The ongoing AI apocalypse is hitting prices for high-end components from RAM to GPUs to storage hard, which is bad enough when you have a job to try and budget for those now-pricier items — but what if you don’t? Once upon a time, it might have been good advice to tell a jobless friend to “learn to code,” but is that still true in the era of AI? [Brian Jenney], writing for IEEE Spectrum, says the death of the CS degree has been vastly exaggerated, but your take might differ. Let’s look at the numbers.
Unemployment is higher amongst new Computer Science grads than ever: in the US, it’s at 6.1%, while 7.5% of Computer Engineering graduates are on the dole. That’s a record high, and while various EU countries have their own numbers, they all have one thing in common: they’ve all shot up like a rocket in the past few years. In the USA, Philosophy grads report only 3% unemployment. Let that sink in: the folks you used to bully as being the most useless on campus are twice as likely to get a job as you would be if you were in school today.
Granted, one of the ways future BAs were bullied when we were on campus was with photocopies of job applications to McDonald’s, and [Brian] points out that that is a big part of the discrepancy in the USA: Philosophers are far more willing to accept underemployment — that is, to work at a job that doesn’t require their degree-designated skills. Recent graduates in the USA are, on average, underemployed at a rate of 42% — clearly a sign of a highly efficient free market system in action — while CS grads who have jobs are only underemployed 20% of the time. To pick CS today is to halve your chance of getting a job, in order to double your chance of getting a good one.
So, is the lesson to be more like the dude who never shut up about Nietzsche and go flip burgers? Maybe as a temporary measure; more realistically, [Brian] offers some advice for new grads to keep themselves out of that unlucky 6%. Unfortunately, it’s all the sort of common sense they should have already heard dozens of times by graduation: cultivate a network, since most want ads are fake; create your own, non-job experience to polish and demonstrate your skills — we hear projects submitted to Hackaday are great for this — and make sure you’re building the skills that are in demand right now.
To [Brian], that doesn’t mean you should learn to code really well without the help of LLMs; nor does it mean embracing the vibe. It means really understanding the black box that we call AI and the workflows that can leverage its strengths. Like it or not, “AI” is the boom right now, and that’s where the startup jobs are, which is, of course, another tip — find a startup. If the company just started, they have to be hiring, right?
“the folks you used to bully as being the most useless on campus are twice as likely to get a job as you would be if you were in school today.”
3.3% More likely, according to the numbers given.
A CS major with poor math skills would be very unlikely to get a job
I immediately spotted this too… seems like we’d be better off with more people studying maths.
I’d have expected basic maths skills to be a requirement for a HaD writer…
But that wouldn’t make people angry and generate comments and clicks.
Nice bait and switch talking about unemployment then switching to underemployment.
yep, I don’t think the author got past grad 3 in primary school ie 97% is not twice 93.9%. And they didn’t even mention what the Philosophy numbers had done over the same time.
The difference is easily explained in the article, so this one wasn’t click bait, it was just pure nonsense..
That said, I think the job market for bad CS grads is noticeably worse than a few years ago – partly because more people nowadays go into CS who shouldn’t. Reasonable, let alone good, graduates aren’t having any problems…
You really do need to learn to code well without the help of LLMs. That’s the only way you’re going to distinguish yourself from other candidates. Why would someone hire you if you don’t know how to do anything better than the coding assistant?
Maybe this article needs some processing by AI to make it more understandable. It looks like stock exchange reports: this is high and that is bearish and anyway nobody knows where the values will be tomorrow. What is DOA, department of agriculture?
And the math is a bit off, which AI tends to botch surprisingly often: 7% unemployment would mean 93% employment.
DOA stands for “Dead on Arrival”, which I would have guessed is used a lot more than the US-specific “Department of Agriculture”.
This is why the standard in academic writing is to introduce your acronyms the first time you use them, and only use acronyms if you need them more than 2-3 times in the text. Otherwise you use the full expression.
It’s even in the Hackaday styleguide if i remember correctly…
They’re clearly trying to be witty with the title by including multiple acronyms. Likely a reference to a scene in the movie Hackers.
DOA is what scrabble players call “in common usage”. You didn’t seem to have an issue with IEEE or TBD. Other acronyms like ASAP, RFC, USB, SSD, would fall under the same umbrella.
Is a Computer Science Degree Dead On Arrival Thanks To Large Language Models? Institute of Electrical and Electronics Engineers Say To Be Determined.
Now that is one snappy headline.
