The accounting note is not true in the traditional sense. The field in the US is just getting offshored to India/PH/Eastern Europe for better or for worse. There is even a big push to lower the educational requirements to attain licensure in the US (Big 4 partners want more bodies and are destroying the pipeline for US students). Audit quality will continue to suffer and public filers will issue bunk financials if they aren't properly attested to.
The reports from the usual "offshoring centers" aren't exactly inspiring. It's a bloodbath over there.
Seems like the capabilities of current systems map onto "the kind of labor that gets offshored" quite well. Some of the jobs that would get offloaded to India now get offloaded to Anthropic's datacenters instead.
It's amusing to see programmers in the US promoting remote work.
Do those people really believe they're the most intellectually superior to the rest of the world? If a job can be done purely remotely, what stops the employer from hiring someone who lives in a cheaper place?
And being "superior" doesn't necessarily mean extraordinary coding skills. The vast majority of code to be written doesn't require that. What it requires however is a combination of common sense and good understanding of the underlying business. This is in short supply in many of the locations the jobs are being offshored to. But let's be honest, it was also on short supply in the corporate IT departments being offshored, though not quite to the same degree.
You’ll get downvoted but in my experience, which may not be representative of the entire population, this is true.
A mid-size US tech company I know well went fully remote after a lot of insistence from the workforce, prior to the pandemic they were fully in office.
Soon enough they started hiring remotely from EU, and now the vast majority of their technical folks are from there. The only US workers remaining are mostly GTM/sales. I personally heard the founder saying “why should we pay US comp when we can get extremely good talent in EU for less than half the cost”. EU workers, on average, also tend to not switch job as frequently, so that’s a further advantage for the company.
Once you adapt to remote-only, you can scoop some amazing talent in Poland/Ukraine/Serbia/etc for $50k a year.
I think most programmers in the US simply don't realize how much they earn compared to the rest of the world.
I'm not talking about rural Chinese villages whose name you can't pronounce. I'm talking about highly educated programmers who can communicate fluently in English, in cities like Beijing or Munich. If people in SV know how little their counterparts make in these places, they'd be much more opposed to remote work.
And that was before LLM. Today practically the entire planet can write passable English.
How many of these jobs are getting offshored because of AI?
Language barriers, culture, and knowledge are some of the biggest challenges to overcome for offshoring. AI potentially solves many of those challenges
Language barriers: LLMs are language models and all of the major ones are built in English, speaking that language fluently is surely a prerequisite to interacting with them efficiently?
Knowledge: famously LLMs "know" nothing and are making things up all of the time and sometimes approximate "knowledge"
My experience: hosted LLMs are very good, but even 30B models you run locally are quite poor (at least in Romanian). To some degree they still hallucinate words (they don't conjugate properly sometimes).
Language barriers: The outsourced workers I know use AI to help them ask and answer questions about things in English they don’t perfectly understand because English is their second language. They use it to write better English from English with grammatical mistakes
Knowledge: True to an extent, but my assumption here is that it would be used to fill in gaps or correct misunderstandings. Not wholesale doing my job. At least that’s often how I use it
I worry things will be lost in translation (maybe would have already), Or the LLMs will fill in the gaps with wrong information, like some sort of weird telephone game.
That said, I have one ESL on my team who uses LLMs a lot like that and it's fine so who knows.
I think you are conflating 2 things. AI could be going after new entry level jobs in software engineering. I am not a professional engineer but an accountant by trade (I like writing software as a hobby lol) but this article looks like evidence that IIT grads will have a harder time getting these jobs that AI is attacking. My comment rests on the fact that the report doesn't really reconcile with AI destroying entry level jobs for accounting, but rather this type of work being offshored to APAC/India. There are still new COEs being built up for mid cap companies for shared services in India to this day and I don't mean Cognizant and Wipro, but rather the end customer being the company in question with really slick offices there.
I think the article doesn't really prove AI is the culprit but I think this other article disproves that offshoring is. If offshoring was the culprit why is it only affecting the most junior employees? I think the case is still open but AI is the leading candidate.
Do you have any actual evidence that supports the headline? The article does not. It simply mentions 13% decline in relative employment and then blames AI with no actual evidence. Given what I know about the current state of AI and off-shoring, I think off-shoring is a million times more likely to be the culprit than AI.
The entire account department at my firm has moved to Poland. That’s nice for them, but as a US citizen it does mean the writing is on the wall. On the plus side I learned a fun fact. Malgorzata is a more common name than I had ever imagined.
IT help was outsourced to India years ago. I expect them to be replaced with AI the minute their government stops handing the firm big contracts because I’ve never spoken to anyone from that group who was actually better than a chat bot.
Have you seen how the profession has worked post SOX? Did you know 2016 was the peak year where you had accounting students enrolled in uni in the states? I want you to think laterally about this.
Maybe for software engineering but not for accounting. I've had to interface with many offshored teams and interviewed at places where accounting ops were in COE centers in EU/APAC.
Every executive publicly saying obviously* false things like X job will be done by AI in 18 months is putting downward pressure on the labor market. The pressure is essentially peer pressure among executives: are we stupid for continuing to hire engineers instead of handing our engineering budget to Anthropic?
* - Someone should maintain a walkback list to track these. I believe recent additions are Amodei of Anthropic and the CEOs of AWS and Salesforce. (Benioff of Salesforce, in February: "We're not going to hire any new engineers this year." Their careers page shows a pivot from that position.)
I wonder how much this actually matters? I understand that for an auditor, having a quality reputation matters. But if all audits from all firms are bad, how much would the world economy suffer?
Likewise for the legal profession, if all judges made twice the number of mistakes, how much would the world suffer?
> Likewise for the legal profession, if all judges made twice the number of mistakes, how much would the world suffer?
Is this hyperbole? It seems like the real question being asked here is "would the world be worse off without deterministic checks and balances", which I think most people would agree is true, no?
I read it as assuming the deterministic checks and balances are already absent. We have the illusion of determinism but, in practice, audits (and justice) are mostly theatre as it is.
From that perspective, lowering the quality of something that is already non-rigourous might not have any perceivable effect. It’s only a problem if public perception lowers, but that’s a marketing issue that the big-4 already have a handle on.
The current system is not long term stable, and poor accounting is part of the reason more people don't know that. Even worse accounting would speed up the decline.
Then you would have to think twice about the company you may be giving money to (ie the stock market and private bank loans). That's the whole objective of this. Every company is going to need an accountant in one way or another and you don't really need to follow strict GAAP for management requirements (what else is EBIDTA for if anything?), but it's something completely different than saying: I made x dollars and spent y dollars, here is what I have and what I owe, please give me money.
At the end of the day it is a question of convenience/standards, if GAAP didn't exist maybe firms could use a modified accrual standard that is wholly compliant with tax reporting and that's it.
I'm not sure it's that our job is the most automatable, but that the interface is the easiest to adapt to our workflow.
I have a feeling language models will be good at virtually every "sit at a desk" job in a virtually identical capacity, it's just the act of plugging an AI into these roles is non-obvious.
Like every business was impacted by the Internet equally, the early applications were just an artifact of what was an easy business decision.. e.g. it was easier to start a dotcom than to migrate a traditional corporate process.
What we will see here with AI is not the immediate replacement of jobs, but the disruption of markets with offerings that human labor simply can't out-compete.
> I'm not sure it's that our job is the most automatable
I don't know. It seems pretty friendly to automation to me.
When was the last time you wrote assembly? When was the last time you had map memory? Think about blitting memory to a screen buffer to draw a square on a screen? Schedule processes and threads?
These are things that I routinely did as a junior engineer writing software a long time ago. Most people at that time did. For the most part, the computer does them all now. People still do them, but only when it really counts and applications are niche.
Think about how large code bases are now and how complicated software systems are. How many layers they have. Complexity on this scale was unthinkable not so long ago.
It's all possible because the computer manages much of the complexity through various forms of automation.
Expect more automation. Maybe LLMs are the vehicle that delivers it, maybe not. But more automation in software is the rule, not the exception.
RAD programming held the same promise, as did UML, flow/low/no code platforms.
Inevitably, people remember that the hard part of programming isn't so much the code as it is putting requirements into maintainable code that can respond to future requirements.
LLMs basically only automate the easiest part of the job today. Time will tell if they get better, but my money is on me fixing people's broken LLM generated businesses rather than being replaced by one.
Indeed. Capacity to do the hard parts of software engineering well may well be our best indicator of AGI.
I don't think LLMs alone are going to get there. They might be a key component in a more powerful system, but they might also be a very impressive dead end.
This has been my argument as well. We've been climbing the abstraction ladder for years. Assembly -> C -> OOP ->... this just seems like another layer of abstraction. "Programmers" are going to become "architects".
The labor cost of implementing a given feature is going to dramatically drop. Jevons Paradox paradox will hopefully still mean that the labor pool will just be used to create '10x' the output (or whatever the number actually is).
If the cost of a line of code / feature / app becomes basically '0', will we still hit a limit in terms of how much software can be consumed? Or do consumers have an infinite hunger for new software? It feels like the answer has to be 'it's finite'. We have a limited attention span of (say) 8hrs/person * 8 billion.
If it gets to the point where I can no longer find a tech job I am just going to buy a trailer, live somewhere cheap, and just make money doing odd jobs while spending most of my time programming what I want. I don't want to participate in a society where all I have for job options is a McJob or some Amazon warehouse.
This is the best thing engineers can do. I moved to building as a solo founder. I am building an LLM enabled coding product and I teach. I'm hosting a session on Claude Code today, 134 guests signed up. I'm gradually planning to make money teaching for a few months while building the product.
That's actually a good idea. Now I just need to come up with an idea for an SAAS app. I was thinking originally or making one of the games on my project backlog and seeing how much I could make off it. Or creating one of the many idea I have for websites and webapps and see where they go.
Would be more difficult depending on where you live. My plan was to talk to others online and see if I could find someone willing to live such a simple life with me, maybe starting with an LDR first (I'm sort of doing that already)
Unrelated to the discussion, but I love these kinds of backup plans. I've found that most guys I talk to have one. Just a few days ago a guy was telling me that, if his beloved wife ever divorces him, then he'd move to a tropical island and become a coconut seller.
(My personal plan: find a small town in the Sonoran Desert that has a good library, dig a hole under a nice big Saguaro cactus, then live out my days reading library books in my cool and shady cave.)
Mine is forrest fire fighter. Surely with climate change there will not be a shortage of work, and while dangerous and bad for you, it seems kind of fun.
The future seems very uncertain right now and we are living in weird times. Its always a good idea to have a backup plan in case your career path doesn't work out!
AI training costs are increasing around 3x annually across each of the last 8 years to achieve its performance improvements. Last year, spending across all labs was $150bn. Keeping the 3x trend means that, to keep pace with current advances, costs should rise to $450bn in 2025, $900bn in 2026, $2.7tn in 2027, $8.1tn in 2028, $25tn in 2028, and $75tn in 2029 and $225tn in 2030. For reference, the GDP of the world is around $125tn.
I think the labs will be crushed by the exponent on their costs faster white-collar work will be crushed by the 5% improvement exponent.
Be careful you're not confusing the costs of training an LLM and the spending from each firm. Much of that spending is on expanding access to older LLMs, building new infrastructure, and other costs.
"Model collapse" is a popular idea among the people who know nothing about AI, but it doesn't seem to be happening in real world. Dataset quality estimation shows no data quality drop over time, despite the estimates of "AI contamination" trickling up over time. Some data quality estimates show weak inverse effects (dataset quality is rising over time a little?), which is a mindfuck.
The performance of frontier AI systems also keeps improving, which is entirely expected. So does price-performance. One of the most "automation-relevant" performance metrics is "ability to complete long tasks", and that shows vaguely exponential growth.
Given the number of academic papers about it, model collapse is a popular idea among the people who know a lot about AI as well.
Model collapse is something demonstrated when models are recursively trained largely or entirely on their own output. Given most training data is still generated or edited by humans or synthetic, I'm not entirely certain why one would expect to see evidence of model collapse happening right now, but to dismiss it as something that can't happen in the real world seems a bit premature.
We've found in what conditions does model collapse happen slower or fails to happen altogether. Basically all of them are met in real world datasets. I do not expect that to change.
In 2025 you can add quality to jpegs. Your phone does it and you don't even notice. So the rhetorical metaphor employed holds up, in that AI is rapidly changing the fundamentals of how technology functions beyond our capacity to anticipate or keep up with it.
This is an especially bad example, a nice shiny grille is going to be strongly reflecting stuff that isn't already part of the image (and likely isn't covered well by adjacent pixels due the angle doubling of reflection).
Sure, you can view an LLM as a lossy compression of its dataset. But people who make the comparison are either trying to imply a fundamental deficiency, a performance ceiling, or trying to link it to information theory. And frankly, I don't see a lot of those "hardcore information theory in application to modern ML" discussions around.
The "fundamental deficiency/performance ceiling" argument I don't buy at all.
We already know that LLMs use high level abstractions to process data - very much unlike traditional compression algorithms. And we already know how to use tricks like RL to teach a model tricks that its dataset doesn't - which is where an awful lot of recent performance improvements is coming from.
And if you get that "sometimes" down to "rarely" and then "very rarely" you can replace a lot of expensive and inflexible humans with cheap and infinitely flexible computers.
That's pretty much what we're experiencing currently. Two years ago code generation by LLMs was usually horrible. Now it's generally pretty good.
I think humans who think they can't be replaced by a next token predictor think too highly of themselves.
LLMs show it plain and clear: there's no magic in human intelligence. Abstract thinking is nothing but fancy computation. It can be implemented in math and executed on a GPU.
"What's actually happening" is all your life you've been told that human intelligence is magical and special and unique. And now it turns out that it isn't. Cue the coping.
We've already found that LLMs implement the very same type of abstract thinking as humans do. Even with mechanistic interpretability being in the gutters, you can probe LLMs and find some of the concepts they think in.
But, of course, denying that is much less uncomfortable than the alternative. Another one falls victim to AI effect.
> "What's actually happening" is all your life you've been told that human intelligence is magical and special and unique. And now it turns out that it isn't. Cue the coping.
People have been arguing this is not the case for at least hundreds of years.
> These models are extremely unreliable when unsupervised.
> It doesn't feel like that will change fundamentally with just incrementally better training.
I could list several things that I thought wouldn't get better with more training and then got better with more training. I don't have any hope left that LLMs will hit a wall soon.
Also, LLMs don't need to be better programmers than you are, they only need to be good enough.
