The future of software? Imagine a bot, stamping on a human face – forever As we have said before, the software industry has a decades-long history of cost-cutting, commoditization, and a successful sales model of "pile 'em high, sell 'em cheap." This has worrying consequences if your skill set is the next one to be commoditized. But there may be ways out of this narrowing commercial bottleneck. The …
Hmm.
Not so sure I completely agree with everything in the article. See, the problem is, to get to a stage where you can simply automate a deployment end-to-end is hard. Really hard. It gets harder the more complex your system is. You're trying to write code to bring disparate technologies together - Terraform, Ansible, your Hypervisor - and these are not designed to be used in that way. They were never designed with each other in mind, let alone interoperability. I think, what we're doing, is just moving the technical debt elsewhere and kicking it down the road.
Say you had a few technical guys who know how to build systems. They understand the systems. They know the ins-and-outs. They write automation code to make their lives easier. C-Suite show them the door once the automation is working. Roll in a cheaper resource with the only knowledge is 'run this script and magic happens'. Which is great for a year or two until those disparate technologies release the latest and greatest features that are not entirely backwards compatible. Your deployments start to fail in weird and wonderful ways. Problem is, you have no proper techies left who can diagnose the problem, you have no techies maintaining the code, and no techie with enough knowledge to work around the problem until it can be resolved. I've literally watched this happen time and time again. If it didn't, I'd be out of a job.
So, I am not surprised that techies are becoming increasingly hard to find. On the one hand, businesses are punishing them for being technical. On the other hand, once businesses realise their mistake, they struggle to hire the technical people who have the knowledge they are after.
YMMV depending on complexity.
honestly, I don't think they'll care. C-Suites are measured by their KPIs, which is either shareholder value, or whatever it is that contributes to that value — typically cost reduction, or revenue increase within that particular financial quarter, but usually both.
that, and the fact that companies live and die based on their cash on hand, basically means that if they can delay the consequences of their actions to three years or more for making sure that they'll hit their targets, they'll do it. even if it bites them, specifically them, in the ass.
that's a problem for future them to deal with, and it'll be dealt with then, with whatever tools that they have at the time. it takes particularly strong-willed teams at the top to resist this, and it helps when they're not bound by multiple parties demanding their cut there and then.
there's a certain kind of personality type that gravitates towards this role, and it's not one that prioritises long-term thinking and risk reduction. and, like, this isn't even being judgemental about it — if you're in that situation, faced with those incentives, you'll do what it takes to fulfil those priorities, however perverse they are... or you'll be replaced with someone who will.
< "this isn't even being judgemental about it — if you're in that situation, faced with those incentives..."
Exactly. The concept that incentives matter seems lost on too many people, in business, politics, and elsewhere. Your 20-year plan might be the absolute best one for the organization, but if you're fired after 4 years and the plan is then scrapped, what good did it do anyone? Sadly, the incentives in place often make squeezing out every possible cent of profit THIS YEAR the only real choice. Next year? Rinse and repeat. Do this for a few years, then move on to the next job at the next place before it all comes crashing down. It's somebody else's problem now.
The banks have already done this (many times probably), pushed out the experienced expensive people to replace them with automation or cheap junior staff or outsourced cheap staff (or all three). Then realised that their systems were in fact held together with wet string and beard hair from the guys they just pushed out. The banks then brought these guys back but they had to offer money that couldn't be refused. To departmental budgets it might have been catastrophic, but to the business it didn't even register.
This will be the same, C-suite will be reassured by whichever sycophant is deemed worthy to talk to them that throwing out all their experienced staff will save %s, keep major shareholders happy and secure their bonus for another year. Shit falling apart later isn't their problem they have people to blame. If they have to bring back people to fix stuff in an expensive way they can blame individual departments for going over budget.
But this isn't like other industrial changes. If you worked in agriculture you could retrain and work in manufacturing; if you were a factory worker you could retrain and work in the new knoweldge industry as a computer programmer, sysadmin etc; if you're a sysadmin or a dev you can't retrain as an AI. And jobs wrangling the AIs that replaced you are going to be few and far between. And you might not be able to retain as a joiner or labourer if you're older.
