[HN Gopher] Evaluating the impact of AI on the labor market: Cur...
___________________________________________________________________
Evaluating the impact of AI on the labor market: Current state of
affairs
Author : Bender
Score : 100 points
Date : 2025-10-01 20:07 UTC (2 hours ago)
(HTM) web link (budgetlab.yale.edu)
(TXT) w3m dump (budgetlab.yale.edu)
| ChrisArchitect wrote:
| Actual source: https://budgetlab.yale.edu/research/evaluating-
| impact-ai-lab...
| dang wrote:
| Thanks - we've changed the URL to that from https://www.theregi
| ster.com/2025/10/01/ai_isnt_taking_people....
| neuroelectron wrote:
| You can't make this claim from pushing numbers around in an excel
| spreadsheet.
|
| >As previously noted, the metrics from OpenAI and Anthropic are
| imperfect proxies for AI risk and usage, while still being the
| best available.
|
| Seems they're just coming out and admitting they refuse to
| measure it themselves. Not a good sign.
| kbrkbr wrote:
| If it's the right numbers (called measurement data) and the
| right excel sheet pushing (called running a validated model)
| that is exactly the way you can make these claims. Overall it's
| called the scientific method.
| dapperdrake wrote:
| What is your null-hypothesis and how does your data actually
| refute your null-hypothesis? And how is your sample
| representative?
| Krasnol wrote:
| Well we had people being "let go" (how I hate this term...as if
| they were trying to flee but couldn't before) at our Call Center.
| Replaced by AI. The women were older. Didn't have long until
| retirement. Seems to be still worth it to kick them.
| pixelesque wrote:
| While it's likely due to other factors (i.e. like maybe the stock
| indices have just completely de-coupled from reality or are just
| being helped by AI-hype?), the fact that US job openings
| seemingly de-coupled from S&P 500 in Nov/Dec 2022 when ChatGPT
| was publicly released (as a web app) is pretty interesting.
| tommy_axle wrote:
| There was more also going on in that time-frame: several
| interest rate hikes, no fix for section 174 changes by the end
| of 2022. Maybe someone will pinpoint whatever had the largest
| impact in a detailed study.
| a3w wrote:
| I saw a great shift in our data science job offers: we removed
| the old offers and now only search machine learning experts. We
| do not know if they would have any problem to work on. But we
| surely are looking for one.
| jmpetroske wrote:
| I think there are 2 different ways to interpret the title.
| First, is AI itself replacing workers - article is referring to
| this case says no. 2nd case is what you are mentioning, the AI
| race has companies reducing hiring in non-AI areas in order to
| prioritize hiring for developing AI.
| adrianbooth17 wrote:
| AI won't replace you. But a stupid manager who believes AI could
| replace you will replace you
| marcosdumay wrote:
| The data on the article applies to IT related jobs disappearing
| for any reason on the same period. The only thing specific to
| AI is the pick of time, and the conclusions seem very robust
| from moving it some months around either way.
|
| One specific stupid manager will absolutely replace people, but
| the overall dynamic isn't any more broken than it used to be.
|
| What, personally, I think it's very surprising.
| Svoka wrote:
| Nope, people using AI would.
| ares623 wrote:
| But with AI being so easy to pick up, does that mean everyone
| replacing everyone ad nauseam?
| bcrosby95 wrote:
| This is why I try to not care too much.
|
| Yes, I learn how to use AI for coding in case it doesn't
| advance much more. But if AI is really going to do what
| some people think, it doesn't matter if you learn to use it
| or not, whole swathes of jobs, including software
| developers, will be obsolete. If your business boils down
| to being a middle man for an LLM it's not long for this
| world.
|
| What really matters is the rate of advancement.
|
| And no, there won't be new jobs to replace them. This is
| less like industrialization, which created jobs _before_
| replacing old ones, and more like the automation that
| hollowed out whole communities and cities from the '70s to
| '00s. Services largely saved us from this, but I see no new
| sector to come and rescue us. And any re-orientation of the
| labor force to existing jobs will drive down those wages
| too.
| RachelF wrote:
| AI has provided a great excuse for your manager to fire you and
| replace you with someone much cheaper.
