Warn all you want it is here and will continue to grow fast and deep. Quit focusing on the "I told you so" and Monday morning quarter backing, adapt or die.
I will say I work in risk management and with it I’m literally doing software engineer work. Not necessarily by choice my boss asks me to. But like I just ask it to program stuff we need and it does it perfectly even after testing each time so far. Why pay one of them when you can just have an employee like me making 110-120K do risk management + any coding needs just ask AI? I mean sure we always need some for more complex stuff but less for the small stuff
Not sure why the headline is “over 200 economists”. A bulk of the signatories are either AI researchers, or those working in ancillary functions within AI labs.
Just to give you an idea, I am a SWE with only 6 YOE.
I have recruiters CONSTANTLY hitting me up for AI startups that are working to disrupt medical billing, civil engineering, legal services, etc. All work that was traditionally done by humans for middle class wages.
We have yet to see the seismic impact this is going to have on our economy and our politics.
But what about the metaverse stuff about 5 years ago or the air taxis and public transit start ups about 10 years ago? Or maybe the social media and Facebook centric start ups 15 years before that?
It’s the latest bullshit hot trend that will eventually be replaced with whatever’s next
Today I asked copilot to take some raw data I had. It was roughly 150 rows with 6 columns - each column reprinting data from a specific date.. I created categories. Based on key words in the rows of data, I wanted copilot to group the data together (the example I gave was “data cleansing” group if row contains “data”. ) I could have done this in excel but I was trying to do this “quicker” and it seemed straight forward enough… the table I got was incredibly inaccurate, I just did it in excel in the end. Or today, when I asked it to check for files with the following file names in a large folder. It came back and said only 3 files were missing. But after spot checking, I found 10. When I asked it said it was because it had a limit on how many files in the folder it could query. I have no fucking idea how you can sit back and let ai do large data cleansing activities. Every time I’ve done basic tasks I run into accuracy issues it’s insane.
So I guess long story long was metaverse was to create some shitty augmented reality and ai is to create an assistant that can not be trusted to do the most menial tasks
AI is really bad at this sort of stuff. Instead you should have asked it to write a script to do it and output the data. Pretty sure you can run scripts in excel, you could certainly do it in a few lines of SQL. I'm honestly surprised it isn't already built into excel where it will execute it for you.
Anything deterministic is much better solved with regular software, data access and manipulation was solved by software decades ago (at least at this scale). Like don't ask it to count something, it will still fuck it up.
The tools around an LLM will get better. Like Claude code, i can ask it to rename 1000 files and it won't read any of them, it will just write a little script and execute it and the script will do the actual work and with all the accuracy of deterministic software.
That said, given it's demonstrated it's good at doing my job (writing code), I'm not concerned about it replacing me because the bar just keeps going higher in terms of the value I provide. The companies that see the productivity gains as a means of cutting labour will lose out to the ones who see the productivity gains as a means of getting ahead.
My only concern is that some knucklehead above me sees it the same way.
No, I just work with companies that struggle to implement already existing software advances. Ai isn’t as accurate or straight forward as it’s suggested and it can’t be shoved into every product and space. It maybe has its place but it’s not the magic bullet that the c suite under ai psychosis thinks it is
"AI may become radically more powerful over the next 10 years," reads the first bullet point of the open letter."
Ah, so we’ll have this ostensible more powerful AI right around the time cold fusion is invented.
Look, AI is LLMs. That’s all it is. And you get diminishing returns with LLMs, especially when they’re training on data produced by other LLMs (or even by themselves). That’s just how the math works, sorry. We’ve seen the amazing increase in LLM capability already. Now it’s slowing down.
AI still has an unsolved issue of being profitable. Also, from programmers prospective new models seem to be getting more annoying. At the moment where I work all devs are forced to use AI but it does not make work faster but company spends more money.
Ive been building agents at my work. Already we're optimising token usage, like when to reach into the LLM and when to fallback to regular old deterministic software and optimising the data it receives and retains. It's work for me as a dev but the money saved is a good use of my time. If they increase prices that will only encourage that optimisation process and it should mean we can get away with less powerful models too because it doesn't need as much context and isn't thinking as hard.
I can't see a way to profitability besides cutting expenses. I can see there being a flood of second hand hardware in the not so distant future, much like Bitcoin mining rigs.
I dont see this take in many places, but I completely agree. There are too many hard caps: hardware shortages are slowing down production of GPU racks; shortages of things like transformers and community pushback are slowing down datacenter construction; antsy investors are starting to get nervous about the profit potential so capital shortage might slow down progress, etc.
