphasise to myself the importance of finding the right related work and experimental setup before trying to come up with a solution for anything. As one of my great friends put it (for some completely unrelated thing), you must be clear on what you want first, before you can go and get it. In another direction, that also helps with narrowing down the landscape to see that we are not trying to solve everything at once, but instead that we are looking for something specific.
A.