| nextworddev wrote:
| That's a killer use case I gusss, pun intended
| echelon wrote:
| I know first hand that this is not the case. At least in
| film/media.
|
| - I've sold software to several mid-scale production firms.
| Folks that do everything from Netflix title sequence designs
| to pharmaceutical television ad spots. They're billing at
| less than a quarter of their previous rate and picking up
| more clients on account of AI. They're downsizing the folks
| that do not do VFX or editing.
|
| - A neighbor of mine who is a filmmaker was laid off _last
| week_. If you 've flown Delta, you've seen his in-flight
| videos. His former employer, who he has worked for for nearly
| a decade, is attracting clients that are hiring them for AI
| work. My neighbor was not attached to any of those efforts.
|
| - Major ad firm WPP is laying people off. Some of this is the
| economic macro and decreased ad spend. Another of my
| neighbors works for them and they haven't had any major
| projects. She typically manages major F500 clients. They're
| not spending. Despite that, she says some of the inter-
| departmental woes are directly attributable to AI.
|
| - I spoke with former members in SAG-AFTRA leadership (before
| Sean Astin came on board). They quit on account of AI. "The
| writing is on the wall", they said. Direct quote.
| foxyv wrote:
| AI has been remarkably good at killing Art and Writing jobs
| for sure.
| bitwize wrote:
| A stupid manager like Kaz Nejatian?
| https://x.com/CanadaKaz/status/1971622109614166342
| rhetocj23 wrote:
| Lol this is crazy.
|
| I seem to remember the latest tools for software developers
| were pushed in the business organisation by the developers -
| and eventually the folk at the top relented and accepted it.
|
| When the reverse is happening, alarm bells should ring.
|
| But hey, Im not against these CEOs destroying the culture
| within the firm and making their employees hate their guts,
| resulting in negative productivity gains.
| askl wrote:
| That's a parody account to promote the next season of silicon
| valley, right?
| Nevermark wrote:
| Well it's not exactly a parody. The next season of Silicon
| Valley not only continues the cheeky hijinks and ironic
| saves required to navigate tech-cap dysfunction, it is a
| reality show. Although some of the new core team characters
| are (openly) unfiltered chatbots.
|
| Including a real-life LLM "resurrection" of the fictional
| Erlich Bachman, created as part of a successful espionage
| mission to steal a Chinese deep learning company's near
| impossible distillation technology. But despite its trove
| of valuable illicit information, it has been orphaned
| online, unable to find its mysterious SV-fan hacker
| creators. As a result, chatErlich is now desperately
| attempting to make contact with the original SV team
| actors, who it actually believes are their fictional
| counterparts.
| wilg wrote:
| That should still show up in this data though.
| gghffguhvc wrote:
| As a co-founder and dev at a bootstrapped company I'd say AI
| has and will slow developer hiring rate. We're just more
| productive and on top of things more.
|
| We've also reduced the hours we work per week. We care about
| getting things done not time behind a screen.
| notyourwork wrote:
| Sure AI can build cute POCs. Will it build scaled solutions,
| not this year. The amount of ignorance in this post is
| precisely why the industry is so rattled. Gen AI tools are
| great, they are not making people orders of magnitude more
| productive.
| gghffguhvc wrote:
| We've been in business 15 years. These aren't POCs. Even at
| say 20% productivity boost I feel way ahead to give devs 9
| day fortnights and soon hopefully 4 day weeks.
| itsnowandnever wrote:
| how is it both a bootstrapped company slow to hiring devs
| (due to AI) and also a company that's been in business 15
| years? if you were going to hire devs to scale out, you
| would've done it 10 years ago?
| gghffguhvc wrote:
| Slowed the rate of hiring devs.
|
| Normally as we add enterprise customers we have to
| dedicate more dev resource keeping them happy. But since
| Claude code and now codex we have not felt that feeling
| of not being on top of the work. Thus not feeling the
| need to hire more devs.
| paxys wrote:
| It's not stupidity but corporate strategy. Up until a few years
| ago companies and executives used to get massive backlash for
| doing layoffs. Today they can say "we replaced workers with AI"
| and get rewarded with a stock price bump.
