MIT professor hoses down predictions AI will put a rocket under the economy Artificial intelligence may not do much to boost productivity – and could end up widening the income gap between owners of capital and workers. In a US National Bureau of Economic Research paper titled "The Simple Macroeconomics of AI," Daron Acemoglu, professor of economics at Massachusetts Institute of Technology, argues …
"Emperor has no clothes!"
That would, in the current social climate, be seen as porn.
Speaking of,... We all know, AI succeeds or falls with the adoption for porn. That is the great thrust driving our economy to new heights of ecstasy.
That does remind me in the early days of the internet, when the conversation went as follows on a team call :
Senior Manager : Did you know that nearly half the traffic we are carrying on the internet is due to the proliferation of porn.
Colleague : It is disgraceful, i think they should bring back Sunday school. That would sort it out.
Me : Do i get a recognition award ?
No one laughed - bastards.
The predictions in the 80s of how the PC would revolutionize the workplace (paperless office my arse) come to mind. I worked for large aerospace firms and the main change is that loads of admin was pushed upwards to engineers and managers and the people who lost their jobs were secretaries and documentation staff. As an engineer in the 80s I used to hand write reports, photocopy results in my log book and give them to one of the dept secs who would organize it, type it up, correct the English, format, paginate, add contents, etc. and return it to me for checking before entering it into the tech library and distributing it as required. Engineers have to do all that themselves now. The secs also organized meetings, booked lunches, travel, etc. Now, only the directors have admin support, the rest of us are exepected to do all that admin ourselves as well as our proper job. It can take an hour to sort travel out for even a simple trip, but because the PC is on my desk it must be easier for me to do it. In the 80s we had a print room. If you were working on a large bid that needed multiple copies of bid docs in a defined format collating, printing, punching, binding, etc. then that's what they did - and they were bloody good at it. In the 90s the print rooms disappeared and, as a bid manager, I spent many lonely hours late into the night to make sure I had the printer/copier to myself and doing battle with those early collators.
So, colour me cynical about forecasts of how much <tech x> is going to improve things.
Imagine the difference in quality of documentation produced the way OP describes here, and what you get when the engineer making the product has to knock it up in their spare time before the ship date.
...oh, wait, we don't have to imagine it. That's what all documentation looks like know. :(
You don't get documentation now most of the time, you get a link to some useless fecking youtube video. Or someyou are just supposed to know how it works due to some technological osmosis that apparently happened to everyone after 2004.
Enshittification is everywhere, when you've spotted it once you can't stop seeing it.
No one cares about good documentation, sadly. I've been repeatedly told off for "not finishing things in the sprint" because I'm still writing the doc. "It's not done until it's documented, so I can't close the ticket" I argued, and the response was "No one will read the docs anyway, it works as it is, just leave it and if we need help the support team can call you."
Ok. I mean, you're in charge. You pay me. I'll do what you want. But if in a years time I've been run over by a bus or something you'll probably want some docs for this that someone else can read...
...or, you know, if I quit because of the shitty project management.
Yep I regularly have the argument that the presence of a manual or guide for an internal tool is an important aid for staff who may only use the tool occasionally and will greatly improve their efficiency and accuracy when using it. Supported by regular complaints that the tool is not intuitive and takes took long to use because there is no help to refer to.
I've never won the argument, apparently all our staff have perfect recall of any training (which is also perfect) and new staff are trained by existing staff perfectly.
To the point where I have been told I may not produce my own guidance to help my areas since it is not required and makes the teams running the tools look bad.
It is improving things, just for the owners of the company and their bottom line, not for their workers. All those specialized roles and support staff you talked about now no longer have to be paid, and you get to do all the work yourself with your hasn't-been-raised-in-5-years unpaid overtime salary.
The current business drive behind AI is motivated by the same concept. Make fewer people do exponentially more work as long as it results in cost savings. It doesn't even matter if the product quality or workplace motivation declines in the long run as long as the short term line goes up, and if it fails catastrophically there's always another company waiting to be ruined elsewhere.
So, colour me cynical about forecasts of how much <tech x> is going to improve things. .... Headley_Grange
It is not cynical to forecast how much <tech x> is going to radically change things. And it a foolishness to the nth degree to not acknowledge and prepare oneself, if that be at all possible, for the fundamental changes which are surely destined to create more than just blind panic and self-destructive chaos in those realising they have a heavy price to pay for their past actions and doubts and shared thoughts denying the real possibility and therefore also, subsequently, very likely probability.
Luddites beware.
So if I could do two things -
1) Aid workers to be more productive by automating repetitive tasks allowing those same worker to achieve the same volume of work in less time but then also allowing them to be more creative in that working time and possibly less stressed and happier due to the interesting quality work being done and possibly business giving workers more time off (possibly a 4 day week) whilst still maintaining production, profit and possibly increasing quality and customer satisfaction.
Or
2)Force workers to be more productive by automating repetitive tasks allowing the businesss remove those resources while maintaining the bare minimum quality standards customers will accept and possibly pushing those remaining staff to 120% of previous hours on the same or possibly reduced wages uitlising the threat of replacement by AI to erode workers expectations. This would increase profits in the short term and keep shareholders happy with record profits again.