TBH I think we're a while away yet from the wholesale replacement of workers by AI, most AI I've encountered so far has been pretty shit at everything and is anything but intelligent. But I'm sure the board are hearing the hype and are wetting themselves at the prospect of removing vast swathes of their meatbag employees.
"C-Suites are measured by their KPIs, which is either shareholder value, or whatever it is that contributes to that value — typically cost reduction, or revenue increase within that particular financial quarter, but usually both."
When a metric becomes a target it ceases to become a useful metric.
Another phenomena is that very smart people tend to make things harder (than necessary?). Take OOP for example, or front end. Smart guys enjoy complex syntax and high level abstractions. Think Git, Java, C++. Or they (intentionally?) build impenetrable moats around their professional fiefdoms. Or maybe the complications are necessary to reflect the complexity not visible to less sophisticated folk.
I am repeatedly shocked when a large text document hangs in gedit while Ctrl-F typing. They would probably say I am doing it wrong, because Linux is not for GUI users. By making it harder (Vim), they make it easier, as abstractions allow faster operations.
Windows has been a success because of the above, probably. And my Windows never crashed on the same PC where Linux freezes every second session. The Linux folk suggest I do some high level magic, which I do not risk doing in fear it can consume one week of tinkering "just for fun". On the other hand I cannot install newer Windows on old machines or older Windows on new machine.
"Smart guys enjoy complex syntax and high level abstractions."
If it were not for high level abstractions we'd probably have been programming plugboards these last many decades or flipping switches to toggle in binary op-codes.
The entire history of computing is a succession of higher level abstractions. Assemblers abstracted machine codes, relocating assemblers abstracted memory layout, higher level languages abstracted assemblers,processes abstracted flow of control, file systems abstracted disk layout, databases abstracted files. The only GUI programming I ever did for Windows used Delphi and Delphi's OO, specifically its VCL, abstracted all the GUI elements.
I disagree on at least two points - the simple one is that up to D2007 (the last I bought) Delphi abstracted MANY of the GUI elements - not all. In some cases you had to use Windows API calls directly.
The other point is that you are breaking the link between complex and high level abstraction. I'm not even sure I agree on what seems to be your definition of high level abstraction. A quick DDG gave me:
"Abstraction is one of the four cornerstones of Computer Science. It involves filtering out – essentially, ignoring - the characteristics that we don't need in order to concentrate on those that we do."
I'm not sure what assembler or the other high level languages would have filtered out.
Another one:
"the action of removing or separating something from a place or context (= the situation, facts, words, etc. that exist around something), or something that is removed like this:"
I think has *has* told us he doesn't know what he's doing - which is the whole point! He *DOESN'T* want to spend weeks tinkering just to be able to *USE* his computer, he just wants to *USE* the damn thing.
Just like I just want to get into a car, turn it on, and drive it. I'm not a car tinkerer, I do not have the skills or patience to have to "fiddle" with a car in order to get to work or some groceries. Same with printers. If my printer is dead, I replace it. Yes, I'll offer it to anybody who wants to tinker with it before I dump it, but when I ask "does anybody want this printer, it tears the paper and dumps toner everywhere" the *WRONG* response is "if you take the top off and undo these screws....."
*NO* I want a printer so I CAN PRINT STUFF. *****NOT****** to tinker with and get in the way of actually getting the printing work that I want to do done.
Similarly, the OP wants a computer so he can *USE* the computer, not to tinker with it and get in the way of actually doing the computer stuff he wants to do.
If they don’t know what they're doing, they shouldn’t be fucking about under the hood of the car. If they can avoid that, Linux wont freeze frequently. I have multiple Linux computers, I don’t fuck with them, they just work.
I have Windows for work and my Mrs has Apple. They all work, just don’t fuck about with them, it’s really easy.
If you do want to change stuff, make sure you know what you’re doing.
Smart guys enjoy complex syntax and high level abstractions.
I think you have that backwards. I suspect that what they really like is simple syntax and as few abstractions as possible, because the more abstraction layers there are the harder it is to understand what's being done.