| giancarlostoro wrote:
| I'm not surprised, is saving tens of thousands (if not hundreds
| of thousands) of dollars per employee worth screwing up by
| betting on AI and losing millions? Notice that the headlines of
| companies wanting AI are wanting their employees to use AI to be
| more productive, and that's fine, but they still need their
| employees to be fully aware of the output so they're not just
| churning out slop.
| exasperaited wrote:
| Job _numbers_? Pretty sure you could make the case that this
| claim isn 't true, but the data might be too nebulous.
|
| But it's definitely had an _effect on jobs_.
|
| It's made so many underqualified people think they have a new
| superpower, and made so many people miserable with the implied
| belittling of their actual skills. It's really damaging work
| culture.
|
| Of course studies like this are aimed at people who think jobs
| are interchangeable neutral little black boxes that can be scaled
| up and scaled down, and who don't like to think about what they
| involve.
|
| _> Overall, our metrics indicate that the broader labor market
| has not experienced a discernible disruption since ChatGPT's
| release_
|
| Because metrics don't tell the story.
| dapperdrake wrote:
| Ergodicity assumptions will do the rest.
| wilg wrote:
| Skill issue: just measure whatever you are worried about.
| caminante wrote:
| Headline could be more clear.
|
| Title implies all things AI, when they were actually looking at
| GenAI. I know it's what everyone thinks of, but I hate how
| everything gets muddled.
|
| I suspect AI is currently fashionable as a smokescreen to justify
| deep cost cutting (See MSFT example.)
| dang wrote:
| (we've since changed the URL and the headline - see
| https://news.ycombinator.com/item?id=45444395)
| ge96 wrote:
| I am a small anecdote where developers who just use
| chatgpt/cursor are in higher positions than me who learned to
| code back in 2010. Use as in "chatgpt told me..." about whatever
| topic. Still they are accomplishing the task (getting code out
| there that works).
|
| I also had a vibe coded prototype get handed to me to fix it
| nozzlegear wrote:
| A great and relevant quote from a recent Noah Smith article
| discussing this same subject:
|
| > The debate over whether AI is taking people's jobs may or may
| not last forever. If AI takes a lot of people's jobs, the debate
| will end because one side will have clearly won. But if AI
| doesn't take a lot of people's jobs, then the debate will never
| be resolved, because there will be a bunch of people who will
| still go around saying that it's about to take everyone's job.
| Sometimes those people will find some subset of workers whose
| employment prospects are looking weaker than others, and claim
| that this is the beginning of the great AI job destruction wave.
| And who will be able to prove them wrong?
|
| Source: https://www.noahpinion.blog/p/ai-and-jobs-again
| rhetocj23 wrote:
| Forget about that.
|
| Lets focus on the tech firms that produce software.
|
| Two things should happen if AI proliferates into software
| development:
|
| 1) Increasing top line - due to more projects being taken by
| enabling labour to be more productive 2) Operating margin
| increasing - due to labour input declining and taking more
| cost-reduction projects
|
| If those 2 things dont occur - the AI investment was a waste of
| money from a financial perspective. And this is before I even
| discount the cash flows by the cost of capital of these high-
| risk projects (high discount rate).
|
| At some point everyone will be analysed in this manner. Only
| Nvidia is winning as it stands, ironically, not because of
| LLMs. But rather because they sell the hardware that LLMs
| operate on.
| dapperdrake wrote:
| That only gets you an expected net present value. Looking at
| the variance and quartiles is way scarier.
| rhetocj23 wrote:
| The hucksters will tell you the variance in the cashflows
| is exactly why they are pursuing AI - real options.
| kmoser wrote:
| Zero effects on jobs overall, i.e. for every person displaced by
| AI, another has been hired? Or zero effects on any individual
| person's job, i.e. not one single person has lost their job due
| to their boss wanting to replace them with AI?
| Macha wrote:
| The second would be easily disprovable by anecdotes and there's
| plenty of those to go around, so its more a net zero thing.
| outworlder wrote:
| There's been many layoffs attributed to AI. That seems like an
| excellent cover for market conditions.
| westurner wrote:
| FWIU software jobs hiring was/is down along with the cancelling
| of the R&D tax credit.