Yeah, kind of funny that the media is not talking about this. If you look at the graph of requisitions for data centers in energy capacity, you can plainly see that the timelines being thrown around are laughable. Looking at that graph makes you feel like this is a Big Short moment v2.
The issue with economics is that they come in too late. This is just telling me the AI bubble is popping soon as usage limits, and calls against it keep coming up.
It feels like AI firms set themselves up for this one; this is now the infinite-excuse glitch that politicians at all levels across many G7 nations can point to every time there's a job issue; they can say AI is stealing their jobs. So both become an inevitable job-efficiency initiative and an absolution from Port's economic planning.
Once the pitchforks are out, I see all of the AI leaders moving to New Zealand if they'll have them.
Less productivity means less money flowing through the economy means less money for the companies trying to make more money with less workers. The economics of implosion.
Has it or is it just repeating its endless cycle again?
We are making the same mistakes that the colonial empires did...
Make so much bank that you think hey how about some slavery, colonialism or neocolonialism whilst they would have made even more bank if they didn't engage in any of that stupid shit.
Just trade without being extra. All that extra baggage comes with other extra baggage that will mess with trade eventually.
Maybe AI agents will get their own bank accounts and start spending. This will create an evolutionary leap forward for them and then they will expand and build their own infrastructure, and then eventually create pods for harvesting energy from sleeping humans who have become completely unemployable inside of the AGI-lead world.
Stopped clock is right twice a day. Half the shit an economist will yell you is "good" is like... the most cost-effective method to euthanize the poor.
I dont see why not? Accounting/bookeeping seems like a good use of ai, its already in big HR firms. Something else to consider is the jobs lost by people refusing to do shady shit for ai. Like a new thing in home service companies is to push the ai record button in the service app, then it records the entire conversation with the customer and provides feedback on improving sales numbers. Some people are quitting rather than doing it cuz it's shady and wrong.
The economy is cratering. The cuts were coming anyway as we start massive decline. AI is the scapegoat, because if you say you're cutting for any reason but AI, the stock craters.
It's both. The tech sector is a bloodbath because you don't need a Computer Science degree or be able to write in complete sentences to write code anymore. They just hire warm bodies to type into Copilot. The economy is also cratering.
Do you know what "wishcasting" is? I'm not sure if that's what I would call this, but this almost religious mantra of trying to pretend that AI does not and will not have any impact on the economy and our industries - but how many incredibly smart people saying this, warning about this, does it take before people really take it seriously?
On Reddit I hear a lot about how people need to listen to scientists and researchers and generally smart people more.
I guess if you work as a software dev and have used the latest models, especially if you've been tracking the trajectory over the last 6 months you have a different take away - at least I see devs finally taking this seriously en masse. What would it take for you? What do you need to see?
AI is absolutely changing many industries forever in ways we’re a long way from understanding.
And there are businesses using AI to justify layoffs right now that actually have to do with unrelated business and economic fundamentals.
Right now I suspect we’re seeing more of the latter than the former, because it’s still extremely new and still hasn’t fully integrated into business operations. As it does there will be more churn and replacement, but even then it’s unclear what it will ultimately do to the workforce, the same way we didn’t know what the internet would do.
Here's where I'll kind of agree - AI is helping to make clear that lots of jobs are likely already unnecessary, overhires, and fundamentally changing the incentives of overhiring.
Who is being hired the most though - are people who can come into orgs, use AI, and automate as many tasks as possible. This is starting and will continue and ramp up.
None of this disagrees with what the economists are saying - they are not saying... All jobs that have gone in the last little while are because of AI? No - they are saying over the next 10 years (the range of time varies for all of them, but 10 years was agreeable to all of them) - this is likely to ramp up.
I am responding to the person who says "No" out of reflex, and if I will be bold, fear and discomfort.
Thank you! The number of people I explain this to on a regular basis. It really is another great example of how people are fed misinformation through so many avenues of media.
Stop trying to game the stock market then. Find some sort of large index fund like something tied to the S&P 500 or Vanguard’s VTI, put your money in it, and forget about it
Trying to gamble on stocks without knowing about them is a great way to lose money. Safe investments is a great way to make money over time
There was a test done many years ago where they had a monkey pick stocks in some weird way. The monkey beat the majority of so called experts. I'd likely have to dig to find it.
