I disagree with this cartoon in that getting to the edge is not easier by focusing on only one thing. Not knowing other things is just as likely to hold you back from some insight as not knowing your specialty well enough. It's so common that people unaware of progress in other fields spin their wheels solving problems with known solutions that I think the circle should have been more of a confounded sphere where knowing a bit of EE and ChemE makes your work on BioEng 10x closer to the edge. Knowing more than one field reinforces the strength of your innovation. Simple example, how much easier is it to make progress in applied physics if you know enough EE and machining to design and build the best test apparatus in the world?
Pick a focus and whenever you get to the point that you're a strong expert (or whatever level you feel is useful) and can do the job you want to do you should reach a bit and pick up something you're curious about to add more diversity to your skillset.
Being a generalist isn't bad, I basically have a PhD in being a generalist and tons of people want to work with me because someone who only understands optics will be lost when they need to use a photodiode and a lock in amplifier or someone who wants to test magnetic properties will be lost when they don't know how to actually construct a custom magnet. Among many obvious examples.
Unless you're at a place so big that you can just glide on assuming other people will fill in the gaps in your knowledge, you need both general and specific skills. A PhD doesn't have to mean you are only capable of one thing unless you let it, even if that was the common meaning 20 years ago. There just aren't the jobs in academia to support that kind of singular focus anyway.
We're going to live long enough that it's a terrible career decision to pigeon hole yourself so badly 50 years before you're even theoretically going to retire. Especially with online classes it's only going to get more common for people to have multiple significant skills and major changes in their field over the course of their career. Required even. Very few people are going to go for 50 years with a single technical expertise, certainly jobs for such experts are far harder to find than jobs for people who can do a few things at a more modest level of expertise.
I know a ton about materials science and physics, have genetically engineered viruses, have built chemical reactors, have designed catalysts, have done machine learning, have designed lots and lots of electronics, have programmed as required (not my favorite kind of work, but I've done 8051 assembly up to ARM firmware, Matlab, Scheme, Python, and shudder LabView), am more than half decent at machining, teach karate, perform music, and make art... trying to be the best at just one thing to the exclusion of all else is unnecessary and limiting both professionally and personally.
I was terrible at all of these things when I started. The article echoes my philosophy as well -- you have to learn to learn because you want to and not because you want a grade. I got a C- in differential equations. Two years later in stat mech I was the only one of my friends who recognized LaGrangian multipliers being used (by name even) because when I took the course my goal was to learn differential equations regardless of what grade I got, so I didn't forget it all the second the final was done. If you are the kind of person who just won't do homework unless it's graded, you should work on that because it will hold you back and you're training yourself to rapidly forget information that isn't required to pass a test.
Don't be afraid to try new things that interest you. Be okay with skilled people saying you're bad at something you did your best at, grow a thicker skin since those people are the ones likely to be able to give you advice to become good at it. Professors want you to come to office hours, especially if you're struggling.
You have to be okay with getting that C-, who cares what grade they assign you as long as ten years later you still remember convolutions or whatever. Be solid at a bunch of stuff rather than extremely good at, I don't know, something super specific like the spin states in cobalt platinum alloys or matlab modeling of alloy surface energy same as you shouldn't refuse to learn anything but Python if you want to still be a developer in 2069.