No matter how much better they get, I don't see any actual sign of intelligence, do you?
There is a lot of handwaving around the definition of intelligence in this context, of course.
My definition would be actual on the job learning and reliability i don't need to second guess every time.
I might be wrong, but those 2 requirements seem not compatible with current approach/hardware limitations.
Intelligence doesn't matter. To quote "Superintelligence: Paths, Dangers, Strategies":
> There is an important sense, however, in which chess-playing AI turned out to be a lesser triumph than many imagined it would be. It was once supposed, perhaps not unreasonably, that in order for a computer to play chess at grandmaster level, it would have to be endowed with a high degree of general intelligence.
The same thing might happen with LLMs and software engineering: LLMs will not be considered "intelligent" and software engineering will no longer be thought of as something requiring "actual intelligence".
Yes, current models can't replace software engineers. But they are getting better at it with every release. And they don't need to be as good as actual software engineers to replace them.
Chess was once thought to require general intelligence. Then computing power became cheap enough that using raw compute made computers better than humans. Computers didn't play chess in a very human-like way and there were a few years where you could still beat a computer by playing to its weaknesses. Now you'll never beat a computer at chess ever again.
Similarly, many software engineers think that writing software requires general intelligence. Then computing power became cheap enough that training LLMs became possible. Sure, LLMs don't think in a very human-like way: There are some tasks that are trivial for humans and where LLMs struggle but LLMs also outcompete your average software engineer in many other tasks. It's still possible to win against an LLM in an intelligence-off by playing to its weaknesses.
It doesn't matter that computers don't have general intelligence when they use raw compute to crush you in chess. And it won't matter that computers don't have general intelligence when they use raw compute to crush you at programming.
The proof that software development requires general intelligence is on you. I think the stuff most software engineers do daily doesn't. And I think LLMs will get continously better at it.
I certainly don't feel comfortable betting my professional future on software development for the coming decades.
"It is difficult to get a man to understand something when his salary depends upon his not understanding it" ~ Upton Sinclair
Your stance was the widely held stance not just on hacker news but also by the leading proponents of ai when chatgpt was first launched. A lot of people thought the hallucination aspect is something that simply can't be overcome. That LLMs were nothing but glorified stochastic parrots.
Well, things have changed quite dramatically lately. AI could plateau. But the pace at which it is improving is pretty scary.
Regardless of real "intelligence" or not.. the current reality is that AI can already do quite a lot of traditional software work. This wasn't even remotely true if if you were to go 6 months back.
I think I have a pretty good idea of what AI can do for software engineering, because I use it for that nearly every day and I experiment with different models and IDEs.
The way that has worked for me is to make prompts very specific, to the point where the prompt itself would not be comprehensible to someone who's not in the field.
If you sat a rando with no CS background in front of Cursor, Windsurf or Claude code, what do you suppose would happen?
It seems really doubtful to me that overcoming that gap is "just more training", because it would require a qualitatively different sort of product.
And even if we came to a point where no technical knowledge of how software actually works was required, you would still need to be precise about the business logic in natural language. Now you're writing computer code in natural language that will read like legalese. At that point you've just invented a new programming language.
Now maybe you're thinking, I'll just prompt it with all my email, all my docs, everything I have for context and just ask it to please make my boss happy.
But the level of integrative intelligence, combined with specialized world knowledge required for that task is really very far away from what current models can do.
The most powerful way that I've found to conceptualize what LLMs do is that they execute routines from huge learnt banks of programs that re-combine stored textual information along common patterns.
They're cut and paste engines where the recombination rules are potentially quite complex programs learnt from data.
This view fits well with the strengths and weaknesses of LLMs - they are good at combining two well understood solutions into something new, even if vaguely described.
But they are quite bad at abstracting textual information into a more fundamental model of program and world state and reasoning at that level.
I strongly suspect this is intrinsic to their training, because doing this is simply not required to complete the vast majority of text that could realistically have ended up in training databases.
Executing a sophisticated cut&paste scheme is in some ways just too effective; the technical challenge is how do you pose a training problem to force a model to learn beyond that.
I'd argue that, out of white collar jobs, it is actually one of the least automatable still. I.e. the rest of the jobs are likely going to get disrupted much faster because they are easier to automate (and have been the target of automation by the software industry in the past century). Whatever numbers were seeing now may be too early to reflect this accurately.
Also there are different metrics that are relevant like dollar count vs pure headcount. Cost cutting targets dollars. E.g. entry level developers are still expensive compared to other jobs.
Most "Software Engineering" is just applying the same code in slightly different contexts. If we were all smarter it would have been automated earlier through the use of some higher-level language.
You don’t need to perfect AI to the point of becoming credentialed professionals to gut job markets— it’s not just developers, or creative markets. Nobody’s worried that the world won’t have, say, lawyers anymore — they’re worried that AI will let 20% of the legal workforce do 100% of the requisite work, making the skill essentially worthless for the next few decades because we’d have way too many lawyers. Since the work AI does is largely entry-level work, that means almost nobody will be able to get a foothold in the business. Wash, rinse, repeat to varying levels across many white collar professions and you’ve got some real bad times brewing for people trying to enter the white collar workforce from now on— all without there being a single AI lawyer in the world.
We might end up needing 20% or so less doctors, because all that bureaucracy can be automated. A simple automated form pre-filler can save a lot of time. It’s likely that hospitals will try saving there.
I'm not necessarily opposed to unionization in general but it's never going to save many US software industry jobs. If a unionization drive succeeds at some big tech company then the workers might do well for a few years. But inevitably a non-union startup competitor with a lower cost structure and more flexible work rules will come along and eat their lunch. Then all the union workers will get laid off anyway.
Unionization kind of worked for mines and factories because the company was tied to a physical plant that couldn't easily be moved. But software can move around the world in milliseconds.
Unions _can_ protect against this, but they have to do it via lobbying the government for protectionism, tariffs, restricting non-union competition etc.
Indeed, just look at the CGI VFX industry of Hollywood. US invented it and was the leader for a long time, but now it has been commodified, standardized and run into the ground, because union or not, you can't stop US studios form offshoring the digital asset work to another country where labor is 80% cheaper than California and quality is 80% there. So the US is left with making the SW tools that VFX artist use, as the cutting edge graphics & GPU knowhow is all clustered there.
Similarly, a lot of non-cutting edge SW jobs will also leave the US as tooling becomes more standardized, and other nations upskill themselves to deliver similar value at less cost in exchange for USD.
This was when programmers were making software to time Amazon worker's bathroom breaks so believing "this could never happen to me" was probably an important psychological crutch.
This is, if true, a fundamental shift in the value of labor. There really isn’t a non-Luddite way to save these jobs without destroying American tech’s productivity.
That said, I’m still sceptical it isn’t simply a reflection of an overproduction of engineers and a broader economic slowdown.
Not really. If it’s overproduction, the solution is tighter standards at universities (and students exercising more discretion around which programmes they enroll in). If it’s overproduction and/or outsourcing, the solutions include labour organisation and, under this administration, immigration curbs and possibly services tariffs.
Either way, if it’s not AI the trend isn’t secular—it should eventually revert. This isn’t a story of junior coding roles being fucked, but one of an unlucky (and possibly poorly planning and misinformed) cohort.
It can be oversupply/outsourcing and also secular: You can have basically chronic oversupply due to a declining/maturing industry. Chronic oversupply because the number of engineers needed goes down every year and the pipeline isn't calibrated for that (academia has been dealing with this for a very long time now, look up the postdocalypse). Outsourcing, because as projects mature and new stuff doesn't come along to replace, running maintenance offshore gets easier.
Software isn't eating the world. Software ate the world. New use cases have basically not worked out (metaverse!) or are actively harmful.
Unions won’t solve this for you. If a company just decides they have enough automation to reduce union workforce it can happen the next time contracts get negotiated.
Either way, there are layoff provisions with union agreements.
Isn't that just delaying the inevitable? Yangshan Deep-Water Port in Shanghai is one of the most automated ports. Considering there are more people in China than in the US, China still automated their port.
I'm not making a value judgment on the specific case of dock workers, I'm rather saying that unions can and do prevent automation. If Software Devs had unionized earlier, a lot of positions would probably still be around.
In Hollywood, union bargaining bought some time at least. Unions did mandate limits on the use of AI for a lot of the creation process.
AI is still used in Hollywood but nobody is proud of it. No movie director goes around quoting percentages of how many scenes were augmented by AI or how many lines in the script were written by ChatGPT.
So what your argument is we're so special that we deserve to hold back human progress to have a privileged life? If it's not that what would you want a union to do in this situation?
Do you really believe that all technological progress has bettered humanity? Where’s the four day work week we were promised? I thought automation was supposed to free us from labor.
Unions work in physical domains that need labor “here and now”, think plumbers, electricians, and the like. You can’t send that labor overseas, and the union can control attempts at subversion via labor force importation. But even that has limitations, e.g. union factory workers simply having their factory shipped overseas.
Software development at its core can be done anywhere, anytime. Unionization would crank the offshoring that already happens into overdrive.
Unions can only prevent automation up to a point. Really the only thing that could have reasonably prevented this would have been for programmers to not produce as much freely accessible training data (formerly known as "open source software").
Exactly. I am always so impressed by the fact that developers never see that open source is essentially them giving away free labor to giant corporations. Developers basically programmed their way out of a job, for free. It's the only profession that is proud to have its best work done on unpaid time and used for free by big corporations.
We're not "too important." All a union would do is create extra problems for us.
There are two possibilities:
a) This is a large scale administrative coordination problem
b) We don't need as many software engineers.
Under (a) unionizing just adds more administrators and exacerbates the problem, under (b) unions are ineffective and just shaft new grads or if they manage to be effective, kills your employer (and then no one has a job.)
You can't just administrate away reality. The reason SWEs don't have unions is because most of us (unlike blue collar labor) are intelligent enough to understand this. I think additionally there was something to be said about factory work where the workers really were fungible and it was capital intensive, software development is almost the polar opposite where there's no capital and the value is the theory the programmers have in their head making them a lot less fungible.
Finally we do have legal tools like the GPL which do actually give us a lot of negotiating power. If you work on GPL software you can actually just tell your employer "behave or we'll take our ball and leave" if they do something stupid.
Unions would just delay the inevitable while causing other downsides like compressing salary bands, make it difficult to fire non-performers, union fees, increasing chance of corruption etc.
For a recent example:
> Volkswagen has an agreement with German unions, IG Metall, to implement over 35,000 job cuts in Germany by 2030 in a "socially responsible" way, following marathon talks in December 2024 that avoided immediate plant closures and compulsory layoffs, according to CNBC. The deal was a "Christmas miracle" after 70 hours of negotiations, aiming to save the company billions by reducing capacity and foregoing future wage increases, according to MSN and www.volkswagen-group.com.
I mean, I still don't want to unionize with the guys who find `git` too complicated to use (which is apparently the majority of HN). Also, you guys all hate immigrants which is not my vibe, sorry.
I really hope nobody had themselves convinced that software engineering couldn't be automated. Not with the code enterprise has been writing for decades now (lots and lots and lots of rules for gluing state to state, which are extremely structured but always just shy of being so structured that they were amenable to traditional finite-rule-based automation).
The goal of the industry has always been self-replacement. If you can't automate at least part of what you're working on you can't grow.
... unfortunately, as with many things, this meshes badly with capitalism when the question of "how do you justify your existence to society" comes up. Hypothetically, automating software engineering could lead to the largest open-source explosion in the history of the practice by freeing up software engineers to do something else instead of toil in the database mines... But in practice, we'll probably have to get barista jobs to make ends meet instead.
The experiences people are having when working with big, complex codebases don’t line up with your gloomy outlook. LLMs just fall apart beyond a certain project size, and then the tech debt must be paid.
Is it gloomy? I personally liken it to inventing the washing machine instead of doing laundry by hand, beating it against a washboard, for another hundred years.
I think the initial job loss from AI will come from having individual workers be more productive and eliminate the need to have larger teams to get the same work done.
Eventually, maybe. Right now I see a lot more people wasting time with AI in search of these promised efficiencies. A lot of companies reducing headcount are simply hiding the fact that they are deprioritizing projects or reducing their overall scope because the economy is shit (I know, I know - but it feels worse than reported IMO) and that's the right business cycle thing to do. If you're dramatic and take the DOGE/MAGA approach to management, just fire everyone and the important issues will become obvious where investment is actually needed. It's a headcount 'zero based budget' played out IRL. The truth is, there is a lot of fat to be cut from most large companies and I feel like it's the current business trend to be ruthless with the blade, especially since you have AI as a rose colored scapegoat.
The way I like to describe it is that you can't go from 1 developer to 0 thanks to AI, but you might be able to go from 10 to 9. Although not sure what the exact numbers are.
I'll go further than you. Even if the team is a cost center, it may not make sense to reduce the headcount if there's still more work to do. After all, an internal team that just assists other teams in the company without directly creating value suddenly become more productive could in turn make the other teams more productive. Automatically reducing headcount after a productivity increase is like that effect where people drive more dangerously when wearing seatbelts.
I used to hire someone who worked part time from home to bookmark some of the key pages in thousands of pdfs just so that I can directly jump to those pages instead of spending time myself on finding those pages.
AI can now do it very cheap so no need to give that job to a human anymore.
It presents a difference-in-differences (https://en.wikipedia.org/wiki/Difference_in_differences) design that exploits staggered adoption of generative AI to estimate the causal effect on productivity. It compares headcount over time by age group across several occupations, showing significant differentials across age groups.
Page 3: "We test for a class of such confounders by controlling for firm-time effects in an event study regression, absorbing aggregate firm shocks that impact all workers at a firm regardless of AI exposure. For workers aged 22-25, we find a 12 log-point decline in relative employment for the most AI-exposed quintiles compared to the least exposed quintile, a large and statistically significant effect."
I appreciate the link to differences in differences, I didn't know what to call this method.
The OP's point could still be valid: it’s still possible that macro factors like inflation, interest rates, or tariffs land harder on the exact group they label ‘AI-exposed.’ That makes the attribution messy.
You really do have to account for why this is mainly happening in industries that are adopting AI, why it's almost exclusively impacting entry-level positions (with senior positions steady or growing), and why controlling for broad economic conditions failed to correct this. I doubt very much that these three Stanford professors would be blindsided by the concept of rates and tarriffs.
My personal theory is that the stock market rewards the behavior of cutting jobs as a signal of the company being on the AI bandwagon. Doesn't matter if the roles were needed or not. Line goes up, therefore it is good.
