Picking the right problems to work on is also necessary, and difficult to get right. Worming in on a hard unsolved problem that is eventually incredibly important is the scientific equivalent of winning the lottery, and we know lots of examples of this. Recently, mRNA vaccines, neural networks, and the twin primes conjecture come to mind. But there are doubtless many, many other examples of people working on a possibly-fundamental problem which never actually break through, and we never hear about it.
But I would argue it's still worth making a concerted effort to identify good problems to solve, and specifically thinking about what makes a good 'hard problem.' It's much easier to make progress in an under-studied area, simply because the low hanging fruit haven't been eaten yet. The danger, of course, is that understudied things may be understudied for a reason: they're not actually relevant to anything else, or there's some secret impediment and many others have already failed.
By contrast, cracking away at ridiculous subproblems from the last generation of academics is almost always a dead end, imo; you're competing with people who already know the field extremely well, the low hanging fruit has been picked, and the obvious roads to relevance have probably already been well-explored. Unsolved problems in a long-standing discipline are often unsolved because the right tools are missing or it's impossible within the existing paradigm... Or the problem is just irrelevant, relative to the amount of effort needed to solve it.
Picking the right problems to work on is also necessary, and difficult to get right. Worming in on a hard unsolved problem that is eventually incredibly important is the scientific equivalent of winning the lottery, and we know lots of examples of this. Recently, mRNA vaccines, neural networks, and the twin primes conjecture come to mind. But there are doubtless many, many other examples of people working on a possibly-fundamental problem which never actually break through, and we never hear about it.
But I would argue it's still worth making a concerted effort to identify good problems to solve, and specifically thinking about what makes a good 'hard problem.' It's much easier to make progress in an under-studied area, simply because the low hanging fruit haven't been eaten yet. The danger, of course, is that understudied things may be understudied for a reason: they're not actually relevant to anything else, or there's some secret impediment and many others have already failed.
By contrast, cracking away at ridiculous subproblems from the last generation of academics is almost always a dead end, imo; you're competing with people who already know the field extremely well, the low hanging fruit has been picked, and the obvious roads to relevance have probably already been well-explored. Unsolved problems in a long-standing discipline are often unsolved because the right tools are missing or it's impossible within the existing paradigm... Or the problem is just irrelevant, relative to the amount of effort needed to solve it.
ramble ramble ramble