Literacy itself used to be a niche skill, learned by a select few who made an earning from it (scribes, criers), while the majority of craftsman got along just fine without it. Then the printing press came along, and it brought a proliferation of readable material, and literacy became useful. Not essential, but useful. People could distribute political literature, or send messages cheaply. Literacy bloomed. And as literacy increased, and became expected, it became essential.
We're at the beginning stage of that now. Somewhere around (probably before) the printing press. Programming literacy is becoming useful, but not yet essential.
I have a lot of friends who work in offices surrounded by Excel spreadsheets all day. I know for a fact they would be more productive if they knew how to use it better, write macros, VBA... or heaven forbid, use a database instead. I think this is going to happen more often in the future, too.
Stupid alert: I briefly worked at a startup that did push to production by remote desktop to the server and manually restarting the IIS server after copying things into place. They didn't automate it - because "it's the way we always did it." (Yes, they're out of business now.)
Automation also has the potential to make a task less error-prone and its output more consistent. A botched or forgotten manual task can be difficult to clean up after, so there's risk management too. For instance: in one incident, a coworker at a former employer accidentally ran the "send thousands of people email about how much we're about to pay them in a few days" script on input from the wrong year because January had just ended and he had to input a value containing the last day of the month (e.g. 2013-01-31) by hand.
That is, if the benefit is worth the costs.... do so. Otherwise, reconsider.
And few things make as glorious as a mistake as an automated job going out of control. There is a reason one of the most important questions you should ask on any automated task is "how do I make it stop?"
There will always be a need for people who understand computer science, but those engineers will be like compiler engineers or hardware engineers are today...enabling a much larger group of programmers to produce the software that provides value to users. And as we're able to provide improved layers of abstraction to make software development easier, we'll see computing become more accessible to more programmers who will build software with a much richer understanding of the domain.
So I'm torn about initiatives like this. I see coding as the new literacy...it will be a skill required by almost every professional who is an expert in his or her domain. But they won't be coding in today's languages and coding will be significantly easier than it is today. So we need to find a way to teach children to develop using tomorrow's tools today. And that means extracting the part of programming in today's languages that is intrinsic to programming while doing our best to avoid the part of coding that is just pain inflicted by today's limited abstractions. That separation may be difficult (or even impossible) to achieve, but that doesn't mean we should try. Because anyone who doesn't learn how to instruct a computer to accomplish tasks is going to be automated out of a job (https://www.youtube.com/watch?v=7Pq-S557XQU)
That is an extremely unusual definition of "literacy."
"Coding is the new literacy" implies that coding will be necessary for anyone who wants to interact with modern life, not just for experts. Imagine trying to use the freeway system if you could not actually read streetsigns, for example. i.e. if you were illiterate.
That's the whole point. Computers and robots are creating a world in which there really isn't room for anyone who isn't an expert. We've already automated the low-hanging fruit...we've got a fraction of the manufacturing jobs because computers and machines can do that work faster, cheaper and more effectively. Google and such are working on less low-hanging fruit (people who drive automobiles for a living, lawyers...there's a bunch of examples in that youtube clip). That bar for offering utility beyond what can be automated is steadily rising.
If you're not an expert in something, you're going to be pretty useless in the world we're creating.
Are computers everywhere and do people use them every day? Yes. But mechanical ability isn't the new literacy and just as many people interact with engines on a daily basis and would be helped by it in daily life. I think it is much more useful to think of programming like woodworking, leatherworking, sewing, soldering, etc. Very useful skills which can improve your life, but not at all necessary, and "only a few people only need to be really good at it" as you stated.