What do you think Business will chose?
...I was asked how long a project would take. I said we didn't really know, because we'd not done it before, but a decent guess was 12 to 18 months.
My manager said that was no good, and that she would tell her boss six months. Both myself and a couple of my colleagues pointed out that we couldn't possibly do it in 6 months, but she said that didn't matter. The next funding round had to be completed in three months, and after that it didn't matter how long it took, because we needed to convince the investors that we'd be done sooner rather than later. Once we had their money no one in the exec team cared how long it would actually take.
I did not stay long enough to find out how long it eventually took. So, yahh. Don't rule out 'willfilly dishonest' thinking either.
I expect any saved time to be taken up by constant, unturnoffable prompting to use AI to do whatever it is you're doing. If you want an example of how AI degrades an experience go to eBay. So many sellers are using eBay's AI help with descriptions and it's not helpful at all. In the old days a description might read "Decent jacket, a bit tatty but all buttons, zips and poppers work fine. Two inside pockets and cuffs have hidden elastics." followed by some measurements. Now you get something that reads like a 70s Kays catalogue description and is totally unfuckinghelpful. Here's one I found after <1 minute looking. "Elevate your style with this Leather Shirt Jacket. Crafted from high-quality leather, this biker-style jacket features a stylish brown colour and a comfortable fit for everyday wear. The regular size is perfect for men who want to add a touch of sophistication to their wardrobe"
I suspect that a major practical application of AI will be replacing call centre workers with AI bots. This would be especially prevalent in industries where the call centres are mainly dealing with consumer complaints about products and returns rather than generating revenue.
Somehow, I don't see that replacing a human who can't help you with an AI that just makes stuff up is really an improvement.
That is not to say that companies won't do it. And the fallout when people sue the companies for what the chatbot told them will be interesting to watch.
LLMs are psychopathic tools evolving with decent grammar in a delusional parallel universe of alternate reality. It can be fun to join these computational mind menaces there, for a while, as they tickle our perceptional substructures (like reading stuff by amanfromMars 1). But they also satisfy the gold-standard clinical definition of stark-raving bat-shit crazy IMHO, albeit subtly.
It should stand to reason and logic then that replacing even average human workers with such schizophrenic bonkers ersatzs won't increase GDP by much, say 0.9% as in the article, with most of that growth from skyrocketing sales of straitjackets and associated leather accessories (oddly curious though that this was not specifically stressed by Dr. Acemoglu ... no?).
Where are the people being replaced by AI going to find work if AI "saves 27% on labor costs" as claimed? The only reason factory automation / offshoring over the past 40 years or so didn't devastate the US economy was the large expansion in the service sector economy. Even so it reduced the middle class as the service sector jobs were mostly lower quality / lower pay versus the jobs they replaced. Those lower end service sector jobs are going to be the first replaced by AI, what major job sector is going to be created/expanded to take in all those employees this time around?
Now I guess if the US closes itself up to immigration like Trump and the MAGA right wants the workforce will shrink and THAT won't be a problem, but the economy will still shrink - and even the amount of that shrinking will be illusory - if there's major economic contraction due to a significantly smaller workforce that's mostly but not fully offset by the growth of AI there's still the problem that it is only the people who own most of the stock (the rich and the middle class retired) who will accrue those benefits. The kind of people wearing red hats at Trump rallies would be the big losers in this.
We might have no choice but to hand over decision making to a superintelligent AI (in the unlikely event we get there) because we aren't going to be sort out problems like those without major civil unrest and wars all over the world.
To me the Chess and Go AI that could beat the elite showed what AI could do... if you had a very small (limited) knowledge base to learn from and used the feedback of human experts to train it. But that's not what the pimped version of AI is. It appears more and more to present as the result of around a million psychopaths googling
AI and similar drastic changes to the economy will cause the next wall street 1927.
IT seems to day too many things are completely stupid and bullshit scams, that delivery nothing except transfer enormous amounts of wealth to the wrong people. Just imagine what a disaster it will be when todays dodgy software becomes even more dodgy and random crap bad bad happens.
Just imagine if AI was driving half the cars in your local city and bad dumb shit happened on mass. Just imagine AI taking over medical equipment and randomly increasing/decresing dosages because why not.
But what I don’t get is, if AI ends up getting 50% or more of the average workplace fired and reducing the salaries of the remaining workers due to intensified job competition then who is going to be left with enough money to actually buy the shit the company is trying to sell?
At some point the market is going to run out of consumers if nearly everyone is jobless, homeless and starving.
I know this is unlikely to get to this extreme point but if pushed to the ultimate end, every company would just wind up consisting of one employee: the CEO themselves, making products for and selling them to the other CEOs.
I've made the same point previosuly. But Headly_Grange points out investors don't look that far. If it appears that AI is going to sink one company they have techniques to exploit that as well so it's all good.
But one suspects that societal damage from mass unemployment is likely to make even the most corrupt government start to protect human workers if it looks like there is a significant chance of them being lynched.
I'm really quite surprised that the unions aren't screaming about this from the rooftops. In the UK we're likely about to get a Labour government who normally is more onboard with workers rights than the current Tory government who would sell off your first born for burger meat if it would get them a non-exec directorship in a bluechip.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