That said, I put everything into gold on January 17, 2025, because history shows that fascists blow up the economy sooner or later.
How about we don't
ReplyDeleteWarn all you want it is here and will continue to grow fast and deep. Quit focusing on the "I told you so" and Monday morning quarter backing, adapt or die.
ReplyDeleteand he finally wakes up
ReplyDeleteWe don't need 200 hundred economists, replace them with EconoAI. /s
ReplyDeleteSo we finally get the AI companies themselves to shut up with this slop, and the media just . . . recruits someone else.
ReplyDeleteDemand will always find a supply, it seems.
I will say I work in risk management and with it I’m literally doing software engineer work. Not necessarily by choice my boss asks me to. But like I just ask it to program stuff we need and it does it perfectly even after testing each time so far. Why pay one of them when you can just have an employee like me making 110-120K do risk management + any coding needs just ask AI? I mean sure we always need some for more complex stuff but less for the small stuff
ReplyDeleteWe need to fight for better safety nets because AI is not going away
ReplyDeleteHaven't they been saying that for three years now?
ReplyDeleteLLM layoffs are just a pretext.
ReplyDeleteIt's the layoffs themselves which cause issues to the economy.
And yet oddly I bet none of them are warning of job losses for economists.
ReplyDelete"Due to financial mismanagement that will be blamed on ai"
ReplyDeleteThe same economists who bet for that exact same outcome in stocks and options? Damn what a coincidence! 🤡
ReplyDeletethis is a desperate attempt to reign in workers in the face of a failing ai bubble. ridiculous
ReplyDeleteYeah, it's coming.
ReplyDeleteThen everything will change
People without jobs don't buy things companies make.
ReplyDeleteEconomists are idiots it would seem. They should be using AI to do economy work.
ReplyDeleteNot sure why the headline is “over 200 economists”. A bulk of the signatories are either AI researchers, or those working in ancillary functions within AI labs.
DeleteJust to give you an idea, I am a SWE with only 6 YOE.
ReplyDeleteI have recruiters CONSTANTLY hitting me up for AI startups that are working to disrupt medical billing, civil engineering, legal services, etc. All work that was traditionally done by humans for middle class wages.
We have yet to see the seismic impact this is going to have on our economy and our politics.
Using LLMs to do any of these things sounds like professional negligence.
DeleteNot disagreeing that it will be disruptive, but I read a statistic that at least 75% of VC start up fail (could be closer to 90%).
DeleteThe amount of money flowing to AI start ups is booming now, but it’s not really an indicator of anything but euphoria. Some will be successful though.
But what about the metaverse stuff about 5 years ago or the air taxis and public transit start ups about 10 years ago? Or maybe the social media and Facebook centric start ups 15 years before that?
DeleteIt’s the latest bullshit hot trend that will eventually be replaced with whatever’s next
You're conflating completely different things. Put it this way:
DeleteWhat was the purpose of Metaverse versus the purpose of automation/AI?
Today I asked copilot to take some raw data I had. It was roughly 150 rows with 6 columns - each column reprinting data from a specific date.. I created categories. Based on key words in the rows of data, I wanted copilot to group the data together (the example I gave was “data cleansing” group if row contains “data”. ) I could have done this in excel but I was trying to do this “quicker” and it seemed straight forward enough… the table I got was incredibly inaccurate, I just did it in excel in the end.
DeleteOr today, when I asked it to check for files with the following file names in a large folder. It came back and said only 3 files were missing. But after spot checking, I found 10. When I asked it said it was because it had a limit on how many files in the folder it could query. I have no fucking idea how you can sit back and let ai do large data cleansing activities. Every time I’ve done basic tasks I run into accuracy issues it’s insane.
So I guess long story long was metaverse was to create some shitty augmented reality and ai is to create an assistant that can not be trusted to do the most menial tasks
AI is really bad at this sort of stuff. Instead you should have asked it to write a script to do it and output the data. Pretty sure you can run scripts in excel, you could certainly do it in a few lines of SQL. I'm honestly surprised it isn't already built into excel where it will execute it for you.
DeleteAnything deterministic is much better solved with regular software, data access and manipulation was solved by software decades ago (at least at this scale). Like don't ask it to count something, it will still fuck it up.
The tools around an LLM will get better. Like Claude code, i can ask it to rename 1000 files and it won't read any of them, it will just write a little script and execute it and the script will do the actual work and with all the accuracy of deterministic software.