This is a complete reversal in the past where having a high headcount was an easy signal of a company's growth (i.e. more people, means more people building features, means more growth).
Investors are lazy. They see one line go down, they make the other line go up.
CEOs are lazy. They see line go up when other line goes down. So they make other line go down.
(I am aware that "line go up" is a stupid meme. But I think it's a perfect way to describe what's happening. It is stupid, lazy, absurd, memetic. It's the only thing that matters, stripped off of anything that is incidental. Line must go up.)
If we replace all juniors with AI, in a few years there won't be skilled talent for senior positions.
AI assistance is a lot different than AI running the company. Making expensive decisions. While it could progress, bear in mind that some seniors continue to move up in the ranks. Will AI eventually be the CEO?
We all dislike how some CEOs behave, but will AI really value life at all? CEOs have to have some place to live, after all.
The AI will at least be cheaper than a CEO, it might also be more competent and more ethical. The argument against making a Large Language Model the CEO seems to mostly be about protecting the feelings of the existing CEO, maybe the Board should look past these "feelings" and be bold ?
Software development is one of the listed industries. Well before AI we have seen that few companies wanted entry level devs due to the training and such.
Reducing in call centers has been going on for a while as more people use automated solutions (not necessarily AI) and many of the growing companies make it hard to reach a real person anyways (Amazon, Facebook, etc). I feel like AI is throwing fuel on the existing fire, but isn't as much of a driver as the headlines suggest.
American workers are truly under attack from all sides. H1B. Outsourcing. What's left? The blue collar manufacturing is mostly gone. White collar work well on its way out. Why is our own government (by the people for the people) actively assisting in destroying American's ability to get jobs (H1B)? Especially in these conditions. I'm no racist or idiot but it's unacceptable. I didn't expect the gov to actively be conspiring with big corps to make my economic position weaker. Unbelievable breach of trust. We need to demand change from our government.
And when we're all out of work except for the doctors and nurses, electricians and plumbers, there will be nobody to contribute to consumer spending. And we will suffer, at the hand of the government that assisted in this scam.
It’s an unpopular opinion in the current environment but it’s the program that allows international talent to connect with local capital that creates all the jobs in tech.
The biggest problem right now is that there is no distinction between companies replacing Americans labor with cheap labor and entrepreneurial talent that creates jobs. Everyone is on the same visa.
Having to work with ESL contractors from firms like Cognizant or HCL is true pain. Normally it would be like 3-4 US employees working on something and then its like 20-30 ESL outsourced people working on something. The quality is so poor though its not worth it.
My current org nuked their contract w HCL after 2 years because how shitty they are and now everything is back onshore. Millions wasted lol. Corporations are so silly sometimes.
They also need 5 people to do the work of one us worker. And then another US worker to guide and do some qa on the output they produce . I don't see how it saves money. There are other countries with lower wages than the US where this doesn't happen such as Poland or Australia.
> You really do have to account for why this is mainly happening in industries that are adopting AI
Correlation is not causation. The original research paper does not prove a connection.
> I doubt very much that these three Stanford professors would be blindsided by the concept of rates and tarriffs.
They are nonetheless subject to publish or perish pressure and have strong incentives to draw publishable attention-grabbing results even where the data is inconclusive.
Tariffs are just a massive government revenue generating consumption tax on particular industries. We would expect unemployment among the young trying to enter those industries to be hit hardest.
Is there some central authority that’s telling people to blame this all on AI, or how is everyone reaching this conclusion and ignoring the other obvious factors you stated?
It is in their interest to find explanations for reductions in labor that don't assign the blame to corporate greed.
For example, a call center might use the excuse of AI to fire a bunch of people. They would have liked to just arbitrarily fire people a few years ago, but if they did that people would notice the reduction in quality and perhaps realize it was done out of self-serving greed (executives get bigger bonuses / look better, etc). The AI excuse means that their service might be worse, perhaps inexcusably so, but no one is going to scrutinize it that closely because there is a palatable justification for why it was done.
This is certainly the type of effect I feel like underlies every story of AI firing I've heard about.
How is firing a bunch of people because you made a machine that you believe can do their jobs not textbook corporate greed? It seems like the worst impulses of Taylorism made manifest?
Since this article is about AI, and since this comment seems rather low effort compared to the Stanford study, I went ahead and used low effort to analyze the report compare it to this comment. Here's my low effort AI response:
> Prompt: Attached is a paper. Below is an argument made against it. Is there anything in the paper that addresses the argument?: High interest rates + tariff terror -> less investment -> less jobs
> High rates/firm shocks: They add firm–time fixed effects that absorb broad firm shocks (like interest-rate changes), and the within-firm drop for 22–25-year-olds in AI-exposed roles remains.
> “Less investment” story: They note the 2022 §174 R&D amortization change and show the pattern persists even after excluding computer occupations and information-sector firms.
> Other non-AI explanations: The decline shows up in both teleworkable and non-teleworkable jobs and isn’t explained by pandemic-era education issues.
> Tariffs: Tariffs aren’t analyzed directly; broad tariff impacts would be soaked up by the firm–time controls, but a tariff-specific, task-level channel isn’t separately tested.
Fitting, since it came up with unrelated information (the R&D tax thing) and the 3rd bullet point. Also started talking about tariffs as if it had addressed them, then notes that it doesn't address them.
Blaming AI is better because it helps corporations convince the working class that there jobs are in long-term danger so they collectively settle for less favorable work terms and compensation, unlike if they are convinced that it is going to gradually improve with the upcoming monetary easing cycle..
End of ZIRP and the Sec. 179 change for engineering salaries probably explains more of this (plus the increase in outsourcing). I’m sure some decision makers also threw AI into the mix but the financials of hiring software engineers in the US was already challenging before AI “took everyone’s job”.
I generally agree that AI is the scapegoat, but not for those same reasons. Despite the lack of job growth and the tariffs, recent data shows the economy grew about 3%. Even if it's not AI as the primary driver, efficiency seems to have increased.
How does that make sense? Wouldn’t high interest rates and tariffs cause more expensive engineers to have disproportionate opportunity? I remember during 2008 it was much easier for my employer to justify junior engineers than senior ones.
CEOs citing savings from AI should be able to show higher profits soon. The fact that they’re not means those tall tales are coming home to roost soon.
Every day when I am out in the city, I am amazed by how many jobs we have NOT managed to replace with AI yet.
For example, cashiers. There are still many people spending their lives dragging items over a scanner, reading a number from a screen, holding out their hand for the customer to put money in, and then sorting the coins into boxes.
Self checkout has been a thing for ages. Heck in Japan the 711s have cashiers but you put the money into a machine that counts and distributes change for them.
Supermarkets are actually getting rid of self checkouts due to crime. Surprise surprise, having less visible "supervision" in a store results in more shoplifting than having employees who won't stop it anyway.
It’s also just resulting in atrocious customer experience.
I can go to Safeway or the smaller chain half a block away.
The Safeway went all in on self checkouts. The store is barely staffed, shelves are constantly empty, you have to have your receipt checked by security every time, they closed the second entrance permanently, and for some reason the place smells.
Other store has self checkouts but they also have loads of staff. I usually go through the normal checkout because it’s easier and since they have adequate staff and self checkout lines it tends to be about the same speed to.
End result is I don’t shop at Safeway if I can avoid it.
Cashiers should not, and will not prevent theft. They're not paid nearly enough to get in danger, and it is not their job.
I'm sure you can find videos of thefts in San Francisco if you need a visual demonstration. No cashier is going to jump in front of someone to stop a theft.
True, but having a cashier standing there waiting to scan your items will prevent most normal people from stealing. Sure, some will brazenly walk right past with a TV on their shoulder, but most people won't.
If there's no cashier and you're doing it yourself, a whole lot more people will "forget" to scan a couple items, and that adds up.
There's usually a security person or two in the store, looking over the self checkouts. I agree that job prevents a lot of people from becoming opportunistic thiefs, but I'm making a distinction between cashiers and security. Today the store needs both.
> The fact that a cashier does not prevent all thefts does not mean a cashier does NOTHING for theft.
Yes, for one thing, it ignores that a very large share of retail theft is insider theft, and that cash handling positions are the largest portion of that.
They absolutely do. It’s not the cashiers being security, it’s having adequate staffing making people less likely to steal. Its not stopping crimes that have occurred it’s just reducing opportunistic theft.
You ban them from coming back in after a few warnings. Stores seem really icy about facial recognition right now though. The optics are pretty bad (a play on words pun?)
Detecting theft does not mean theft is prevented. You then need the government to prosecute, and impose sufficient punishment to deter theft. This is not cheap, nor a given that it will happen.
So take the broken god awful experience of self checkout and add another layer of “I think you did something wrong so now you have to stand around waiting for an actual person”?
You mean ordering kiosks and self-checkout machines? We have automated it, it's just not everywhere has implemented it.
The one I'm desperately waiting for is serverless restaurants—food halls already do it but I want it everywhere. Just let me sit down, put an order into the kitchen, pick it up myself. I promise I can walk 20 feet and fill my own drink cup.
Japan does this a lot of places, and it makes the experience much easier.
And I think the entire mid and low range restaurants could replace servers with a tablet and people would be happier. I'm not sure how it doesn't make more money for the restaurant too, making it so easy to order more during a meal.
You seem to like self-checkout processes. I don't. I avoid any place where I have to interact with a screen. Be it a screen installed on-premise or the screen on my phone. It is not a relaxing experience for me.
Serverless restaurants have been common in Australia for decades. You just get a buzzer and then need to go pick up your food when it is ready. There's a single person behind the bar to take orders and pour beer/wine/soda.
Seems like perfect option for robots (not humanoid). Bring me my food. You can still keep people in kitchen for a bit, but well servers in many restaurants are not really needed.
There used to be two humans standing at the cash register,
now because of software, automatic change machines, and cameras there is only one. One of those humans' jobs got automated.
Call it what you like but replacing the work of humans one for one is difficult and usually not necessary. Reformulating the problem to one that machines can solve is basically the whole game. You don't need a robot front desk worker to greet you, you just need a tablet to do your check in.
This. And I do their work a lot more slowly because it's not my regular job, and I actually already had to do some of the work (getting the items out of my trolley and onto the conveyor). Now I stand there forever fumbling with barcodes, trying to get bags to stay open, switching between getting items out of the trolley and scanning. The old checkout system is so much more efficient when you are buying anything more than a couple of items at a time.
Yeah this is like saying Aldi “automated” cart return. They didn’t, they got every shopper to do the work themselves. Automated cart return would be if you just gave the cart a little “giddyup!” when you were done and it found its way home. Or those cart conveyor belts at Ikea, it’s only part of the process but that part is automated.
[edit] Aldi did automate the management of getting shoppers to do that work, because there’s not a person standing there taking and handing out quarters, but (very simple) machines. Without those machines they might need a person, so that hypothetical role (the existence of which might make the whole scheme uneconomical) is automated. But they didn’t automate cart return, all that work’s still being done by people.
Pharmacists are my favourite. They're a human vending machine that is bad at counting and reading. But law protects them. Pretty good regulatory capture.
Please actually understand what pharmacists actually do and _why_ AI is not a good replacement for them yet, unless you want to die of certain drugs interactions.
Hahaha, this drug interaction nonsense is what online people tell each other. It isn't even real. It's like "nice trigger discipline" or "the postal police don't fuck around" and shit like that. Just something that is not true but for some reason is internet urban legend.
Retail pharmacists are human vending machines. You don't need AI. It's a computer prescription written by a far more qualified human which is then provided to a nigh-illiterate half-wit who will then try as hard as possible to misread it. Having then misread it, the patient must then coax them out of their idiocy until they apologize and fulfill what's written.
Meanwhile some Internet guy who gets all his information from the Internet will repeat what he's heard on the Internet. I know this because anyone passingly acquainted with this would have at least made the clarification between compounding pharmacists and retail pharmacists or something.
Pharmacists are a fantastic example. My pharmacy is delivered my prescription by computer. They text me, by computer, when it's ready to pick up. I drive over there … and it isn't ready, and I have to loiter for 15 minutes.
Also, after the prescription ends, they're still filling it. I just never pick it up. The autonomous flow has no ability to handle this situation, so now I get a monthly text that my prescription is ready. The actual support line is literally unmanned, and messages given it are piped to /dev/null.
The existing automation is hot garbage. But C-suite would have me believe our Lord & Savior, AI, will fix it all.
The only way AI could fix this if it said "replace the pharmacist with a vending machine and hire a $150k junior engineer to make sure the DB is updated afterwards", which you never know, Claude Opus 4 might suggest. At that point, we'll know AGI has been achieved.
> Some examples of these highly exposed jobs include customer service representatives, accountants and software developers.
We seem to be in this illogical (delusional?) era where we are being told that AI is 'replacing' people in certain sectors or types of work (under the guise that AI is better or will soon be better than humans in these roles) yet those same areas seem to be getting worse?
- Customer service seems worse than ever as humans are replaced with "AI" that doesn't actually help customers more than 'website chatbots' did 20 years ago.
- Accounting was a field that was desperate for qualified humans before AI. My attempts to use AI for pretty much anything accounting related has had abysmal results.
- The general consensus around software development seems to be that while AI is lowering the barrier of entry to "producing code", the rate of production of tech debt and code that no one "owns" (understands) has exploded with yet-to-be-seen consequences.
> "The general consensus around software development seems to be that while AI is lowering the barrier of entry to "producing code", the rate of production of tech debt and code that no one "owns" (understands) has exploded with yet-to-be-seen consequences."
^ This. (Tho I'm not sure about it being "general
consensus".) Vibe code is the payday loan (or high-interest credit card) of tech debt. Demo-quality code has a way of making it into production. Now "everyone" can produce demos and PoCs. Companies that leverage AI as a powerful tool in the hands of experienced engineers may be able to iterate faster and increase quality, but I expect a sad majority to learn the hard way that there's no free lunch, and shipping something you don't understand is a recipe for disaster.
I’m suss about this paper when it makes this claim:
“where AI is more likely to
automate, rather than augment , human labor.”
Where is AI currently automating human labor? Not Software Engineering. Or - what’s the difference between AI that augments me so I can do the job of three people and AI that “automates human labor”
I was also curious about this. Table A1 on page 56 lists examples of positions that are automated vs augmented, and these are the positions the authors think are going to be most augmented (allegedly taken from [0]):
- Chief Executives
- Maintenance and Repair Workers, General
- Registered Nurses
- Computer and Information Systems Managers
After skimming [0], I can't seem to find a listing of jobs that would be augmented vs automated, just a breakdown of the % of analyzed queries that were augmenting vs automating, so I'm a bit confused where this is coming from.