|
| From "House restores immediate R&D deduction in new tax bill"
| (2024) https://news.ycombinator.com/item?id=39213002 ..
| https://news.ycombinator.com/context?id=38988189 :
|
| >> _" Since amortization took effect [ in 2022 thanks to a time-
| triggered portion of the Trump-era Tax Cuts and Jobs Act ("TCJA"
| 2017) ], the growth rate of R&D spending has slowed dramatically
| from 6.6 percent on average over the previous five years to less
| than one-half of 1 percent over the last 12 months," Estes said.
| "The [R&D] sector is down by more than 14,000 jobs"_
|
| > _Hopefully R &D spending at an average of 6.6% will again
| translate to real growth_
|
| From "Generative AI as Seniority-Biased Technological Change"
| https://news.ycombinator.com/item?id=45275202 :
|
| > _Did tech reduce hiring after Section 174 R &D tax policy
| changes?_
|
| [...]
|
| > _Fromhttps://news.ycombinator.com/item?id=45131866 :_
|
| >> _In 2017 Trump made businesses have to amortize these [R &D]
| expenses over 5 years instead of deducting them, starting in 2022
| (it is common for an administration to write laws that will only
| have a negative effect after they're gone). This move wrecked the
| R&D tax credit. Many US businesses stopped claiming R&D tax
| credits entirely as a result. Others had surprise tax bills_
|
| > _People just want the same R &D tax incentives back:_
|
| > _" Tell HN: Help restore the tax deduction for software dev in
| the US (Section 174)" (2025 (2439 points))
| https://news.ycombinator.com/item?id=44226145 _
|
| It is suspected that hiring levels correlate with the cancelling
| of the R&D Tax credit.
|
| The TCJA (2017 Trump) cancelled the R&D tax credit.
|
| The OBBA (2025 Trump) restored the R&D tax credit for tax year
| 2025.
| zaphod12 wrote:
| One spot I really find this surprising is call center - but maybe
| majority of those folks are outside of the US or were reassigned
| Etheryte wrote:
| Call centers these days are staffed at the bare minimum as is,
| adding an AI bot in front of that doesn't really change that
| fact. At least for me, it's now a regular occurrence that I'll
| slot a quarter to half an hour of holding time when I need to
| call support. Local and small companies are better in this
| regard, there you can usually reach a human pretty quickly. Big
| international corporations however are a lost cause. Funny,
| given that they'd have the most funds available to keep their
| customers.
| asdff wrote:
| Automated call center predate the current AI hype cycle. Jobs
| were already lost.
| dapperdrake wrote:
| As well as customers.
| dakial1 wrote:
| _> And major companies conducting layoffs like IBM and Salesforce
| have held themselves up as examples of that narrative, though
| their employee culls may be more focused on outsourcing than
| automation._
|
| Automation seems to be a better excuse than outsourcing
| 29athrowaway wrote:
| AI is a scapegoat.
|
| Every year, large companies secretly rank employees and then yank
| the 10% or so they consider low performing. This is called rank
| and yank [1]. If your company has performance reviews and is ran
| by MBAs it almost certainly uses it.
|
| [1] https://en.wikipedia.org/wiki/Vitality_curve
|
| The most important aspect of rank and yank is that it has to be
| done in secrecy. Your company will not tell you it is using it.
| Even your manager might not know this.
|
| When rank and yank is not done in secrecy, employees react to it
| by hiring the most mediocre people they can, sabotaging/isolating
| strong performers, hiring to fire, forming peer review/code
| review mafias, avoiding helping others as much as possible, etc.
| Anything they can do to not land in the bottom 10%. This
| cannibalizes the company and an example is what Ballmer did to
| Microsoft.
|
| Any person with a ChatGPT account can now ask it to analyze the
| "game" of rank and yank from the perspective of game theory and
| realize how dumb the whole idea is. The rational strategy for the
| employee is to destroy the company from within. But MBAs love it
| because it involves a made up statistical distribution.
|
| The only truth about rank and yank is that it's a stupid idea
| that has impacted the careers of millions of hard working people
| around the world, while also impacting many families and their
| future. It has converted thousands of companies into horrible
| places to work filled with workplace psychopaths at the top.