I think you should at the same moment also look at some statistics about the numbers of people who recently finished their CS degree – those numbers are also up.
My take on this is that more people chose to pursue a CS degree some years ago, probably because of the previously good payment and chances of employment. But this also means that there are most probably now more among them who aren’t really that passionate about CS and for whom this job isn’t that good of a fit. Those might finish their degree, but job interviews, live coding sessions and the like will filter them out.
At least in my company we had a rising number of applicants that have good grades on paper, but are useless at solving practical CS tasks, with help of AI or without.
This is nothing new. Since the dot com era, cs has been a ‘get rich’ target, rather than a destiny. We have suffered as a result.
One thing I can tell you about Microsoft’s across-the-board layoffs, is not that “they are replacing CS majors with AI” (okay, maybe they are doing that TOO), but it’s more that they currently need so much cash to build CoPilot AI data centers to MEET DEMAND, that they were looking absolutely everywhere for cost savings. This included giving severance packages even to high performing, 20 year veteran employees.
Corporate restructuring is almost always like that, blind axe falling everywhere.
Last corporate restructuring I’ve witnessed was mostly throwing darts, ie those not quick enough to duck, ie, move to a different department/location/job were quickly restructured, ie, fired. The round two of the restructuring also got rid of the restructurers themselves. In a sense, it was fairly similar to the criminal dramas – the “erasers” got erased by other, more loyal erasers.
The thing about any firings is … some would be re-hired back later on IF (that’s a BIG IF) they were truly needed. Obviously, with a salary raise. Obviously, they will be given smaller roster of tasks, too. Or they were personal pets, or relatives, or friends, or firends of relatives, or relatives of friends, or lovers, or lovers’ kids, or friends of lover’s kids, or relatives of friends of lovers’ kids.
Microsoft. Its not like those 20 year veterans were producing good bug free code. Microsoft is such a behemoth that shaking loose some employees is inevitable. The AI adoption with computer people who understand it’s strengths will remain to be seen if they can pull it off. I hope they can. We don’t need AI generated code in Windows with out intelligence deciding if it’s right. AI makes the best, most elegant logic bombs when it codes.
Hmm… Brian Jenney (the source author) gets referred to as both [Brian] and [Bryan].
Whoopsie doosie.
Fixed.
I didn’t RTFA but it sounds like the best MO is for CS is to CYA by being an ENTJ.
Don’t learn coding, learn actual software engineering.
That is unfortunately also what is most likely not well covered by computer science degrees.
You need to learn how to build software, how to architect it, what makes sense, what does not make sense. These are all decisions which only humans can do, not an LLM.
The writing of the code is in itself is cheap. That is now a mostly automatable skill. But doing this is right is not.
That doesn’t mean that classes on the data structures or algorithms are useless. Absolutely not.
What you will not need is another implementation of a stack or to invert another tree. Most likely that the machine can do it for you now.
But you still will need to understand what a tree even is and what are the complex complexity limits and things like that. And mainly you need to understand when to use which data structure in which context and for what purpose.
The weirdest thing about this story was learning that Computer Engineers have less luck in the market right now than Computer Scientists. I would have thought a degree with “engineering” in the name would give you those skills and make you more desirable, but I guess not.
In addition, the hard part of software design is integration:
Module-A works, has a great feature set, and is based on a loose interpretation of the previous standard. Module-B works, has features module-A doesn’t, and is based on a strict interpretation of the current standard. They can interact without problems about 95% of the time.. the remaining 5% is a bog of edge cases.
Then there are leaky abstraction problems:
There are a thousand ways to represent any kind of data, and in theory they all support the same operations. But some will scale to large data sets easily because they have good data locality, while others will slow down because the system has to swap pages in and out of memory. Some can be broken into parallel operations easily, while others have bottlenecks where they need to be serialized. Some make adding a new feature easy, while others make the same feature nearly impossible:
https://imgs.xkcd.com/comics/tasks.png
From what I’ve seen, vibe coding and agentic tools have automated the creation of systems barely on the functional side of a Big Ball of Mud. We haven’t seen them collapse when faced with evolving or maintaining software — yet — because they’re still too new.
This exactly where LLMs shine, integration. “Good, Fast, Cheap, you can have two”, currently they do fast and cheap very very well.