When the Stanford paper looked at augment vs automate, they used the data from Anthropic's AI Economic Index. That paper defined the terms like this:
We also analyze how AI is being used for tasks, finding 57% of usage suggests augmentation of human capabilities (e.g., learning or iterating on an output) while 43% suggests automation (e.g., fulfilling a request with minimal human involvement).
From the data, software engineers are automating their own work, not augmenting. Anthropic's full paper is here:
I see a worrisome trend. On one hand, many of my proto-boomer friends are suffering from age-ism , and memes claim that over-50-year-olds are unemployable. Not 100% fidelity, but there's some truth.
Then I hear about a lot of youngsters struggling to find work, and see articles like this.
Well, who's left? Is there a sweet spot at like 31 that are just cleaning up?
beside the point, but over 50 = proto-boomer? You mean para-boomer, maybe? Gen X is <=60, I believe, so you referring to the cusp boomer/genx I think..
It emphasizes "AI adoption linked to 13% decline," which implies causation. The study itself only claims "evidence consistent with the hypothesis."
The article also largely highlights job loss for young workers, while only briefly mentioning cases where AI complements workers.
The study's preliminary status -- it is not peer reviewed -- is noted but only once and at end. If the article was more balanced it would have noted this at the beginning.
Articles on the same subject by the World Economic Forum, McKinsey, and Goldman Sachs are more balance and less alarmist.
Thinly veiled economic propaganda aside, I am dealing with a different AI mess everyday. Technical debt is exploding everywhere I turn. There is an ever larger part of me these days that wishes I could just call the bluff all at once and let all the companies in question learn the inevitable lessons here the hard way.
The worst thing for me would be just needing to get a job like I had before being a dev, the stakes are so much grander for all the companies. It's only really existential for the side of this that isn't me/us. I've been working since I was 15, I can figure it out. I'll be more happy cutting veggies in a kitchen than every single CEO out there when all is said and done!
My company did exactly this earlier in the year. It was a blatant lie and everyone who works here knew it. None of the people laid off were actually replaced with AI, the work they did was just eliminated.
It’s also just natural cost cutting from businesses that were previously massively over hiring, and outside of AI don’t exactly have a ton of areas with huge growing investment.
Plus slashing jobs like this keeps the plebs in line. They don’t like software engineers having the money and job security to raise a stink over things. They want drones terrified of losing everything.
Is it pure AI or a guy in India that can cover the gap using AI to create good enough slop to pass?
Everyone is doubling down on hiring IN India right now. H1B isn’t even a thing. It’s offshoring to Indians that are utilizing AI to ship good enough slop. Everyone’s India office is rapidly expanding.
Slow down people. Let's stop jumping to biases and see what we have here.
Note upfront: I'm not suggesting AI is not having an impact. That would be foolish. But I will say there's *a lot* less to the conclusion of this study, simply because the data is questionable. It's not that they did anything wrong per se. I won't say that here because it'll end up a HN cluster fuck. Cluster fuck aside, the caveats and associated doubt are enough to say, "Don't bet the farm on this study." Great bander for the bar? Sure.
It's an interesting study but I've seen it called "absolute proof" and other type things. Don't be fooled, it's not that.
> "This study uses data from ADP, the largest payroll processing firm in America. The company provides payroll services for firms employing over 25 million workers in the US. We use this information to track employment changes for workers in occupations measured as more or less exposed to artificial intelligence"
a) I'm calling this out because I've seen posts on LinkedIn saying it was a sample of 25M. Nope! ADP simply does payroll for that many.
b) The size of the US workforce is ~165M, making ADP's coverage ~15% of the workforce.
c) Do the business ADP server come from particular industries, are of a particular size, in particular geographic locations? etc.? It's not only about the size of the sample - which we'll get to shortly - but the nature of the companies - which we'll also get to shortly.
> "We make several sample restrictions for our main analysis sample."
d) It's great that they say this, but it should raise an eyebrow.
> "We include only workers employed by firms that use ADP’s payroll product to maintain worker earnings records. We also exclude employees classified by firms as part-time from the analysis and subset to people between the age of 18 and 70."
e) Translation: we did a slight bit of pruning (read: cherry-picking).
> "The set of firms using payroll services changes over time as companies join or leave ADP’s platform. We maintain a consistent set of firms across our main sample period by keeping only companies that have employee earnings records for each month from January 2021 through July 2025."
f) Translation: More cherry-picking.
> "In addition, ADP observes job titles for about 70% of workers in its system. We exclude workers who do not have a recorded job title."
g) Translation: More cherry-picking.
> "After these restrictions we have records on between 3.5 and 5 million workers each month for our main analysis sample, though we consider robustness to alternative analyses such as allowing for firms to enter and leave the sample."
h) 3.5M to 5.0M feels like a large enough sample... if it wasn't so "restricted." Furthermore, there's no explanation on the 1.5M delta, and how adding or removing that much impacts the analysis.
i) And they considered that why? And did what they did why? It's a significant assumpt that gets nothing more than a hand wave?
> "While the ADP data include millions of workers in each month, the distribution of firms using ADP services does not exactly match the distribution of firms across the broader US economy."
j) Translation: as mentioned above ADP !== a representation of the broader economy.
> "Further details on differences in firm composition can be found in Cajner et al. (2018) and ADP Reserch (2025)."
j) Great there's a citation, but given the acknowledgement of the delta isn't at least a line or two in order? Something about the nature of the delta, and THEN mention the citation?
k) Editorial: You might think this hand-wave is ok, but to me it's usually indicative of a tell and a smell.
l) Finally, do understand the nature of academia and null research (which has been mentioned on HN). In short, there is a (career / financial) incentive to find something novel (read: worth publishing). You advance your career by doing not-null
research.
Again, I'm not suggesting anything nefarious per se. But this study is getting A LOT of attention. All things considered, more than it objectively deserves.
__Again: I'm not suggesting AI is not having an impact. That would be foolish.__
This study feels pretty weak. Software as a occupation is collapsing, but it's not due to AI. Articles and "studies" like this are just a smoke screen to keep your eye off the ball.
As I see it, it's really the lack of "capitalists" willpower to be actually capitalist.
We can't call it incompetence because neither those whom we have come to know as capitalists nor their advisors are incompetent, which means they quite literally do not want to offset any decline in jobs or (job creation) that can be linked to progress.
That's not strange. A "capitalist" wants market participation to grow, infinitely, which is possible. Who we came to know as capitalists don't care about the markets, actual market growth or market participation. They only care about the growth of the value of the markets, "however" that happens.
I highly recommend that journalists and economists dig a bit more radically honest into the matter. There'd be more value in that, more blog posts, more articles, more discussions on all platforms, and thus more participation.
I mean it's a scapegoat vs straw man vs actual culprit kind of situation ... isn't it?
Short-term, discrete numbers like these are interesting to look at, but they don't really tell us much about the long-term trajectory. In parallel: [1].
The accounting note is not true in the traditional sense. The field in the US is just getting offshored to India/PH/Eastern Europe for better or for worse. There is even a big push to lower the educational requirements to attain licensure in the US (Big 4 partners want more bodies and are destroying the pipeline for US students). Audit quality will continue to suffer and public filers will issue bunk financials if they aren't properly attested to.
The reports from the usual "offshoring centers" aren't exactly inspiring. It's a bloodbath over there.
Seems like the capabilities of current systems map onto "the kind of labor that gets offshored" quite well. Some of the jobs that would get offloaded to India now get offloaded to Anthropic's datacenters instead.
And some jobs, offshored or not, are just human frontend to datacenters.
It's amusing to see programmers in the US promoting remote work.
Do those people really believe they're the most intellectually superior to the rest of the world? If a job can be done purely remotely, what stops the employer from hiring someone who lives in a cheaper place?
And being "superior" doesn't necessarily mean extraordinary coding skills. The vast majority of code to be written doesn't require that. What it requires however is a combination of common sense and good understanding of the underlying business. This is in short supply in many of the locations the jobs are being offshored to. But let's be honest, it was also on short supply in the corporate IT departments being offshored, though not quite to the same degree.
You’ll get downvoted but in my experience, which may not be representative of the entire population, this is true.
A mid-size US tech company I know well went fully remote after a lot of insistence from the workforce, prior to the pandemic they were fully in office.
Soon enough they started hiring remotely from EU, and now the vast majority of their technical folks are from there. The only US workers remaining are mostly GTM/sales. I personally heard the founder saying “why should we pay US comp when we can get extremely good talent in EU for less than half the cost”. EU workers, on average, also tend to not switch job as frequently, so that’s a further advantage for the company.
Once you adapt to remote-only, you can scoop some amazing talent in Poland/Ukraine/Serbia/etc for $50k a year.
I think most programmers in the US simply don't realize how much they earn compared to the rest of the world.
I'm not talking about rural Chinese villages whose name you can't pronounce. I'm talking about highly educated programmers who can communicate fluently in English, in cities like Beijing or Munich. If people in SV know how little their counterparts make in these places, they'd be much more opposed to remote work.
And that was before LLM. Today practically the entire planet can write passable English.
How many of these jobs are getting offshored because of AI?
Language barriers, culture, and knowledge are some of the biggest challenges to overcome for offshoring. AI potentially solves many of those challenges
> AI potentially solves many of those challenges
Isn't it exactly the opposite?
Language barriers: LLMs are language models and all of the major ones are built in English, speaking that language fluently is surely a prerequisite to interacting with them efficiently?
Knowledge: famously LLMs "know" nothing and are making things up all of the time and sometimes approximate "knowledge"
Nope, LLMs are quite functional in non-english languages. My partner regularly works with ChatGPT in Turkish
My experience: hosted LLMs are very good, but even 30B models you run locally are quite poor (at least in Romanian). To some degree they still hallucinate words (they don't conjugate properly sometimes).
Language barriers: The outsourced workers I know use AI to help them ask and answer questions about things in English they don’t perfectly understand because English is their second language. They use it to write better English from English with grammatical mistakes
Knowledge: True to an extent, but my assumption here is that it would be used to fill in gaps or correct misunderstandings. Not wholesale doing my job. At least that’s often how I use it
I worry things will be lost in translation (maybe would have already), Or the LLMs will fill in the gaps with wrong information, like some sort of weird telephone game.
That said, I have one ESL on my team who uses LLMs a lot like that and it's fine so who knows.
Found this article from last year saying IIT grads are facing the same grim outlook as technology hiring in India for new grads has also dried up
https://www.bloomberg.com/news/newsletters/2024-05-30/tough-...
So, that doesn't seem like a likely culprit unless you have some convincing evidence.
I think you are conflating 2 things. AI could be going after new entry level jobs in software engineering. I am not a professional engineer but an accountant by trade (I like writing software as a hobby lol) but this article looks like evidence that IIT grads will have a harder time getting these jobs that AI is attacking. My comment rests on the fact that the report doesn't really reconcile with AI destroying entry level jobs for accounting, but rather this type of work being offshored to APAC/India. There are still new COEs being built up for mid cap companies for shared services in India to this day and I don't mean Cognizant and Wipro, but rather the end customer being the company in question with really slick offices there.
I think the article doesn't really prove AI is the culprit but I think this other article disproves that offshoring is. If offshoring was the culprit why is it only affecting the most junior employees? I think the case is still open but AI is the leading candidate.
Do you have any evidence of this because the rationale seems like a coping strategy or conspiracy theory how it's being suppositioned.
Do you have any actual evidence that supports the headline? The article does not. It simply mentions 13% decline in relative employment and then blames AI with no actual evidence. Given what I know about the current state of AI and off-shoring, I think off-shoring is a million times more likely to be the culprit than AI.
The entire account department at my firm has moved to Poland. That’s nice for them, but as a US citizen it does mean the writing is on the wall. On the plus side I learned a fun fact. Malgorzata is a more common name than I had ever imagined.
IT help was outsourced to India years ago. I expect them to be replaced with AI the minute their government stops handing the firm big contracts because I’ve never spoken to anyone from that group who was actually better than a chat bot.
Have you seen how the profession has worked post SOX? Did you know 2016 was the peak year where you had accounting students enrolled in uni in the states? I want you to think laterally about this.
This is exactly right.
The H1B pipeline has not decreased at all whereas millions of American workers have been laid off.
Maybe for software engineering but not for accounting. I've had to interface with many offshored teams and interviewed at places where accounting ops were in COE centers in EU/APAC.
Yup, 95% of the AI hype is to apply pressure on the labor market and provide cover for offshoring/downsizing.
Where is the evidence for this? Who is "applying pressure on the labour market"?
Every executive publicly saying obviously* false things like X job will be done by AI in 18 months is putting downward pressure on the labor market. The pressure is essentially peer pressure among executives: are we stupid for continuing to hire engineers instead of handing our engineering budget to Anthropic?
* - Someone should maintain a walkback list to track these. I believe recent additions are Amodei of Anthropic and the CEOs of AWS and Salesforce. (Benioff of Salesforce, in February: "We're not going to hire any new engineers this year." Their careers page shows a pivot from that position.)
Maybe it's a good time to ask for advice. Which IT job roles and companies are least vulnerable to offshoring? Defense contractors and the like?
> Audit quality will continue to suffer
I wonder how much this actually matters? I understand that for an auditor, having a quality reputation matters. But if all audits from all firms are bad, how much would the world economy suffer?
Likewise for the legal profession, if all judges made twice the number of mistakes, how much would the world suffer?
> Likewise for the legal profession, if all judges made twice the number of mistakes, how much would the world suffer?
Is this hyperbole? It seems like the real question being asked here is "would the world be worse off without deterministic checks and balances", which I think most people would agree is true, no?
I read it as assuming the deterministic checks and balances are already absent. We have the illusion of determinism but, in practice, audits (and justice) are mostly theatre as it is.
From that perspective, lowering the quality of something that is already non-rigourous might not have any perceivable effect. It’s only a problem if public perception lowers, but that’s a marketing issue that the big-4 already have a handle on.
They don’t though. Marketing hits reality all the time. The Big 4 will survive, but you can only gaslight people for so long.
The all-in on AI shows a lack of imagination around innovation.
The current system is not long term stable, and poor accounting is part of the reason more people don't know that. Even worse accounting would speed up the decline.