The problem is that low level stuff tends to have really complex syntax out of necessity because you're having to do all the difficult bits yourself; and while high-level stuff has nice simple syntax, it's frustrating because it regularly won't do what you want it to but rather it does what whoever wrote the abstraction thought you needed.
Getting that balance right is really, really hard.
In order to actually make use of all this whizzy technology and automation somebody somewhere actually needs to be able to clearly articulate what the bloody machine needs to do.
When AI reaches the point that it can translate a set of requirements that can be summarised as "we need a thing... to do ... a job, no not like that, more like ... umm, well I'll know it when I see it" and it can reconcile the fact that because there are half a dozen voices involved in these requirements who all see the problem differently and therefore want the job done differently (or more likely not at all) then we're really in trouble.
A few decades ago now, I was working on a system for a large UK government department. We had an all day workshop with a group of 'empowered users' to agree on feature changes for the next release. After several hours turning one of the proposals around everyone in the room decided the feature was in fact useless and it was removed from the requirements.
The next day there was a phone call from someone very senior in the department to say ignore them, they weren't empowered to make decisions just to rubberstamp the original list..
And if you were very lucky you didn't have to tear up what you'd written to fit it in. I had that about a fortnight before go-live. It converted a simple product list into one where the BoM had to be taken into account with some hastily written recursive code. At the end it worked but I wasn't convinced I understood it in detail once I'd finished and any changes were strictly forbidden.
> Microsoft pushing hard for over a decade to insert scriptable layers in between the GUI front-ends and the server back-ends. This is the main reason PowerShell exists.
Bash + Python would provide the same functionality. Whatever happened to all those complaints about Linux and old-fashioned config files.
I always got the impression that powershell exists out of some sort of institutional stubbornness and refusal to admit when they are wrong.
Like, after a decade of Windows admins moaning about not being able to properly script common administrative tasks someone at Microsoft finally cracked and said: "OK, fine, we'll concede that we do need a shell for administrative scripting, but we're still not going to ship one of the existing ones that you all already know how to use. We're making our own. And it'll be better. And completely different. For reasons. And you'll all have to learn everything all over again. So there."
Well, it's fortunate that none of our jobs require "thinking" isn't it?
I have been experimenting with LLMs to see what they can produce, and I have to say that none of them are even close to producing real world applications for complex business domains of the kind I've spent much of my career writing custom apps for. This is because they're not trained at all on business processes for individual businesses, which are mostly highly specific and also not publicly available for LLM training.
Amusingly, I got one to generate a high frequency traiding bot for currency dealing, but it was just the basic structure for the app, and had a comment saying "put your trading logic here". It's at this point that all us experienced programmers who've been replaced can expect to get the phone calls asking if we'd be available to do a bit of work, and can start quoting hourly rates in the thousands.
Seems to me that you're indeed taking the right steps to remain relevant, as "Terminator AI" guns for our jobs, by taking ownership of your (our) evolving means of production. Perty much all ElReg readers have the ability to successfully tame that RotM by becoming better bullriders of those wretched cybercreatures imo, by mastering key roughstock slide flexions and spine rotations (metaphorically) through Tobias' LLM fitness training regimes for beginners, intermediate RAG rodeo artists, and advanced agentic "cowpersons" for example.
Can't let this coming AIpocalypse of automated wealth concentration pass us by without a fight can we!?!? That this tech currently manifests mostly as crapola is doubtless a lurid ruse ...
(PS. if your job is counting blood cells for healthcare companies, consider owning a Countess, instead of having it stamping on your face - forever ... ouch!)
I asked a couple for Pascal* code to generate a UUID of a specific type (might have been type 1 or type 5). One generated a output statement with a literal UUID embedded in it. The produced code for a type 4 which wasn't what was asked for.
"Sneakily trying to see if they would spew out code from the Free Pascal Compiler library.
At least it didn't simply reproduce the FOSS library code which slightly surprised me.
Ultimately, training a LLM to do that sort of thing would need more than feeding it training code. It would also need to understand the RFCs and that's a completely different task. Unfortunately the library failed on that one: if you're using one of the MAC/timestamp formats and take a random or arbitrary string instead of a MAC address (frequently done for security purposes) you should set the least significant bit and last time I looked it doesn't. I suppose I should flag that up..