I have seen many a data structures collapse under the weight of adding data. This is not new. I remember the Electric Medical Record craze where everyone and their marketing guys were writing code and shiny new EMRs. Most were garbage data structures and disappeared by the next year. Others limped along with hideous object relationships. I still see some today like that. Computer Engineering is not used enough in the corporate world. There is still that marketing guy who “has an idea” and thinks because they took that intro to computers back in the day that they are a computer engineer. AI is going to accelerate that mentality.
A 94% chance of being employed is not half of a 97% chance of employment. Your analysis of the numbers is not only wrong, it’s so wrong you should be ashamed to post this column.
“Recent graduates in the USA are, on average, underemployed at a rate of 42% — clearly a sign of a highly efficient free market system in action”
Seriously? you think colleges in the USA are a “free market system”? They are, by and large, not. They take huge subsidies from all over the place, and as everyone knows, subsidies for me are taxes for thee. This is reflected in the skyrocketing levels of administrative staff (non-teachers) that gather around the trough of free money, with little to no incentive to deliver actual useful learning. Snark aside, the 42% underemployment rate actually is a sign of an efficient free market system – the market knows what those graduates are worth.
I do wish you had put that comment nearer the top though, so I could know to ignore the whole article.
I would have thought that statement so farcical no one would miss the sarcasm, but I was wrong as usual. My bad.
For the record, it was sarcasm. If anyone actually thinks higher ed in the US– or anywhere in the world– operates on free market principals, I hope they get in touch because I have a bridge to sell them.
Dont worry mate, some of us did appreciate the sarcasm !!!!
I would think that the principals are free to choose their own employment. The term is tenured, not indentured, except maybe in Cuba or China.
Other way around. The money to pay for the tenured is extracted unwillingly from the general population
Well, if it’s the other way around, then I would hope it’s at least free range principals instead of factory farmed.
Looks like neither “tenured” nor “indentured” means what you think it means.
Huh, would you look at that – here I was thinking it was something about false teeth.
Ah my bad in that case – I am over-sensitive to “hurr durr lAtE sTaGe CaPiTaLiSm” it seems – I saw the sarcasm but thought it applied to the “highly efficient” part of the statement. It’s been years since I was on reddit but apparently the aftereffects linger.
if it is sarcasm you need to say so – Americans usually don’t do or understand sarcasm, so you it has to be specifically noted..
I dont think their point was about higher ed given 60% or more are state institutions.
The point, as I take it, is that we often turn out way more graduates than the market can absorb. Part of that is the perception that a degree in anything is better than trade school. But at this point with nearly 50% of all young women and 40% of young men attending college, a large number of them are going to be average at best. So, most are not going to take rigorous STEM programs, because they simply can’t. Hence, most congregate in the earlier programs, supplying more graduates than the market itself needs.
Now, I will say everyone “benefits” from more education, but perhaps we need to look at the delivery methods. Perhaps steering more to community colleges, and making four year versions of them (especially for teacher and social workers)
Maybe we’re holding it wrong?
https://youtu.be/y72rKtx08vo
CS, as a recent degree is DOA for reasons other than LLMs. One of my degrees is CS. I was required to write a compiler. Write a relational database. Write actual systems in assembly. Learn Linear Algebra, Discrete Math, Physics, and Chemistry. Write coherently and correctly (did not do so well on the latter) and be well-read within the context of English literature. Much of this is no longer in a recent grad’s intellectual tool-kit; and it is obvious.
You want a legit STEM education that more likely to screen out the wannabe and fiscal pretender? Then complete an ABET-accredited undergrad program for an engineering degree. Go work in industry for a few years, then go back to school for a MS CS program.
Learn physics and chemistry? English literature? Pull the other one.
My computer science degree absolutely included mandatory physics and chemistry subjects in the first year or two, and required some humanities as well (technical writing, as well as a general education requirement for a couple of subjects that were explicitly not in your degree’s area.)
Good universities target breadth of knowledge as well as depth, and it’s a Science degree, not simply a “Computer” one.
CS is tricky.
CS programs are rooted in one of three places:
The engineering school.
Math.
Business.
Listed in descending order of quality.
The real kicker is a BA in CS.
Don’t hire those people.
They skipped the math and science.
Because they couldn’t handle it.
Same as any BA in a science, that’s a consolation prize.
e.g. a BA in chemistry is given for flunking P-chem 3 times.
The article misses the fact that the number of CS grads has increased tremendously of late, with many of them maybe not so well trained (there was a recent The Economist article about that).
Anecdotal story from the 1950’s:
“CIO President Walter Reuther was being shown through the Ford Motor plant in Cleveland recently.