Then you would have to think twice about the company you may be giving money to (ie the stock market and private bank loans). That's the whole objective of this. Every company is going to need an accountant in one way or another and you don't really need to follow strict GAAP for management requirements (what else is EBIDTA for if anything?), but it's something completely different than saying: I made x dollars and spent y dollars, here is what I have and what I owe, please give me money.
At the end of the day it is a question of convenience/standards, if GAAP didn't exist maybe firms could use a modified accrual standard that is wholly compliant with tax reporting and that's it.
This is all covert AI advertising.
https://esborogardius.substack.com/p/if-ai-doesnt-fire-you-i...
Brutal that software engineering went from one of the least automatable jobs to a job that is universally agreed to be "most exposed to automation".
Was good while it lasted though.
I'm not sure it's that our job is the most automatable, but that the interface is the easiest to adapt to our workflow.
I have a feeling language models will be good at virtually every "sit at a desk" job in a virtually identical capacity, it's just the act of plugging an AI into these roles is non-obvious.
Like every business was impacted by the Internet equally, the early applications were just an artifact of what was an easy business decision.. e.g. it was easier to start a dotcom than to migrate a traditional corporate process.
What we will see here with AI is not the immediate replacement of jobs, but the disruption of markets with offerings that human labor simply can't out-compete.
> I'm not sure it's that our job is the most automatable
I don't know. It seems pretty friendly to automation to me.
When was the last time you wrote assembly? When was the last time you had map memory? Think about blitting memory to a screen buffer to draw a square on a screen? Schedule processes and threads?
These are things that I routinely did as a junior engineer writing software a long time ago. Most people at that time did. For the most part, the computer does them all now. People still do them, but only when it really counts and applications are niche.
Think about how large code bases are now and how complicated software systems are. How many layers they have. Complexity on this scale was unthinkable not so long ago.
It's all possible because the computer manages much of the complexity through various forms of automation.
Expect more automation. Maybe LLMs are the vehicle that delivers it, maybe not. But more automation in software is the rule, not the exception.
RAD programming held the same promise, as did UML, flow/low/no code platforms.
Inevitably, people remember that the hard part of programming isn't so much the code as it is putting requirements into maintainable code that can respond to future requirements.
LLMs basically only automate the easiest part of the job today. Time will tell if they get better, but my money is on me fixing people's broken LLM generated businesses rather than being replaced by one.
Indeed. Capacity to do the hard parts of software engineering well may well be our best indicator of AGI.
I don't think LLMs alone are going to get there. They might be a key component in a more powerful system, but they might also be a very impressive dead end.
This has been my argument as well. We've been climbing the abstraction ladder for years. Assembly -> C -> OOP ->... this just seems like another layer of abstraction. "Programmers" are going to become "architects".
The labor cost of implementing a given feature is going to dramatically drop. Jevons Paradox paradox will hopefully still mean that the labor pool will just be used to create '10x' the output (or whatever the number actually is).
If the cost of a line of code / feature / app becomes basically '0', will we still hit a limit in terms of how much software can be consumed? Or do consumers have an infinite hunger for new software? It feels like the answer has to be 'it's finite'. We have a limited attention span of (say) 8hrs/person * 8 billion.
the cost of creating a line of code dropped to zero. the ongoing cost of having created a line of code has if anything gone up.
LLMs are just another layer of abstraction on top of countless. It’s not going to be the last layer, though.
The only thing that AI is good at is a job that someone has already done before.
If it gets to the point where I can no longer find a tech job I am just going to buy a trailer, live somewhere cheap, and just make money doing odd jobs while spending most of my time programming what I want. I don't want to participate in a society where all I have for job options is a McJob or some Amazon warehouse.
That's plan C, plan B is to one person SAAS a better app than my current company makes.
This is the best thing engineers can do. I moved to building as a solo founder. I am building an LLM enabled coding product and I teach. I'm hosting a session on Claude Code today, 134 guests signed up. I'm gradually planning to make money teaching for a few months while building the product.
That's actually a good idea. Now I just need to come up with an idea for an SAAS app. I was thinking originally or making one of the games on my project backlog and seeing how much I could make off it. Or creating one of the many idea I have for websites and webapps and see where they go.
Is it hard to date with a trailer?
Would be more difficult depending on where you live. My plan was to talk to others online and see if I could find someone willing to live such a simple life with me, maybe starting with an LDR first (I'm sort of doing that already)
Beginning to suspect this person is living in a trailer or cave and collecting info for their UniqueDating SaaS.
Not if it has a hitch.
>Buy a trailer, live somewhere cheap, do odd jobs
Unrelated to the discussion, but I love these kinds of backup plans. I've found that most guys I talk to have one. Just a few days ago a guy was telling me that, if his beloved wife ever divorces him, then he'd move to a tropical island and become a coconut seller.
(My personal plan: find a small town in the Sonoran Desert that has a good library, dig a hole under a nice big Saguaro cactus, then live out my days reading library books in my cool and shady cave.)
Mine is forrest fire fighter. Surely with climate change there will not be a shortage of work, and while dangerous and bad for you, it seems kind of fun.
The future seems very uncertain right now and we are living in weird times. Its always a good idea to have a backup plan in case your career path doesn't work out!
Is it hard to date living under a cactus?
Yes, that's where living under a date palm is better.
it must be easier than dating on top of a cactus
Nah dating under a cactus is easy: just don't be a prick.
> he'd move to a tropical island and become a coconut seller.
Is there a visa for that? Doesn't seem feasible unless he lives in a country that has a tropical island already.
Maybe it's just the nature of being early adopters.
Other fields will get their turn once a baseline of best practices is established that the consultants can sell training for.
In the meantime, memes aside, I'm not too worried about being completely automated away.
These models are extremely unreliable when unsupervised.
It doesn't feel like that will change fundamentally with just incrementally better training.
Does it have to? Stack enough "it's 5% better" on top of each other and the exponent will crush you.
AI training costs are increasing around 3x annually across each of the last 8 years to achieve its performance improvements. Last year, spending across all labs was $150bn. Keeping the 3x trend means that, to keep pace with current advances, costs should rise to $450bn in 2025, $900bn in 2026, $2.7tn in 2027, $8.1tn in 2028, $25tn in 2028, and $75tn in 2029 and $225tn in 2030. For reference, the GDP of the world is around $125tn.
I think the labs will be crushed by the exponent on their costs faster white-collar work will be crushed by the 5% improvement exponent.
Be careful you're not confusing the costs of training an LLM and the spending from each firm. Much of that spending is on expanding access to older LLMs, building new infrastructure, and other costs.
Your math is a bit less than it should be because you doubled instead of trebled 2026
The current trained models are already pretty good enough for many things.
Is that so? Ok let the consumers decide - increase the price and let's see how many users are willing to pay the price.
Are LLMs stackable? If they keep misunderstanding each other, it'll look more like successive applications of JPEG compression.
By all accounts, yes.
"Model collapse" is a popular idea among the people who know nothing about AI, but it doesn't seem to be happening in real world. Dataset quality estimation shows no data quality drop over time, despite the estimates of "AI contamination" trickling up over time. Some data quality estimates show weak inverse effects (dataset quality is rising over time a little?), which is a mindfuck.
The performance of frontier AI systems also keeps improving, which is entirely expected. So does price-performance. One of the most "automation-relevant" performance metrics is "ability to complete long tasks", and that shows vaguely exponential growth.
Given the number of academic papers about it, model collapse is a popular idea among the people who know a lot about AI as well.
Model collapse is something demonstrated when models are recursively trained largely or entirely on their own output. Given most training data is still generated or edited by humans or synthetic, I'm not entirely certain why one would expect to see evidence of model collapse happening right now, but to dismiss it as something that can't happen in the real world seems a bit premature.
We've found in what conditions does model collapse happen slower or fails to happen altogether. Basically all of them are met in real world datasets. I do not expect that to change.
The jpeg compression argument is still valid.
It's lossy compression at the core.
In 2025 you can add quality to jpegs. Your phone does it and you don't even notice. So the rhetorical metaphor employed holds up, in that AI is rapidly changing the fundamentals of how technology functions beyond our capacity to anticipate or keep up with it.
> add quality to jpegs
Define "quality", you can make an image subjectively more visually pleasing but you can't recover data that wasn't there in the first place
You can if you know what to fill from other sources.
Like, the grill of a car. If we know the make and year, we can add detail with each zoom by filling in from external sources
This is an especially bad example, a nice shiny grille is going to be strongly reflecting stuff that isn't already part of the image (and likely isn't covered well by adjacent pixels due the angle doubling of reflection).
Is this like how crypto changed finance and currency
I don't think it is.
Sure, you can view an LLM as a lossy compression of its dataset. But people who make the comparison are either trying to imply a fundamental deficiency, a performance ceiling, or trying to link it to information theory. And frankly, I don't see a lot of those "hardcore information theory in application to modern ML" discussions around.
The "fundamental deficiency/performance ceiling" argument I don't buy at all.
We already know that LLMs use high level abstractions to process data - very much unlike traditional compression algorithms. And we already know how to use tricks like RL to teach a model tricks that its dataset doesn't - which is where an awful lot of recent performance improvements is coming from.
Sure, you can upscale a badly compressed jpeg using ai into something better looking.
Often the results will be great.
Sometimes the hallucinated details will not match the expectations.
I think this applies fundamentally to all of the LLM applications.
And if you get that "sometimes" down to "rarely" and then "very rarely" you can replace a lot of expensive and inflexible humans with cheap and infinitely flexible computers.
That's pretty much what we're experiencing currently. Two years ago code generation by LLMs was usually horrible. Now it's generally pretty good.
I think you are selling yourself short if you believe you can be replaced by a next token predictor :)
I think humans who think they can't be replaced by a next token predictor think too highly of themselves.
LLMs show it plain and clear: there's no magic in human intelligence. Abstract thinking is nothing but fancy computation. It can be implemented in math and executed on a GPU.
Why isn't it then
LLMs have no ability to reason whatsoever.
They do have the ability to fool people and exacerbate or cause mental problems.
what's actually happening is all your life you've been told by experience if something can talk to you is that it must be somewhat intelligent.
Now you get can't around that this might not be the case.
You're like that beetle going extinct mating with beer bottles.
https://www.npr.org/sections/krulwich/2013/06/19/193493225/t...
"What's actually happening" is all your life you've been told that human intelligence is magical and special and unique. And now it turns out that it isn't. Cue the coping.
We've already found that LLMs implement the very same type of abstract thinking as humans do. Even with mechanistic interpretability being in the gutters, you can probe LLMs and find some of the concepts they think in.
But, of course, denying that is much less uncomfortable than the alternative. Another one falls victim to AI effect.
> "What's actually happening" is all your life you've been told that human intelligence is magical and special and unique. And now it turns out that it isn't. Cue the coping.
People have been arguing this is not the case for at least hundreds of years.
Considering we don't understand consciousness at ALL or how humans think, you might want to backtrack your claims a bit.
Any abstraction you're noticing in an LLM is likely just a plagiarized one
this boring reductionist take on how LLMs work is so outdated that I'm getting second hand embarassment.
Sorry, I meant a very fancy next token predictor :)
Lots of technology is cool if you get to just say “if we get rid of the limitations” while offering no practical way to do so.
It’s still horrible btw.
Pretty crazy, and all you have to do is assume exponential performance growth for as long as it takes.
> These models are extremely unreliable when unsupervised.
> It doesn't feel like that will change fundamentally with just incrementally better training.
I could list several things that I thought wouldn't get better with more training and then got better with more training. I don't have any hope left that LLMs will hit a wall soon.
Also, LLMs don't need to be better programmers than you are, they only need to be good enough.
No matter how much better they get, I don't see any actual sign of intelligence, do you?
There is a lot of handwaving around the definition of intelligence in this context, of course. My definition would be actual on the job learning and reliability i don't need to second guess every time.
I might be wrong, but those 2 requirements seem not compatible with current approach/hardware limitations.
Intelligence doesn't matter. To quote "Superintelligence: Paths, Dangers, Strategies":
> There is an important sense, however, in which chess-playing AI turned out to be a lesser triumph than many imagined it would be. It was once supposed, perhaps not unreasonably, that in order for a computer to play chess at grandmaster level, it would have to be endowed with a high degree of general intelligence.
The same thing might happen with LLMs and software engineering: LLMs will not be considered "intelligent" and software engineering will no longer be thought of as something requiring "actual intelligence".
Yes, current models can't replace software engineers. But they are getting better at it with every release. And they don't need to be as good as actual software engineers to replace them.
There is a reason chess was "solved" so fast. The game maps very nicely onto computers in general.
A grandmaster chess playing ai is not better at driving a car than my calculator from the 90s.
Yes, that's my point. AI doesn't need to be general to be useful. LLMs might replace software engineers without ever being "general intelligence".
Sorry for not making my point clear.
I'm arguing that the category of the problem matters a lot.
Chess is, compared to self-driving cars and (in my opinion) programming, very limited in its rules, the fixed board size and the lack of "fog of war".
I think I haven't made my point clear enough:
Chess was once thought to require general intelligence. Then computing power became cheap enough that using raw compute made computers better than humans. Computers didn't play chess in a very human-like way and there were a few years where you could still beat a computer by playing to its weaknesses. Now you'll never beat a computer at chess ever again.
Similarly, many software engineers think that writing software requires general intelligence. Then computing power became cheap enough that training LLMs became possible. Sure, LLMs don't think in a very human-like way: There are some tasks that are trivial for humans and where LLMs struggle but LLMs also outcompete your average software engineer in many other tasks. It's still possible to win against an LLM in an intelligence-off by playing to its weaknesses.
It doesn't matter that computers don't have general intelligence when they use raw compute to crush you in chess. And it won't matter that computers don't have general intelligence when they use raw compute to crush you at programming.
The proof that software development requires general intelligence is on you. I think the stuff most software engineers do daily doesn't. And I think LLMs will get continously better at it.
I certainly don't feel comfortable betting my professional future on software development for the coming decades.
Ive heard this described as a kind vs a wicked learning environment.
"It is difficult to get a man to understand something when his salary depends upon his not understanding it" ~ Upton Sinclair
Your stance was the widely held stance not just on hacker news but also by the leading proponents of ai when chatgpt was first launched. A lot of people thought the hallucination aspect is something that simply can't be overcome. That LLMs were nothing but glorified stochastic parrots.
Well, things have changed quite dramatically lately. AI could plateau. But the pace at which it is improving is pretty scary.
Regardless of real "intelligence" or not.. the current reality is that AI can already do quite a lot of traditional software work. This wasn't even remotely true if if you were to go 6 months back.
How will this work exactly?