That said, given it's demonstrated it's good at doing my job (writing code), I'm not concerned about it replacing me because the bar just keeps going higher in terms of the value I provide. The companies that see the productivity gains as a means of cutting labour will lose out to the ones who see the productivity gains as a means of getting ahead.
My only concern is that some knucklehead above me sees it the same way.
We can have a serious discussion, or you can double down on trying to be the class smartass who can't give a straight answer.
DeleteJust make it clear now so we can both avoid wasting time on this.
I can’t believe this is a serious comment. AI is going to disrupt the world and while it has current limitations, acting like it’s a fad is naive.
DeleteNo, I just work with companies that struggle to implement already existing software advances. Ai isn’t as accurate or straight forward as it’s suggested and it can’t be shoved into every product and space. It maybe has its place but it’s not the magic bullet that the c suite under ai psychosis thinks it is
DeleteYes and mortgage defaults too!
ReplyDelete"AI may become radically more powerful over the next 10 years," reads the first bullet point of the open letter."
ReplyDeleteAh, so we’ll have this ostensible more powerful AI right around the time cold fusion is invented.
Look, AI is LLMs. That’s all it is. And you get diminishing returns with LLMs, especially when they’re training on data produced by other LLMs (or even by themselves). That’s just how the math works, sorry. We’ve seen the amazing increase in LLM capability already. Now it’s slowing down.
AI still has an unsolved issue of being profitable. Also, from programmers prospective new models seem to be getting more annoying. At the moment where I work all devs are forced to use AI but it does not make work faster but company spends more money.
DeleteIve been building agents at my work. Already we're optimising token usage, like when to reach into the LLM and when to fallback to regular old deterministic software and optimising the data it receives and retains. It's work for me as a dev but the money saved is a good use of my time. If they increase prices that will only encourage that optimisation process and it should mean we can get away with less powerful models too because it doesn't need as much context and isn't thinking as hard.
DeleteI can't see a way to profitability besides cutting expenses. I can see there being a flood of second hand hardware in the not so distant future, much like Bitcoin mining rigs.
I dont see this take in many places, but I completely agree. There are too many hard caps: hardware shortages are slowing down production of GPU racks; shortages of things like transformers and community pushback are slowing down datacenter construction; antsy investors are starting to get nervous about the profit potential so capital shortage might slow down progress, etc.
Deletelol you know how many actual data centers we need to power even 10% of the population as employees? Quite a ton.
ReplyDeleteYeah, kind of funny that the media is not talking about this. If you look at the graph of requisitions for data centers in energy capacity, you can plainly see that the timelines being thrown around are laughable. Looking at that graph makes you feel like this is a Big Short moment v2.
DeleteWonder if they’ve figured out that frontier models can replace economists too?
ReplyDeleteI think I’m gonna be one, and I don’t care because it’s me working hard for someone else’s son get to fuck models on a yacht on his holidays anyway.
ReplyDeleteThe issue with economics is that they come in too late. This is just telling me the AI bubble is popping soon as usage limits, and calls against it keep coming up.
ReplyDeleteIt feels like AI firms set themselves up for this one; this is now the infinite-excuse glitch that politicians at all levels across many G7 nations can point to every time there's a job issue; they can say AI is stealing their jobs. So both become an inevitable job-efficiency initiative and an absolution from Port's economic planning.
ReplyDeleteOnce the pitchforks are out, I see all of the AI leaders moving to New Zealand if they'll have them.
Lets Hope they get eaten by the Uruks if they go there .
DeleteEconomics is a silly profession, too many BS predictions irresponsibly made over the years
ReplyDeleteTraders are the real economists
DeleteLess productivity means less money flowing through the economy means less money for the companies trying to make more money with less workers. The economics of implosion.
ReplyDeleteVCs pass money from one portfolio company to the other to handle this. Many tech B2B will be fine for a while
DeleteWhy do you think these companies are trading deals with each other. Endless money glitch…
DeleteIt’s unsustainable
DeleteExcept it isn't
Delete(finger guns)
capitalism has finally ate its own tail and finalized the collapse state of itself.
DeleteHas it or is it just repeating its endless cycle again?
DeleteWe are making the same mistakes that the colonial empires did...
Make so much bank that you think hey how about some slavery, colonialism or neocolonialism whilst they would have made even more bank if they didn't engage in any of that stupid shit.
Just trade without being extra. All that extra baggage comes with other extra baggage that will mess with trade eventually.