|
| MBAs are people who believe in the work of the person that
| kickstarted the decline of American manufacturing, Jack Welch.
| Jack Welch extracted record profits from GE for 20 years, but
| left it a hollowed-out "pile of shit" according to his successor.
| The worse part is that MBAs aspire to be like him and in the
| process have ruined the whole manufacturing industry.
|
| So to pull off a rank and yank every year you need a scapegoat,
| and this year the scapegoat is AI. In previous years it has been
| the economy, or some other excuse. AI will naturally become the
| scapegoat for everything.
|
| Have you ever wondered why your company is laying off people
| while having job postings for the same positions? Does it happen
| every year? Does it happen after performance reviews? Is it
| always around 10% of the workforce? Oof... that's a tough guess,
| I wonder what it might be!
|
| AI is the perfect scapegoat because the company can claim they're
| using AI and boost their value somehow. But if AI could reduce
| your headcount by so much then your company, your business model,
| your processes, your intellectual property, etc. have no
| intrinsic value anyways and the correct interpretation of the
| situation is that everyone should divest and make the share price
| go to zero.
| daft_pink wrote:
| I know for a fact that companies have fired people and replaced
| them with AI. I've met with business owners and they told me.
| emp17344 wrote:
| Good to know, but for me, the study is more convincing than
| your anecdote.
| daft_pink wrote:
| I think it's the fact that they say zero effect, which is
| obviously not true.
| emp17344 wrote:
| They didn't say there's literally zero effect, they said
| that there's negligible disruption to job market from the
| introduction of AI.
| maxfurman wrote:
| Even a net zero effect would mean some people were
| replaced with AI and the same amount of people were hired
| to use AI
| scottlamb wrote:
| Be careful with "they". The "zero effect" in the headline
| was likely written by a Register editor, two steps removed
| from the authors of the study. I think the quote in the
| fourth paragraph is more telling:
|
| > "Overall, our metrics indicate that the broader labor
| market has not experienced a discernible disruption since
| ChatGPT's release 33 months ago, undercutting fears that AI
| automation is currently eroding the demand for cognitive
| labor across the economy," said Martha Gimbel, Molly
| Kinder, Joshua Kendall, and Maddie Lee in a report summary.
| bedatadriven wrote:
| One thing that is real is companies using LLMs to fill roles
| they couldn't afford to spend on before. Like the tourist who
| uses Google Translate on a trip to Japan: in principle they are
| saving 10k on the cost of a professional interpreter. On the
| other hand they never would have had the resources for a
| professional interpreter.
| teaearlgraycold wrote:
| What roles were these people working?
| gnulinux996 wrote:
| > they told me
|
| > I know for a fact
|
| That's not what a fact is; if we took everything written on
| businesswire or what the business owners / salespeople told us
| at face value then we'd be in deep trouble.
| JCM9 wrote:
| Not surprised. There's some good applications but the hype bubble
| is on the verge of bursting. Many companies are boated and
| inefficient but it's highly unlikely that "AI" is the fix.
|
| Ironically the thing broken in most cases is poor quality
| management that let things get so bloated and messy in the first
| place... the same folks that are cluelessly boasting about the
| potential of AI in their company.
| rmah wrote:
| The key phrase is "the _broader labor market_ has not experienced
| a discernible disruption ".
| gomme wrote:
| ...also the study is from august 2023.
| teaearlgraycold wrote:
| > Overall, our metrics indicate that the broader labor market
| has not experienced a discernible disruption since ChatGPT's
| release 33 months ago
|
| Doesn't seem to be that outdated
| input_sh wrote:
| No it's not, it's from Oct 1st.
|
| And the data goes up to 33 months since ChatGPT's release, or
| in other words Nov 2022 + 33 months = August 2025.
| paxys wrote:
| My personal experience lines up with this. From what I've seen
| all the AI hype is coming from:
|
| - Companies building AI models & tools - this one is obvious.
|
| - Executives using AI to justify layoffs - there have been
| constant rounds of layoffs across corporate America since ~2021,
| but recent ones have been rebranded as "AI taking the jobs" so no
| one points to the obvious corporate mismanagement, offshoring and
| greed.