I think I have a pretty good idea of what AI can do for software engineering, because I use it for that nearly every day and I experiment with different models and IDEs.
The way that has worked for me is to make prompts very specific, to the point where the prompt itself would not be comprehensible to someone who's not in the field.
If you sat a rando with no CS background in front of Cursor, Windsurf or Claude code, what do you suppose would happen?
It seems really doubtful to me that overcoming that gap is "just more training", because it would require a qualitatively different sort of product.
And even if we came to a point where no technical knowledge of how software actually works was required, you would still need to be precise about the business logic in natural language. Now you're writing computer code in natural language that will read like legalese. At that point you've just invented a new programming language.
Now maybe you're thinking, I'll just prompt it with all my email, all my docs, everything I have for context and just ask it to please make my boss happy.
But the level of integrative intelligence, combined with specialized world knowledge required for that task is really very far away from what current models can do.
The most powerful way that I've found to conceptualize what LLMs do is that they execute routines from huge learnt banks of programs that re-combine stored textual information along common patterns.
They're cut and paste engines where the recombination rules are potentially quite complex programs learnt from data.
This view fits well with the strengths and weaknesses of LLMs - they are good at combining two well understood solutions into something new, even if vaguely described.
But they are quite bad at abstracting textual information into a more fundamental model of program and world state and reasoning at that level.
I strongly suspect this is intrinsic to their training, because doing this is simply not required to complete the vast majority of text that could realistically have ended up in training databases.
Executing a sophisticated cut&paste scheme is in some ways just too effective; the technical challenge is how do you pose a training problem to force a model to learn beyond that.
Ironic to post that quote about AI considering the hype is pretty much entirely from people who stand to make obscene wealth from it.
>That LLMs were nothing but glorified stochastic parrots.
Well yes , now we know they make kids kill themselves.
I think we've all fooled ourselves like this beetle
https://www.npr.org/sections/krulwich/2013/06/19/193493225/t...
for thousands of years up until 2020 anything that conversed with us could safely be assumed to be another sentient/intelligent being.
No we have something that does that, but is neither sentient or intelligent, just a (complex)deterministic mechanism.
LLMs can code, but they can’t engineer IMO. They lack those other parts of the brain that are not the speech center.
I'd argue that, out of white collar jobs, it is actually one of the least automatable still. I.e. the rest of the jobs are likely going to get disrupted much faster because they are easier to automate (and have been the target of automation by the software industry in the past century). Whatever numbers were seeing now may be too early to reflect this accurately.
Also there are different metrics that are relevant like dollar count vs pure headcount. Cost cutting targets dollars. E.g. entry level developers are still expensive compared to other jobs.
Most "Software Engineering" is just applying the same code in slightly different contexts. If we were all smarter it would have been automated earlier through the use of some higher-level language.
I'm sure those who lost a job to software at some point are feeling a great deal of sympathy for developers who are now losing out to automation.
Despite being the target of a lot of schadenfreude, most software developers aren't working on automation.
Nice watching it tear down recruiters though.
Its the least regulated (not at all). So it will be the first to be changed.
AI lawyers? Many years away.
AI civil engineers? Same thing, there is a PE exam that protects them.
You don’t need to perfect AI to the point of becoming credentialed professionals to gut job markets— it’s not just developers, or creative markets. Nobody’s worried that the world won’t have, say, lawyers anymore — they’re worried that AI will let 20% of the legal workforce do 100% of the requisite work, making the skill essentially worthless for the next few decades because we’d have way too many lawyers. Since the work AI does is largely entry-level work, that means almost nobody will be able to get a foothold in the business. Wash, rinse, repeat to varying levels across many white collar professions and you’ve got some real bad times brewing for people trying to enter the white collar workforce from now on— all without there being a single AI lawyer in the world.
Same thing for doctors. Turns out radiologists are fine, it's software engineers that should be scared.
We might end up needing 20% or so less doctors, because all that bureaucracy can be automated. A simple automated form pre-filler can save a lot of time. It’s likely that hospitals will try saving there.
Which universe is that, the one consisting of the union of AI charlatans and people who don’t understand software engineering?
You know even the CEOs are backtracking on that nonsense right?
Universally? Nah.
Too bad engineers were “too important” to unionize because their/our labor is “too special .”
I think you could find 10,000 quotes from HN alone why SDEs were immune to labor market struggles that would need a union
Oh well, good luck everyone.
I'm not necessarily opposed to unionization in general but it's never going to save many US software industry jobs. If a unionization drive succeeds at some big tech company then the workers might do well for a few years. But inevitably a non-union startup competitor with a lower cost structure and more flexible work rules will come along and eat their lunch. Then all the union workers will get laid off anyway.
Unionization kind of worked for mines and factories because the company was tied to a physical plant that couldn't easily be moved. But software can move around the world in milliseconds.
Unions _can_ protect against this, but they have to do it via lobbying the government for protectionism, tariffs, restricting non-union competition etc.
Indeed, just look at the CGI VFX industry of Hollywood. US invented it and was the leader for a long time, but now it has been commodified, standardized and run into the ground, because union or not, you can't stop US studios form offshoring the digital asset work to another country where labor is 80% cheaper than California and quality is 80% there. So the US is left with making the SW tools that VFX artist use, as the cutting edge graphics & GPU knowhow is all clustered there.
Similarly, a lot of non-cutting edge SW jobs will also leave the US as tooling becomes more standardized, and other nations upskill themselves to deliver similar value at less cost in exchange for USD.
This was when programmers were making software to time Amazon worker's bathroom breaks so believing "this could never happen to me" was probably an important psychological crutch.
Saying “programmers” did this is about as useful as saying humans did it.
This is, if true, a fundamental shift in the value of labor. There really isn’t a non-Luddite way to save these jobs without destroying American tech’s productivity.
That said, I’m still sceptical it isn’t simply a reflection of an overproduction of engineers and a broader economic slowdown.
Yeah I agree that outsourcing and oversupply are the real culprits and AI is a smoke screen. The outcome is the same though.
> outcome is the same though
Not really. If it’s overproduction, the solution is tighter standards at universities (and students exercising more discretion around which programmes they enroll in). If it’s overproduction and/or outsourcing, the solutions include labour organisation and, under this administration, immigration curbs and possibly services tariffs.
Either way, if it’s not AI the trend isn’t secular—it should eventually revert. This isn’t a story of junior coding roles being fucked, but one of an unlucky (and possibly poorly planning and misinformed) cohort.
It can be oversupply/outsourcing and also secular: You can have basically chronic oversupply due to a declining/maturing industry. Chronic oversupply because the number of engineers needed goes down every year and the pipeline isn't calibrated for that (academia has been dealing with this for a very long time now, look up the postdocalypse). Outsourcing, because as projects mature and new stuff doesn't come along to replace, running maintenance offshore gets easier.
Software isn't eating the world. Software ate the world. New use cases have basically not worked out (metaverse!) or are actively harmful.
Unions won’t solve this for you. If a company just decides they have enough automation to reduce union workforce it can happen the next time contracts get negotiated.
Either way, there are layoff provisions with union agreements.
Tell that to dock workers, who have successfully delayed the automation of ports to the extent we see them automated in e.g. the PRC [0].
Hell, they're even (successfully) pushing back against automated gates! [1]
[0] https://www.cnn.com/2024/10/02/business/dock-workers-strike-...
[1] https://www.npr.org/2024/10/03/nx-s1-5135597/striking-dockwo...
Isn't that just delaying the inevitable? Yangshan Deep-Water Port in Shanghai is one of the most automated ports. Considering there are more people in China than in the US, China still automated their port.
I'm not making a value judgment on the specific case of dock workers, I'm rather saying that unions can and do prevent automation. If Software Devs had unionized earlier, a lot of positions would probably still be around.
The dock owner may not have a lot of alternatives to negotiating with the union. If devs unionize, the work can move.
In Hollywood, union bargaining bought some time at least. Unions did mandate limits on the use of AI for a lot of the creation process.
AI is still used in Hollywood but nobody is proud of it. No movie director goes around quoting percentages of how many scenes were augmented by AI or how many lines in the script were written by ChatGPT.
So what your argument is we're so special that we deserve to hold back human progress to have a privileged life? If it's not that what would you want a union to do in this situation?
I’d prefer that my family are financially stable over “human progress”. One benefits me and the other benefits tech companies. Easy choice.
If our ancestors had thought like that we'd all be very busy and "stable" doing subsistence farming like we were doing 10,000 years ago.
Better our children never have to work because the robots do everything and they inherited some ownership of the robots.
Do you really believe that all technological progress has bettered humanity? Where’s the four day work week we were promised? I thought automation was supposed to free us from labor.
Unions wouldnt stop any of this but professionalization would
Unions work in physical domains that need labor “here and now”, think plumbers, electricians, and the like. You can’t send that labor overseas, and the union can control attempts at subversion via labor force importation. But even that has limitations, e.g. union factory workers simply having their factory shipped overseas.
Software development at its core can be done anywhere, anytime. Unionization would crank the offshoring that already happens into overdrive.
Unions can only prevent automation up to a point. Really the only thing that could have reasonably prevented this would have been for programmers to not produce as much freely accessible training data (formerly known as "open source software").
Exactly. I am always so impressed by the fact that developers never see that open source is essentially them giving away free labor to giant corporations. Developers basically programmed their way out of a job, for free. It's the only profession that is proud to have its best work done on unpaid time and used for free by big corporations.
We're not "too important." All a union would do is create extra problems for us.
There are two possibilities:
a) This is a large scale administrative coordination problem
b) We don't need as many software engineers.
Under (a) unionizing just adds more administrators and exacerbates the problem, under (b) unions are ineffective and just shaft new grads or if they manage to be effective, kills your employer (and then no one has a job.)
You can't just administrate away reality. The reason SWEs don't have unions is because most of us (unlike blue collar labor) are intelligent enough to understand this. I think additionally there was something to be said about factory work where the workers really were fungible and it was capital intensive, software development is almost the polar opposite where there's no capital and the value is the theory the programmers have in their head making them a lot less fungible.
Finally we do have legal tools like the GPL which do actually give us a lot of negotiating power. If you work on GPL software you can actually just tell your employer "behave or we'll take our ball and leave" if they do something stupid.
Unions would just delay the inevitable while causing other downsides like compressing salary bands, make it difficult to fire non-performers, union fees, increasing chance of corruption etc.
For a recent example:
> Volkswagen has an agreement with German unions, IG Metall, to implement over 35,000 job cuts in Germany by 2030 in a "socially responsible" way, following marathon talks in December 2024 that avoided immediate plant closures and compulsory layoffs, according to CNBC. The deal was a "Christmas miracle" after 70 hours of negotiations, aiming to save the company billions by reducing capacity and foregoing future wage increases, according to MSN and www.volkswagen-group.com.
I mean, I still don't want to unionize with the guys who find `git` too complicated to use (which is apparently the majority of HN). Also, you guys all hate immigrants which is not my vibe, sorry.
I really hope nobody had themselves convinced that software engineering couldn't be automated. Not with the code enterprise has been writing for decades now (lots and lots and lots of rules for gluing state to state, which are extremely structured but always just shy of being so structured that they were amenable to traditional finite-rule-based automation).
The goal of the industry has always been self-replacement. If you can't automate at least part of what you're working on you can't grow.
... unfortunately, as with many things, this meshes badly with capitalism when the question of "how do you justify your existence to society" comes up. Hypothetically, automating software engineering could lead to the largest open-source explosion in the history of the practice by freeing up software engineers to do something else instead of toil in the database mines... But in practice, we'll probably have to get barista jobs to make ends meet instead.
The experiences people are having when working with big, complex codebases don’t line up with your gloomy outlook. LLMs just fall apart beyond a certain project size, and then the tech debt must be paid.
Is it gloomy? I personally liken it to inventing the washing machine instead of doing laundry by hand, beating it against a washboard, for another hundred years.
If you want to know what will happen to software engineers in the US just follow the path of US factory workers in the 90s.
AI is the popular cover excuse for layoffs.
I can’t think of a single job that modern AI could easily replace.
I think the initial job loss from AI will come from having individual workers be more productive and eliminate the need to have larger teams to get the same work done.
Eventually, maybe. Right now I see a lot more people wasting time with AI in search of these promised efficiencies. A lot of companies reducing headcount are simply hiding the fact that they are deprioritizing projects or reducing their overall scope because the economy is shit (I know, I know - but it feels worse than reported IMO) and that's the right business cycle thing to do. If you're dramatic and take the DOGE/MAGA approach to management, just fire everyone and the important issues will become obvious where investment is actually needed. It's a headcount 'zero based budget' played out IRL. The truth is, there is a lot of fat to be cut from most large companies and I feel like it's the current business trend to be ruthless with the blade, especially since you have AI as a rose colored scapegoat.
The way I like to describe it is that you can't go from 1 developer to 0 thanks to AI, but you might be able to go from 10 to 9. Although not sure what the exact numbers are.
For cost centers, maybe. If your development team or org is a revenue generator with a backlog, I don't see why the team would be trimmed.
I'll go further than you. Even if the team is a cost center, it may not make sense to reduce the headcount if there's still more work to do. After all, an internal team that just assists other teams in the company without directly creating value suddenly become more productive could in turn make the other teams more productive. Automatically reducing headcount after a productivity increase is like that effect where people drive more dangerously when wearing seatbelts.
I used to hire someone who worked part time from home to bookmark some of the key pages in thousands of pdfs just so that I can directly jump to those pages instead of spending time myself on finding those pages.
AI can now do it very cheap so no need to give that job to a human anymore.
Have you taken a Waymo yet?
no
High interest rates + tariff terror -> less investment -> less jobs
But let's blame AI
Let's read the paper instead: https://digitaleconomy.stanford.edu/wp-content/uploads/2025/...
It presents a difference-in-differences (https://en.wikipedia.org/wiki/Difference_in_differences) design that exploits staggered adoption of generative AI to estimate the causal effect on productivity. It compares headcount over time by age group across several occupations, showing significant differentials across age groups.
Page 3: "We test for a class of such confounders by controlling for firm-time effects in an event study regression, absorbing aggregate firm shocks that impact all workers at a firm regardless of AI exposure. For workers aged 22-25, we find a 12 log-point decline in relative employment for the most AI-exposed quintiles compared to the least exposed quintile, a large and statistically significant effect."
I appreciate the link to differences in differences, I didn't know what to call this method.
The OP's point could still be valid: it’s still possible that macro factors like inflation, interest rates, or tariffs land harder on the exact group they label ‘AI-exposed.’ That makes the attribution messy.