"It's at this point that all us experienced programmers who've been replaced can expect to get the phone calls asking if we'd be available to do a bit of work, and can start quoting hourly rates in the thousands."
That point has passed a while back and it is the standard MO we have used for the last 30 years. Get the last of their breed in to fix a few remain faults then wave goodbye to therm.
Do keep up. The latest ChatGPT(ARSEHOLE NAME) model improves the prob rates by a LOT!
Howard, blah blah. I was going to be stoic in repsonse but it isnt worth it with you. And a silver medal, too.
OK.
"Well, it's fortunate that none of our jobs require "thinking" isn't it?"
Understand that intelligence not only scales up and down, but also left and right of those up and down. It is 3d. In simple minds terms, your light switch IS intelligent.
I'm sure oil companies paid better than archaeologists.
You're right there. I got a degree in Archaeology and soon gave up as the pay was pitiful, no future in it at all.
Took a graduate conversion course in systems analysis and design and then got a job working at an oil refinery in the IT department. Best paid job I'd ever had and the work was enjoyable too. Bonus!
So yes, oil companies do pay better than archaeology.
Archaeology you do for love.
I too have a science degree (Neuroscience and biochem) from the late 90s and went into IT because in those days, without a PhD, science jobs in the UK paid less then the checkout at Tesco and had considerably less job security. I don't believe much has changed in the intervening 25 yr.
In fact about 50% of my final year ended up working in IT
It was always my view that as the Civil Service was the largest employer of scientists in the UK, at least back in my day, a bit of over-production of graduates was enough for them to justify keeping the salaries down. They got very worried when scientists started emigrating. It would upset the market they'd constructed.
If it makes you feel better, marketing seems the most likely to be completely replaced by these "AI" systems, leaving no humans whatsoever. It's already started.
Of course, it also means we're all already being bombarded by huge amounts of this type of marketing, all across the Internet...
There is always a doomsday prediction and it does get everyone to turn heads and then again go on with the flow.
Sure commoditization is always the game of the market forces, it also creates new areas and opportunities in that place. The skill requirements reduce for sometime, but that automation needs upgrades as well and this area requires new skilled staffing. This is the same story always and the cost of staffing follows this curve. New sophistication brings new areas , tools and skills.
Sure, AI is a different one, but all processes could not automated with single AI and interoperable models across AI's are still limited. The complexity of new landscape means requirement for skilled professionals to debug the issues and maintain and that could not be removed, it could be postponed in the name of cost savings for a couple of years and later struggling to keep up. This seems to be the current strategy for all organizations to shore up share price.
For example,In this case a customer service bots, they have reduced human staffing and increased availability of support. but added on the IT budget for hardware and maintenance of these bots. Also the recruitment for new skills other than customer support. It did add value but the benefit analysis should also include the added complexity for technical landscape and organization process.
I think if I were a career advisor today I would be recommending the trades as a more profitable and secure option to your aspiring young lad or lady.
With hindsight I would have been in a far better situation today if had trained as an instrument fitter or industrial electrician and certainly plumbing would have been a lot more lucretive.
As one of Douglas Adam's transdimensional white mice (HHGTTG) facing a similar dilemma opined "on the other hand just taking the money and running, then I for one could do with the exercise.”
I was just thinking being a godawful poet might be a sought after skill in a post AI world as every man and his dog with the assistance of ChatGPT could produce very passable verse (perhaps not for a Vogon nor Paula Nancy Millstone Jennings*. ) Perhaps then being very bad at something will be very good... pretty much like C-suite manglement today.
* a specimen The dead swans lay in the stagnant pool./ They lay. They rotted. They turned/Around occasionally./ Bits of flesh dropped off them from/ Time to time.:And sank into the pool's mire./ They also smelt a great deal.
We may not have money, but I'll barter my ability to fix a faucet for some of the eggs your chickens are laying...
Maybe we could establish an IOU of sorts, some object that the community agrees has value, so we can trade our services and goods for that instead of having to find someone who both wants what we have and we want what they have. Since it would keep our record of debts current, maybe we could call it "currency".
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