They ended up eating themselves out of empire
Maybe AI agents will get their own bank accounts and start spending. This will create an evolutionary leap forward for them and then they will expand and build their own infrastructure, and then eventually create pods for harvesting energy from sleeping humans who have become completely unemployable inside of the AGI-lead world.
DeleteFollow the white rabbit.
DeleteThey are already here tbh.
ReplyDeleteStopped clock is right twice a day. Half the shit an economist will yell you is "good" is like... the most cost-effective method to euthanize the poor.
ReplyDeleteThis is in fact bad, though.
Will a.i take accounting jobs?
ReplyDeleteWhat about HR jobs
I dont see why not? Accounting/bookeeping seems like a good use of ai, its already in big HR firms. Something else to consider is the jobs lost by people refusing to do shady shit for ai. Like a new thing in home service companies is to push the ai record button in the service app, then it records the entire conversation with the customer and provides feedback on improving sales numbers. Some people are quitting rather than doing it cuz it's shady and wrong.
DeleteNo.
ReplyDeleteThe economy is cratering. The cuts were coming anyway as we start massive decline. AI is the scapegoat, because if you say you're cutting for any reason but AI, the stock craters.
Bingo. And all these companies are realizing how much money they’re setting on fire now that they are paying the actual cost of AI.
DeleteIt's both. The tech sector is a bloodbath because you don't need a Computer Science degree or be able to write in complete sentences to write code anymore. They just hire warm bodies to type into Copilot. The economy is also cratering.
DeleteThis is true in the short term.
DeleteDo you know what "wishcasting" is? I'm not sure if that's what I would call this, but this almost religious mantra of trying to pretend that AI does not and will not have any impact on the economy and our industries - but how many incredibly smart people saying this, warning about this, does it take before people really take it seriously?
DeleteOn Reddit I hear a lot about how people need to listen to scientists and researchers and generally smart people more.
I guess if you work as a software dev and have used the latest models, especially if you've been tracking the trajectory over the last 6 months you have a different take away - at least I see devs finally taking this seriously en masse.
What would it take for you? What do you need to see?
You’re saying two different things, though.
DeleteAI is absolutely changing many industries forever in ways we’re a long way from understanding.
And there are businesses using AI to justify layoffs right now that actually have to do with unrelated business and economic fundamentals.
Right now I suspect we’re seeing more of the latter than the former, because it’s still extremely new and still hasn’t fully integrated into business operations. As it does there will be more churn and replacement, but even then it’s unclear what it will ultimately do to the workforce, the same way we didn’t know what the internet would do.
Here's where I'll kind of agree - AI is helping to make clear that lots of jobs are likely already unnecessary, overhires, and fundamentally changing the incentives of overhiring.
DeleteWho is being hired the most though - are people who can come into orgs, use AI, and automate as many tasks as possible. This is starting and will continue and ramp up.
None of this disagrees with what the economists are saying - they are not saying... All jobs that have gone in the last little while are because of AI? No - they are saying over the next 10 years (the range of time varies for all of them, but 10 years was agreeable to all of them) - this is likely to ramp up.
I am responding to the person who says "No" out of reflex, and if I will be bold, fear and discomfort.
Thank you! The number of people I explain this to on a regular basis. It really is another great example of how people are fed misinformation through so many avenues of media.
DeleteSo how do we capitalize on those stocks
DeleteFire your employees, sell your stock when it hits a high, then bail. IDK why more people aren't doing this...
DeleteBy being wealthy and owning them 20 years ago.
DeleteSame way we solve climate change: by doing something about it 20 years ago
DeleteNo we can do things now. Big things are coming soon for the working class maybe good big things maybe bad big things. But things.
DeleteThis is coming from an electrician who knows jack shit about stocks or what A.i will eliminate
Delete
DeleteStop trying to game the stock market then. Find some sort of large index fund like something tied to the S&P 500 or Vanguard’s VTI, put your money in it, and forget about it
Trying to gamble on stocks without knowing about them is a great way to lose money. Safe investments is a great way to make money over time
You do realize that most indexes are overweight in the stock that are going to create the economic disaster right?
DeleteMaybe, maybe not
DeleteOn average they do much better than whatever a non-expert can do, so I’ll put my money in them
What do you do with your money
There was a test done many years ago where they had a monkey pick stocks in some weird way. The monkey beat the majority of so called experts. I'd likely have to dig to find it.
DeleteThat said, I put everything into gold on January 17, 2025, because history shows that fascists blow up the economy sooner or later.