Those fixed effects are estimated separately for each age group, controlling for that.
pg. 19, "We run this regression separately for each age group."
Were entry level jobs the first to go in earlier developer downturns?
Is AI being used to attempt to mitigate that effect?
I don't think their methods or any statistical method could decouple a perfectly correlated signal.
Without AI, would junior jobs have grown as quickly as other?
You really do have to account for why this is mainly happening in industries that are adopting AI, why it's almost exclusively impacting entry-level positions (with senior positions steady or growing), and why controlling for broad economic conditions failed to correct this. I doubt very much that these three Stanford professors would be blindsided by the concept of rates and tarriffs.
My personal theory is that the stock market rewards the behavior of cutting jobs as a signal of the company being on the AI bandwagon. Doesn't matter if the roles were needed or not. Line goes up, therefore it is good.
This is a complete reversal in the past where having a high headcount was an easy signal of a company's growth (i.e. more people, means more people building features, means more growth).
Investors are lazy. They see one line go down, they make the other line go up.
CEOs are lazy. They see line go up when other line goes down. So they make other line go down.
(I am aware that "line go up" is a stupid meme. But I think it's a perfect way to describe what's happening. It is stupid, lazy, absurd, memetic. It's the only thing that matters, stripped off of anything that is incidental. Line must go up.)
Given the timeline this is more likely a reversion to the mean following the end of zero interest rate policy.
Juniors become seniors.
If we replace all juniors with AI, in a few years there won't be skilled talent for senior positions.
AI assistance is a lot different than AI running the company. Making expensive decisions. While it could progress, bear in mind that some seniors continue to move up in the ranks. Will AI eventually be the CEO?
We all dislike how some CEOs behave, but will AI really value life at all? CEOs have to have some place to live, after all.
The AI will at least be cheaper than a CEO, it might also be more competent and more ethical. The argument against making a Large Language Model the CEO seems to mostly be about protecting the feelings of the existing CEO, maybe the Board should look past these "feelings" and be bold ?
I'll re-explain.
A human CEO might do morally questionable things. All do not, of course, but some may.
Yet even so, they need a planet with air, water, and some way to survive. They also may what their kids to survive.
An AI may not care.
It could be taking "bad CEO" behaviour to a whole new level.
And even if the AI had human ethics, humans play "us vs them" games all the time. You don't get much more "them" than an entirely different lifeform.
The AI most certainly does not care, because it is a computer program. It also doesn't want to buy a boat.
It also doesn’t care if the company goes bankrupt tomorrow without paying out their bonus.
Software development is one of the listed industries. Well before AI we have seen that few companies wanted entry level devs due to the training and such.
Reducing in call centers has been going on for a while as more people use automated solutions (not necessarily AI) and many of the growing companies make it hard to reach a real person anyways (Amazon, Facebook, etc). I feel like AI is throwing fuel on the existing fire, but isn't as much of a driver as the headlines suggest.
The jobs are going to India
American workers are truly under attack from all sides. H1B. Outsourcing. What's left? The blue collar manufacturing is mostly gone. White collar work well on its way out. Why is our own government (by the people for the people) actively assisting in destroying American's ability to get jobs (H1B)? Especially in these conditions. I'm no racist or idiot but it's unacceptable. I didn't expect the gov to actively be conspiring with big corps to make my economic position weaker. Unbelievable breach of trust. We need to demand change from our government.
Efficiency rules all.
It just doesn't make sense to pay someone $10 when you can pay someone else $2
And when we're all out of work except for the doctors and nurses, electricians and plumbers, there will be nobody to contribute to consumer spending. And we will suffer, at the hand of the government that assisted in this scam.
It’s an unpopular opinion in the current environment but it’s the program that allows international talent to connect with local capital that creates all the jobs in tech.
Nearly half the unicorns in the country were found by foreigners living in the country. https://gfmag.com/capital-raising-corporate-finance/us-unico...
The biggest problem right now is that there is no distinction between companies replacing Americans labor with cheap labor and entrepreneurial talent that creates jobs. Everyone is on the same visa.
They will come back (eventually).
Having to work with ESL contractors from firms like Cognizant or HCL is true pain. Normally it would be like 3-4 US employees working on something and then its like 20-30 ESL outsourced people working on something. The quality is so poor though its not worth it.
My current org nuked their contract w HCL after 2 years because how shitty they are and now everything is back onshore. Millions wasted lol. Corporations are so silly sometimes.
They also need 5 people to do the work of one us worker. And then another US worker to guide and do some qa on the output they produce . I don't see how it saves money. There are other countries with lower wages than the US where this doesn't happen such as Poland or Australia.
> You really do have to account for why this is mainly happening in industries that are adopting AI
Correlation is not causation. The original research paper does not prove a connection.
> I doubt very much that these three Stanford professors would be blindsided by the concept of rates and tarriffs.
They are nonetheless subject to publish or perish pressure and have strong incentives to draw publishable attention-grabbing results even where the data is inconclusive.
Tariffs are just a massive government revenue generating consumption tax on particular industries. We would expect unemployment among the young trying to enter those industries to be hit hardest.
Do you understand that American employers don't have to pay American tariffs?
Everyone pays mate
i'm curious who you think pays american tarrifs
You first
> Do you understand that American employers don't have to pay American tariffs?
Except they do, if their raw materials, tools, etc., are imported.
More investment -> more return on investment -> "AI is increasing worker efficiency" -> This is good for AI.
Less investment -> more layoffs -> "AI is replacing workers" -> This is good for AI.
A computer does something good -> "That's AI" -> This is good for AI.
A computer does something bad -> "It needs more AI" -> This is good for AI.
It seems more true than the "this is good for bitcoin" meme now that bitcoin seems to track the dollar very closely
Is there some central authority that’s telling people to blame this all on AI, or how is everyone reaching this conclusion and ignoring the other obvious factors you stated?
It is in their interest to find explanations for reductions in labor that don't assign the blame to corporate greed.
For example, a call center might use the excuse of AI to fire a bunch of people. They would have liked to just arbitrarily fire people a few years ago, but if they did that people would notice the reduction in quality and perhaps realize it was done out of self-serving greed (executives get bigger bonuses / look better, etc). The AI excuse means that their service might be worse, perhaps inexcusably so, but no one is going to scrutinize it that closely because there is a palatable justification for why it was done.
This is certainly the type of effect I feel like underlies every story of AI firing I've heard about.
How is firing a bunch of people because you made a machine that you believe can do their jobs not textbook corporate greed? It seems like the worst impulses of Taylorism made manifest?
> It is in their interest to find explanations for reductions in labor that don't assign the blame to corporate greed.
Exactly.
Since this article is about AI, and since this comment seems rather low effort compared to the Stanford study, I went ahead and used low effort to analyze the report compare it to this comment. Here's my low effort AI response:
> Prompt: Attached is a paper. Below is an argument made against it. Is there anything in the paper that addresses the argument?: High interest rates + tariff terror -> less investment -> less jobs
> High rates/firm shocks: They add firm–time fixed effects that absorb broad firm shocks (like interest-rate changes), and the within-firm drop for 22–25-year-olds in AI-exposed roles remains.
> “Less investment” story: They note the 2022 §174 R&D amortization change and show the pattern persists even after excluding computer occupations and information-sector firms.
> Other non-AI explanations: The decline shows up in both teleworkable and non-teleworkable jobs and isn’t explained by pandemic-era education issues.
> Tariffs: Tariffs aren’t analyzed directly; broad tariff impacts would be soaked up by the firm–time controls, but a tariff-specific, task-level channel isn’t separately tested.
Fitting, since it came up with unrelated information (the R&D tax thing) and the 3rd bullet point. Also started talking about tariffs as if it had addressed them, then notes that it doesn't address them.
Blaming AI is better because it helps corporations convince the working class that there jobs are in long-term danger so they collectively settle for less favorable work terms and compensation, unlike if they are convinced that it is going to gradually improve with the upcoming monetary easing cycle..
I'm sorry, have you read the paper, or did you just want to recite those here?
End of ZIRP and the Sec. 179 change for engineering salaries probably explains more of this (plus the increase in outsourcing). I’m sure some decision makers also threw AI into the mix but the financials of hiring software engineers in the US was already challenging before AI “took everyone’s job”.
Here's the study:
https://digitaleconomy.stanford.edu/publications/canaries-in...
It looks like they're looking at data for the last few years, not just the last few months.
I haven't read it, and maybe you can disagree with their opinions, but there does appear to be a slow down in college graduates recently.
I generally agree that AI is the scapegoat, but not for those same reasons. Despite the lack of job growth and the tariffs, recent data shows the economy grew about 3%. Even if it's not AI as the primary driver, efficiency seems to have increased.
How does that make sense? Wouldn’t high interest rates and tariffs cause more expensive engineers to have disproportionate opportunity? I remember during 2008 it was much easier for my employer to justify junior engineers than senior ones.
(High interest rates + tariff terror -> less investment -> less jobs) + AI
Do you consider things to be that single-faceted, that other factors cannot realistically be a part of the equation?
I have to admit that something is "single-faceted" would be a nice break from hearing that something is "complex and multifaceted".
Well, you do have CEOs out there saying it...
2 things can be true
Less investment? You must be trolling. I encourage you to look at the about of stupid money that has been “invested” into LLMs.
> But let's blame AI
The thing whose exact purpose is to replace labor? Must be a conspiracy going on to suggest its linked to reducing labor. Bias! Agenda!
The jobs are going overseas
CEOs citing savings from AI should be able to show higher profits soon. The fact that they’re not means those tall tales are coming home to roost soon.
Nah, its going to be like when everyone included "bitcoin" in their quarterly reports and the market goes nuts, until it stops
Every day when I am out in the city, I am amazed by how many jobs we have NOT managed to replace with AI yet.
For example, cashiers. There are still many people spending their lives dragging items over a scanner, reading a number from a screen, holding out their hand for the customer to put money in, and then sorting the coins into boxes.
How hard can it be to automate that?
They don’t need AI for that, they just cut staff to the bare minimum and put in self checkouts.
And then they hire supervisors, helpers and checkout guards/security. I hope it at least makes sense on paper.
>How hard can it be to automate that?
Self checkout has been a thing for ages. Heck in Japan the 711s have cashiers but you put the money into a machine that counts and distributes change for them.
Supermarkets are actually getting rid of self checkouts due to crime. Surprise surprise, having less visible "supervision" in a store results in more shoplifting than having employees who won't stop it anyway.
It’s also just resulting in atrocious customer experience.
I can go to Safeway or the smaller chain half a block away.
The Safeway went all in on self checkouts. The store is barely staffed, shelves are constantly empty, you have to have your receipt checked by security every time, they closed the second entrance permanently, and for some reason the place smells.
Other store has self checkouts but they also have loads of staff. I usually go through the normal checkout because it’s easier and since they have adequate staff and self checkout lines it tends to be about the same speed to.
End result is I don’t shop at Safeway if I can avoid it.
Amazon could not do it. They claimed they could, but it was just indians watching the video and tabulating totals overseas
The hard part is preventing theft, not adding numbers.
Cashiers should not, and will not prevent theft. They're not paid nearly enough to get in danger, and it is not their job.
I'm sure you can find videos of thefts in San Francisco if you need a visual demonstration. No cashier is going to jump in front of someone to stop a theft.
That's not the type of theft they were talking about. Rather, self scanners purposely not scanning items to get them for free, etc
I had a roommate in college who used to stuff containers of beef into produce bags full of kale, and weigh that on the self-service scanner.
True, but having a cashier standing there waiting to scan your items will prevent most normal people from stealing. Sure, some will brazenly walk right past with a TV on their shoulder, but most people won't.
If there's no cashier and you're doing it yourself, a whole lot more people will "forget" to scan a couple items, and that adds up.
There's usually a security person or two in the store, looking over the self checkouts. I agree that job prevents a lot of people from becoming opportunistic thiefs, but I'm making a distinction between cashiers and security. Today the store needs both.
Pretty sure if a "security person" worked so well, Walmart wouldn't be severely reducing self checkouts at their stores to Walmart Plus members only.
That might be regional, then. I wouldn't say $COUTNRY is exactly a high-trust society, but it's not quite that bad for us over here.
A thief doesn't know what a cashier will do. And a cashier is an eye witness or can yell "hey stop them!"
You're doing the all or nothing fallacy. The fact that a cashier does not prevent all thefts does not mean a cashier does NOTHING for theft.
> The fact that a cashier does not prevent all thefts does not mean a cashier does NOTHING for theft.
Yes, for one thing, it ignores that a very large share of retail theft is insider theft, and that cash handling positions are the largest portion of that.
Cashiers absolutely do something for theft.
They absolutely do. It’s not the cashiers being security, it’s having adequate staffing making people less likely to steal. Its not stopping crimes that have occurred it’s just reducing opportunistic theft.
Is the theft really happening at the checkout?
And if so, why can't we detect it via camera + AI?
You detect someone leaving your store with a 4€ item. What then?
You ban them from coming back in after a few warnings. Stores seem really icy about facial recognition right now though. The optics are pretty bad (a play on words pun?)
Who is going to stop them from coming back in?
There are stores that are abandoning self-checkouts completely and going back to cashiers as the theft rose to unsustainable numbers.
Checkouts are often only egress points. So having pair of eyes over them does have some effect compared to having none at all.
Detecting theft does not mean theft is prevented. You then need the government to prosecute, and impose sufficient punishment to deter theft. This is not cheap, nor a given that it will happen.
So take the broken god awful experience of self checkout and add another layer of “I think you did something wrong so now you have to stand around waiting for an actual person”?
No thanks.
You mean ordering kiosks and self-checkout machines? We have automated it, it's just not everywhere has implemented it.
The one I'm desperately waiting for is serverless restaurants—food halls already do it but I want it everywhere. Just let me sit down, put an order into the kitchen, pick it up myself. I promise I can walk 20 feet and fill my own drink cup.
Japan does this a lot of places, and it makes the experience much easier.
And I think the entire mid and low range restaurants could replace servers with a tablet and people would be happier. I'm not sure how it doesn't make more money for the restaurant too, making it so easy to order more during a meal.
You seem to like self-checkout processes. I don't. I avoid any place where I have to interact with a screen. Be it a screen installed on-premise or the screen on my phone. It is not a relaxing experience for me.
Serverless restaurants have been common in Australia for decades. You just get a buzzer and then need to go pick up your food when it is ready. There's a single person behind the bar to take orders and pour beer/wine/soda.