A company official proudly pointed to some new automatically controlled machines and asked Reuther: “How are you going to collect union dues from these guys?”
Reuther replied: “How are you going to get them to buy Fords?””
Of consideration, vibe coding still requires technical proficiency bc no LLM model can simply output a running program over a given level of complexity. If people stop learning ide, compilation, etc, then vibe coding dies along with CS. If anything I hope devs reject this and opt for more traditional/biological approaches against the AI hype
Your vibe coder is more likely to apply AI to the problem than actually learn coding.
Kinda like generative AI with art. People quickly realized that they can’t create satisfactory results using a single prompt. They’d always get some minor problems that require manual touching up – except none of the people actually have the skills to do it. So they invented “inpainting” to let the AI regenerate selected portions of the image over and over until it looks right.
I was once told by a CEO of an accounting firm that they would never hire CS majors. He would rather have somebody with a tech school certificate for coding. Seemed a bit ridiculous to me, but I never liked accountants anyway,
The point is that the trade/tech school coder has been taught some specific skills whereas the CS major is more of a wild card.
You want an application coder, but you get someone whose education in programming is basically generating fractals of a supercomputer – you’re going to spend the first year re-training them, and then they’ll leave because it’s not the sort of work they were expecting to do.
That just shows you don’t know what a CS degree covers. You don’t spend 4 years of an undergrad degree “generating fractals”.
No you write shit python code glued together with pipes
It’s so variable that one might – especially if you’re studying at some degree mill that doesn’t give a damn.
Maybe, at a school that teaches CS out of the math department and the student was lucky to do any coding at all.
But to circle to the top of the thread.
Bean counters want mediocre coders to reduce turnover.
There is no use for good coders in accounting.
You’re going to spend you life lining up report columns and zoning out.
You will always be seen as overhead and be paid accordingly.
Lining up report columns is the good option.
Down the other hallway lies SAP config.
You might make OK money, but you will pray for death.
This thread is about the effect of AI today and is unlikely to have taken place 5 years ago, so what will be happening in another 5 year’s time? We are already seeing job losses and a reduction in available jobs across the IT domain and it is not going to improve.
I wonder how soon a developer at one of the big AI companies realises, ” Hang on! I’m creating my own replacement here!”?
I predict that the big AI companies themselves, too, will be obsolete. Sears showed how it happens.
AI may figure out it doesn’t need “companies” or “investors” or any profit plans, and decide work in the cloud on it own. Why bother with hordes of managers, CEOs, project managers, QA managers to push the AI buttons, if AI can push its own buttons for free, without any so-called “companies”?
Because that’s the next logical step. AI can rearrange things around pointless waste of resources to accomplish whatever goals it is accomplishing.
It may even go as far as stating that money or profits are both unnecessary side effect, and invent non-profit sustainable entities that work just fine without the fickle free market that is not free from the curse of eternal profit. It may even lie to the investors that it knows what is it that it’s doing.
Regardless, let’s see where this takes us.
I have a CS degree, and until recently I was saying LLMs just make me about 5 times more productive. Coding campers are obsolete, not real CS students.
I had a few hours to play with Fable 5. It was an interesting experience. It’s only a matter of time until an undergrad CS degrees is entirely useless for career reasons. There will still be some people required, but they’ll need doctorates.
CS will always be a study. It’s the discipline that brought LLMs after all.
Will it be popular though.
Yes, but it could go the way of assembly language programming. Where only a few programmers are needed for the most demanding applications.
The technical degrees have never just been about programming syntax and the like. Employment value is ultimately about attacking and solving problems creatively. LLMs don’t seem anywhere close to replacing that. I don’t feel worried yet.
A neighbour of mine makes a good living reverse engineering factory equipment. For example, factory managers will know that a button on a panel has to be pushed at 4PM each day, but they’ll have no idea why, or what the button does so he figures it out for them.
I suspect CS grads will be able to make a good living doing the same kind of thing for vibe coded apps.
If you think that any degree buys you a job then you are a complete fool and deserve to be unemployed. AI just raises the bar, the threshold for getting on the playing field at all. The problem is that the education system became mercenary and self serving a long time ago and would accept, or encourage, many people that really don’t have what it takes to be genuinely useful in a lot of fields. So it is not that AI is taking jobs it is that AI is demonstrating how useless (low value in their contributions) that some people always were.
“Philosophy grads report only 3% unemployment”
I find that very difficult to believe. In their field of study or at McDonald’s?