The creeping IT apocalypse

> Repetition is a sure warning sign. If you’re building the same integrations, patching the same servers over and over again every day, congratulations – you’ve already become a robot. It’s only a matter of time before a small shell script makes it official.

Absolutely - if something is repetitive, it's a candidate for automation. This is true across all disciplines. Only the as-yet unautomatable human judgement, insight and communication is safely valuable.

On the other hand, "go away or I will replace you with a very small shell script" has been a BOFH joke since the 90s.

I've just started my career in ML, and I already sort of feel that a majority of the work that people around me are doing can be done by an automated pipeline (like Python's AutoML) and throwing enough compute power at the problem. It's quite worrying.

This has been happening since the 1970s. Or earlier. I don’t really think of it as an apocalypse. IT skills have never had a long shelf life. Any time you are a technology expert at your company, in IT, the technology landscape will shift under you. This is the Red Queen Hypothesis in action. People who ran mainframes in the 1980s became trusted experts and then most of the jobs evaporated. Same thing happened to people running critical VAX or Unix systems. Your skills are only valuable as long as the related technology is.

The same thing happens to programming positions.

But I think the good news is missing from this article—IT jobs are, overall, sticking around or increasing in number. (According to the Bureau of Labor Statistics, the jobs are growing “faster than average for all occupations”). You do have to keep updating your skill set, but it’s not like manufacturing, where efficiencies eliminate jobs altogether or move them to completely different sectors. And there is that ageism to worry about, and uncertainty.

I’m personally more worried about some of the other remaining white-collar office jobs, like the accountants, paralegals, HR, various banking positions, etc.

Thx; Agree, though I find many "cloud-hosted ERP products" are actually very traditional ERP products, just hosted by vendor (I remember the "Lightbulb" moment when I realized the much-vaunted Oracle HCM Cloud is just Oracle Fusion tech stack underneath, that somebody else manages; in that case, I don't know if it's necessarily _fewer_ people that work on it, they just all work at Oracle datacentre instead of dispersed among the clients :)

Haven't worked on Workday and the like yet to understand how "Cloudy" they trully are.

Programmers have been predicted to be going out of jobs since the days of COBOL. There is a reason that is not happening(yet), or hardware companies would have been all shipping pluggable chips, we'd just config-connect them and be done. The reason is in the word itself `Software`.

The real problem with these ready-made plumb-and-plug modules is sooner or later these are either too slow, or expensive, or just a pain to refactor/redo. Eventually you just come back and realize you need a more granular control over things, and anything you are likely to come up with resembles a programming language.

I had this moment of realization myself while having to change a complicated graph in Pentaho Kettle a few months back. The graph looks bonkers hard and brittle, changing anything requires redoing all the dependent elements of the graph, and if you have a graph complicated enough you will be forced to rewrite it. The real trouble there is no functional/unit testing with these things. And then you realize, you are just better off with a full fledged ETL language/programming language. The second problem I faced was running into performance issues. Want to change the sort algorithm? Running into heap space issues? Want better logging? Want a better threading model? All the best. Nothing is possible.

This is above and beyond the need for meta-programming facilities. At that point whatever GUI graph you draw is worse than any verbose code you will write.

Regarding programmable tools, we already have those. Vim, Emacs, Microsoft Excel all give you a degree of meta control over the tool and what you want to do with it. But that's that, and it is often hard to bend this tools to your command.

These are just a few reasons why there won't be an apocalypse soon.