I don't use self-checkouts at the stores, nor would I eat at automated or self-service restaurants. I have a kitchen for that already.
But it's good if both are available, as apparently there will be customers for both.
Seems like perfect option for robots (not humanoid). Bring me my food. You can still keep people in kitchen for a bit, but well servers in many restaurants are not really needed.
Self check-out machines aren't automation.
There used to be two humans standing at the cash register, now because of software, automatic change machines, and cameras there is only one. One of those humans' jobs got automated.
Call it what you like but replacing the work of humans one for one is difficult and usually not necessary. Reformulating the problem to one that machines can solve is basically the whole game. You don't need a robot front desk worker to greet you, you just need a tablet to do your check in.
I do their work. No work got automated.
This. And I do their work a lot more slowly because it's not my regular job, and I actually already had to do some of the work (getting the items out of my trolley and onto the conveyor). Now I stand there forever fumbling with barcodes, trying to get bags to stay open, switching between getting items out of the trolley and scanning. The old checkout system is so much more efficient when you are buying anything more than a couple of items at a time.
Yeah this is like saying Aldi “automated” cart return. They didn’t, they got every shopper to do the work themselves. Automated cart return would be if you just gave the cart a little “giddyup!” when you were done and it found its way home. Or those cart conveyor belts at Ikea, it’s only part of the process but that part is automated.
[edit] Aldi did automate the management of getting shoppers to do that work, because there’s not a person standing there taking and handing out quarters, but (very simple) machines. Without those machines they might need a person, so that hypothetical role (the existence of which might make the whole scheme uneconomical) is automated. But they didn’t automate cart return, all that work’s still being done by people.
Pharmacists are my favourite. They're a human vending machine that is bad at counting and reading. But law protects them. Pretty good regulatory capture.
Please actually understand what pharmacists actually do and _why_ AI is not a good replacement for them yet, unless you want to die of certain drugs interactions.
Hahaha, this drug interaction nonsense is what online people tell each other. It isn't even real. It's like "nice trigger discipline" or "the postal police don't fuck around" and shit like that. Just something that is not true but for some reason is internet urban legend.
Retail pharmacists are human vending machines. You don't need AI. It's a computer prescription written by a far more qualified human which is then provided to a nigh-illiterate half-wit who will then try as hard as possible to misread it. Having then misread it, the patient must then coax them out of their idiocy until they apologize and fulfill what's written.
Meanwhile some Internet guy who gets all his information from the Internet will repeat what he's heard on the Internet. I know this because anyone passingly acquainted with this would have at least made the clarification between compounding pharmacists and retail pharmacists or something.
Pharmacists are a fantastic example. My pharmacy is delivered my prescription by computer. They text me, by computer, when it's ready to pick up. I drive over there … and it isn't ready, and I have to loiter for 15 minutes.
Also, after the prescription ends, they're still filling it. I just never pick it up. The autonomous flow has no ability to handle this situation, so now I get a monthly text that my prescription is ready. The actual support line is literally unmanned, and messages given it are piped to /dev/null.
The existing automation is hot garbage. But C-suite would have me believe our Lord & Savior, AI, will fix it all.
The only way AI could fix this if it said "replace the pharmacist with a vending machine and hire a $150k junior engineer to make sure the DB is updated afterwards", which you never know, Claude Opus 4 might suggest. At that point, we'll know AGI has been achieved.
Looks like the study pretty arbitrarily picks "exposed industries" and notes that employment rate there has declined.
> Some examples of these highly exposed jobs include customer service representatives, accountants and software developers.
We seem to be in this illogical (delusional?) era where we are being told that AI is 'replacing' people in certain sectors or types of work (under the guise that AI is better or will soon be better than humans in these roles) yet those same areas seem to be getting worse?
- Customer service seems worse than ever as humans are replaced with "AI" that doesn't actually help customers more than 'website chatbots' did 20 years ago.
- Accounting was a field that was desperate for qualified humans before AI. My attempts to use AI for pretty much anything accounting related has had abysmal results.
- The general consensus around software development seems to be that while AI is lowering the barrier of entry to "producing code", the rate of production of tech debt and code that no one "owns" (understands) has exploded with yet-to-be-seen consequences.
> "The general consensus around software development seems to be that while AI is lowering the barrier of entry to "producing code", the rate of production of tech debt and code that no one "owns" (understands) has exploded with yet-to-be-seen consequences."
^ This. (Tho I'm not sure about it being "general consensus".) Vibe code is the payday loan (or high-interest credit card) of tech debt. Demo-quality code has a way of making it into production. Now "everyone" can produce demos and PoCs. Companies that leverage AI as a powerful tool in the hands of experienced engineers may be able to iterate faster and increase quality, but I expect a sad majority to learn the hard way that there's no free lunch, and shipping something you don't understand is a recipe for disaster.
Yeah, in the same way ice cream is linked to homicides!
A recession could also explain this drop.
I’m suss about this paper when it makes this claim:
“where AI is more likely to automate, rather than augment , human labor.”
Where is AI currently automating human labor? Not Software Engineering. Or - what’s the difference between AI that augments me so I can do the job of three people and AI that “automates human labor”
I was also curious about this. Table A1 on page 56 lists examples of positions that are automated vs augmented, and these are the positions the authors think are going to be most augmented (allegedly taken from [0]):
- Chief Executives
- Maintenance and Repair Workers, General
- Registered Nurses
- Computer and Information Systems Managers
After skimming [0], I can't seem to find a listing of jobs that would be augmented vs automated, just a breakdown of the % of analyzed queries that were augmenting vs automating, so I'm a bit confused where this is coming from.
[0]: https://arxiv.org/abs/2503.04761
When the Stanford paper looked at augment vs automate, they used the data from Anthropic's AI Economic Index. That paper defined the terms like this:
We also analyze how AI is being used for tasks, finding 57% of usage suggests augmentation of human capabilities (e.g., learning or iterating on an output) while 43% suggests automation (e.g., fulfilling a request with minimal human involvement).
From the data, software engineers are automating their own work, not augmenting. Anthropic's full paper is here:
https://arxiv.org/html/2503.04761v1
Sounds like a snake eating it's own tail.
What is the effective difference between augment and automate? Either way, fewer man hours are needed to produce the same output.
> What is the effective difference between augment and automate?
The paper says one of those is impacted, and the other isn't.
So, yeah, not only that's what the GP is asking, but I'd like to know it too.
If your job is to swing a hammer, then hammer swinging robot automates your job.
If your job is to swing a hammer, then drill robot augments your job (your job is now swing hammer and drill hole).
How that is different from drill bot automating human driller's job is an exercise left to the reader.
> What is the effective difference between augment and automate?
If the field has a future.
The total output isn't going to stay the same, though.
I see a worrisome trend. On one hand, many of my proto-boomer friends are suffering from age-ism , and memes claim that over-50-year-olds are unemployable. Not 100% fidelity, but there's some truth.
Then I hear about a lot of youngsters struggling to find work, and see articles like this.
Well, who's left? Is there a sweet spot at like 31 that are just cleaning up?
31 would line up with the post house bubble boom recovery
beside the point, but over 50 = proto-boomer? You mean para-boomer, maybe? Gen X is <=60, I believe, so you referring to the cusp boomer/genx I think..
genx is now proto-boomer
Proto as a prefix means it's first or at least before.
Surely this must be linked to a general slowing of the economy.
Sensationalist, alarmist, b.s. article.
It emphasizes "AI adoption linked to 13% decline," which implies causation. The study itself only claims "evidence consistent with the hypothesis."
The article also largely highlights job loss for young workers, while only briefly mentioning cases where AI complements workers.
The study's preliminary status -- it is not peer reviewed -- is noted but only once and at end. If the article was more balanced it would have noted this at the beginning.
Articles on the same subject by the World Economic Forum, McKinsey, and Goldman Sachs are more balance and less alarmist.
Thinly veiled economic propaganda aside, I am dealing with a different AI mess everyday. Technical debt is exploding everywhere I turn. There is an ever larger part of me these days that wishes I could just call the bluff all at once and let all the companies in question learn the inevitable lessons here the hard way.
The worst thing for me would be just needing to get a job like I had before being a dev, the stakes are so much grander for all the companies. It's only really existential for the side of this that isn't me/us. I've been working since I was 15, I can figure it out. I'll be more happy cutting veggies in a kitchen than every single CEO out there when all is said and done!
The study is bs. While executives are blaming AI, it is nowhere near levels of replacement.
What I bet is happening under the covers is reprioritization of work, offshoring or both.
Why bet? In the news recently Australian bank CBA was caught offshoring positions and claiming the jobs had been replaced by AI.
> What I bet is happening under the covers is reprioritization of work, offshoring or both.
AI has been frequently used as an explanation for layoffs.
Before AI, layoffs would be a positive signal to investors, but they'd be demoralizing to staff and/or harm the brand.
Now, you can say, "Wow, we're so good at technology, we're eliminated ___ jobs!" and try to get the best of both worlds.
My company did exactly this earlier in the year. It was a blatant lie and everyone who works here knew it. None of the people laid off were actually replaced with AI, the work they did was just eliminated.
Yeah, unquestioning “journalists” have allowed them to turn laying off thousands into an ad for their new tech.
It’s also just natural cost cutting from businesses that were previously massively over hiring, and outside of AI don’t exactly have a ton of areas with huge growing investment.
Plus slashing jobs like this keeps the plebs in line. They don’t like software engineers having the money and job security to raise a stink over things. They want drones terrified of losing everything.
Is it pure AI or a guy in India that can cover the gap using AI to create good enough slop to pass?
Everyone is doubling down on hiring IN India right now. H1B isn’t even a thing. It’s offshoring to Indians that are utilizing AI to ship good enough slop. Everyone’s India office is rapidly expanding.
Now it is getting harder and harder for young people
Slow down people. Let's stop jumping to biases and see what we have here.
Note upfront: I'm not suggesting AI is not having an impact. That would be foolish. But I will say there's *a lot* less to the conclusion of this study, simply because the data is questionable. It's not that they did anything wrong per se. I won't say that here because it'll end up a HN cluster fuck. Cluster fuck aside, the caveats and associated doubt are enough to say, "Don't bet the farm on this study." Great bander for the bar? Sure.
It's an interesting study but I've seen it called "absolute proof" and other type things. Don't be fooled, it's not that.
https://digitaleconomy.stanford.edu/wp-content/uploads/2025/...
From the original study:
> "This study uses data from ADP, the largest payroll processing firm in America. The company provides payroll services for firms employing over 25 million workers in the US. We use this information to track employment changes for workers in occupations measured as more or less exposed to artificial intelligence"
a) I'm calling this out because I've seen posts on LinkedIn saying it was a sample of 25M. Nope! ADP simply does payroll for that many.
b) The size of the US workforce is ~165M, making ADP's coverage ~15% of the workforce.
https://www.statista.com/statistics/191750/civilian-labor-fo...
c) Do the business ADP server come from particular industries, are of a particular size, in particular geographic locations? etc.? It's not only about the size of the sample - which we'll get to shortly - but the nature of the companies - which we'll also get to shortly.
> "We make several sample restrictions for our main analysis sample."
d) It's great that they say this, but it should raise an eyebrow.
> "We include only workers employed by firms that use ADP’s payroll product to maintain worker earnings records. We also exclude employees classified by firms as part-time from the analysis and subset to people between the age of 18 and 70."
e) Translation: we did a slight bit of pruning (read: cherry-picking).
> "The set of firms using payroll services changes over time as companies join or leave ADP’s platform. We maintain a consistent set of firms across our main sample period by keeping only companies that have employee earnings records for each month from January 2021 through July 2025."
f) Translation: More cherry-picking.
> "In addition, ADP observes job titles for about 70% of workers in its system. We exclude workers who do not have a recorded job title."
g) Translation: More cherry-picking.
> "After these restrictions we have records on between 3.5 and 5 million workers each month for our main analysis sample, though we consider robustness to alternative analyses such as allowing for firms to enter and leave the sample."
h) 3.5M to 5.0M feels like a large enough sample... if it wasn't so "restricted." Furthermore, there's no explanation on the 1.5M delta, and how adding or removing that much impacts the analysis.
i) And they considered that why? And did what they did why? It's a significant assumpt that gets nothing more than a hand wave?
> "While the ADP data include millions of workers in each month, the distribution of firms using ADP services does not exactly match the distribution of firms across the broader US economy."
j) Translation: as mentioned above ADP !== a representation of the broader economy.
> "Further details on differences in firm composition can be found in Cajner et al. (2018) and ADP Reserch (2025)."
j) Great there's a citation, but given the acknowledgement of the delta isn't at least a line or two in order? Something about the nature of the delta, and THEN mention the citation?
k) Editorial: You might think this hand-wave is ok, but to me it's usually indicative of a tell and a smell.
l) Finally, do understand the nature of academia and null research (which has been mentioned on HN). In short, there is a (career / financial) incentive to find something novel (read: worth publishing). You advance your career by doing not-null research.
Again, I'm not suggesting anything nefarious per se. But this study is getting A LOT of attention. All things considered, more than it objectively deserves.
__Again: I'm not suggesting AI is not having an impact. That would be foolish.__
This study feels pretty weak. Software as a occupation is collapsing, but it's not due to AI. Articles and "studies" like this are just a smoke screen to keep your eye off the ball.
Why is it collapsing?
As I see it, it's really the lack of "capitalists" willpower to be actually capitalist.
We can't call it incompetence because neither those whom we have come to know as capitalists nor their advisors are incompetent, which means they quite literally do not want to offset any decline in jobs or (job creation) that can be linked to progress.
That's not strange. A "capitalist" wants market participation to grow, infinitely, which is possible. Who we came to know as capitalists don't care about the markets, actual market growth or market participation. They only care about the growth of the value of the markets, "however" that happens.
I highly recommend that journalists and economists dig a bit more radically honest into the matter. There'd be more value in that, more blog posts, more articles, more discussions on all platforms, and thus more participation.
I mean it's a scapegoat vs straw man vs actual culprit kind of situation ... isn't it?
Short-term, discrete numbers like these are interesting to look at, but they don't really tell us much about the long-term trajectory. In parallel: [1].
[1] "Nvidia Forecasts Decelerating Growth After Two-Year AI Boom" <https://news.ycombinator.com/item?id=45053175>
yet another clickbait "ai is taking jobs" study that doesn't investigate whether or not the employment decrease is directly caused by the ai adoption
[dupe] Discussion: https://news.ycombinator.com/item?id=45025978
And a better source article.