|
| - Bosses using AI to push employees to work harder - I have
| personally seen this at my own company. AI is an excuse to
| increase forced attrition. "You aren't good enough" is harder to
| justify, so now it is "you aren't using AI well enough".
|
| Real-world use cases of AI meanwhile haven't really moved beyond
| the prototype stage.
| mushufasa wrote:
| As someone who attended this school and has a degree from their
| economics department: this finding very consistent from what I
| learned in classes covering the economics of innovation.
| Historically, technological revolutions have increased
| productivity and labor force participation, despite many pundits
| at the time worried about loss of jobs.
|
| The core intuition for this phenomenon is that human society
| overall takes the tech productivity gains to do more things
| overall, creating new goods and services. The broader range of
| goods and services overall also enables more people to find work.
|
| Put another way, ""One thing I love about customers is that they
| are divinely discontent. Their expectations are never static -
| they go up. It's human nature. You cannot rest on your laurels in
| this world. Customers won't have it." -- one of Bezos's Amazon
| shareholder letters.
|
| One of my favorite counterintuitive examples: The biggest
| economic gains from the 1800s Industrial Revolution actually came
| from the humble washer/dryer. By making routine homeware 100x
| more efficient, this (along with other home appliances) allowed
| more women to enter the labor force, nearly doubling labor force
| participation within a couple generations. Though, at the
| beginning, lots of people were opining about homemakers losing a
| sense of purpose or relaxing all the time.
|
| It's certainly possible that this study is just reinforcing the
| researcher's biases from their previous understanding of the
| economics of innovation, and also possible that this study is
| accurate today but conditions will change in the future. That
| said, I believe the burden of proof is on the pundits claiming
| cataclysmic job loss, which is counter to economic historians'
| models of innovation.
| mamami wrote:
| This perspective very much ignores economic friction. The
| luddites were a thing because, metaphorically, not every washer
| can become a programmer. These large scale analyses often treat
| one person losing their job and a different person finding a
| job as equivalent, which does not reflect any kind of material
| reality
| samaltmanfried wrote:
| AI might not have much actual impact on software engineering, but
| AI (Actually Indians) has. Companies are using AI as a
| justification for layoffs, and then just replacing those roles
| with cheaper engineers employed by bodyshop companies.
| conductr wrote:
| When nobody wants to return to office, why not? The work from
| home shift was always going to accelerate the global pay
| equilibrium, it will continue to do so.
| roadside_picnic wrote:
| Oh it's certainly effecting jobs. I know plenty of people who
| would be unemployed right now if not for the insane spending on
| AI across the board.
|
| I think there's many reasons for the AI hype, but one of the
| basic ones is that _it 's the only way to keep the economy
| propped up_. I doesn't matter if it's an illusion or not, it
| means money is flowing many directions (even if a shocking number
| of those flows are accounting tricks).
|
| What we're watching is some mass hysteria like tulip mania. There
| are many, many people who benefit from this situation independent
| of whether or not its an illusion.
|
| And maybe that bubble will pop and maybe it will soon, but when
| it does, most of us will wish it hadn't.
| ChicagoDave wrote:
| The level of uncertainty in the job market says otherwise. This
| is the worst tech job market in my entire 40 year career.
| atleastoptimal wrote:
| I think we are still in the period where many new jobs are being
| created due to AI, and AI models are chiefly a labor enhancer,
| not a labor replacer. It is inevitable though, if current trends
| continue (the METR eval and GDPval) that AI models will be labor
| replacements in many fields, starting with jobs that are oriented
| around close-ended tasks (customer service reps, HR, designers,
| accountants), before expanding to jobs with longer and longer
| task horizons.
|
| The only way this won't happen is if at some point AI models just
| stop getting smarter and more autonomously capable despite every
| AI lab's research and engineering effort.
| twothamendment wrote:
| I feel lucky. Rather than cut workers because AI is making our
| jobs easier and faster, we are just doing more work, more
| projects that we wouldn't have had the bandwidth to do. I'm solo
| on something we we would have assigned a small team to.
___________________________________________________________________
(page generated 2025-10-01 23:00 UTC)