But candidly, I do not see AI as being the basis for growth
Sigh.
there is one thing that worries me most of all, and that is that AI creates the power for big business to dispense with labour
That is growth. Moron.
There is now more labour available to deal with some other human want, desire or problem. Which means that we get two wants, desires, problems, solved out of our same supply of that scarce resource, human labour. This is, by definition., economic growth.
Just another Worstall fallacy as people are pushed out of well-paying satisfying jobs to become door greeters at Walmart. Well they’re satisfying a human need or desire aren’t they?
@Southerner
I once read a remark that gave me pause: that the Highland Clearances had been a good thing because they had caused a bunch of tough, hard-working people to quit wasting their time subsistence crofting and instead emigrate, then to have a disproportionately positive role in creating the modern world.
From this we learn that work isn’t about “jobs”. It’s about creating value. Mining coal wit’ pick and shovel was a “job”. Do you still want people doing that and dying of pneumoconiosis in their 50s? Or would you rather a machine did it and those miners instead did other, more healthy work?
AI is very hungry for energy and we have the most expensive energy in the world.
AI data centres can be sited anywhere in the world.
So no AI centres will be built in the UK.
@ Southerner
“Just another Worstall fallacy as people are pushed out of well-paying satisfying jobs”
When do you think we should have pressed the pause button? Steam engines? The ‘spinning jenny’?
Ah Norman, but AI will never replace actual manual labour.
What it does is replace your average pen-pusher. Even, dare I say it…, several levels of Decision-Makers..
And given the actual level of “intelligence” of current AI…. That’s saying something about the people in the jobs it replaces.. I sincerely doubt they could actually handle a real job.
@ Southerner
“Well-paying, satisfying” but capable of being done by a computer – seems an odd combination.
AI is very hungry for energy and we have the most expensive energy in the world.
AI data centres can be sited anywhere in the world.
So no AI centres will be built in the UK.
Unable to understand cause and effect the government will insist that AI data centre are built here for data security reasons and then wonder why nobody takes them up on the offer and they end up paying £Bns in subsidies.
That’s probably moot anyway. Starmer can either have ever closer realignment with the EU, the worlds greatest regulator and AI/tech laggard, or have AI, but he isn’t getting both.
While passing time due to a cancelled ‘plane last night I caught some BBC News 24 discussing this. A chap was saying “we had traffic lights before cars, so we need the same for AI”.
WTactualF? He really thinks traffic lights existed before motorised vehicles created the need for traffic lights? Traffic lights were invented *BECAUSE* of the effects of the existance of cars, cars were not a creation of the exitance of traffic lights. Or was I in some sort of feaver dream at being awake for 27 hours?
Hold on a mo, I need to clear the drifts from the back yard to stop it snowing.
For people’s amusement:
Traffic lights were first invented in 1868 by John Peake Knight, a railway signaling engineer in London. The primary reason for their invention was to regulate traffic flow and reduce the need for police officers to manually control vehicle circulation.
Early Development
The first traffic signal system was established on December 9, 1868, at Parliament Square in London. Knight’s original design consisted of:
* Three semaphore arms
* Red and green gas-powered lamps for night-time use
However, the initial system was short-lived. On January 2, 1869, the gas-powered signal exploded, injuring a police officer and causing the project to be abandoned as a public health hazard.
@jgh depends on what you define as a “traffic light”, really.
Traffic wardens, and even semaphore systems in some cities when the main, if not all, traffic was still horse and carts/coaches.
Because clogged city arteries are not just a thing since cars were invented..
From what I can read, both development/adoption of cars and electric traffic lights sort of evolved in parallel from the 1920’s
Cars replacing horses, traffic lights replacing traffic cops as things got automated. With both hitting their “modern” incarnations in the 1950’s. Anything after that is just jiggling the basic designs and improvements in efficiency.
Cars developed because of convenience/efficiency in transport, traffic lights mainly to reduce the costs of regulating that traffic.
What Southerner said. The whole AI thing will not make the UK any wealthier. Or rather it will make a few very much wealthier, a small but significant proportion much poorer, and everyone in the middle a small amount wealthier, but not noticeably so. So the gains and the losses will be concentrated. Which is revolution fodder……
“There is now more labour available to deal with some other human want, desire or problem. Which means that we get two wants, desires, problems, solved out of our same supply of that scarce resource, human labour. This is, by definition., economic growth.”
How did that theory work out in the post industrial parts of the UK then? Why is Merthyr Tyfil a shit hole and not one of the wealthiest parts of the UK right now? After all once the factories shut there were loads of people freed up to do other things and make the whole area wealthier, right?
OT: talking of computery stuff, did you see this in El Reg Tim? The govt will apply all sorts of new rules to sites visible and used in the UK. If all your stuff is totally offshore from the UK then you could probably tell Ofcom to eff off when they come calling, like Guido has done in the past.
However it’ll sound the death knell for the likes of Little Snoring-on-sea Locomotive Appreciation Society’s website. You could almost see it as a way of closing down all the little stuff so they only need to spend effort on the big boys.
@Grikath
Ah Norman, but AI will never replace actual manual labour.
What it does is replace your average pen-pusher. Even, dare I say it…, several levels of Decision-Makers..
It’s going to replace those people who tell you “Computer say no” isn’t it? So at least one will be conversing with the actual computer & find out why “Computer say no” One might even be able to have an intelligent conversation with it to the point where “Computer say yes” That’s progress! It can hardly be worse.
I wish they’d fucking get on with it with voice-activated phone answering decision trees. Right now the only way to get past some of these is to say “I want to close my account”. That gets you through to a decision-making human in short order.
“Arthur the Cat
Traffic lights were first invented in 1868 by John Peake Knight, a railway signaling engineer in London”
His son founded the JP Knight shipping company – still family owned today. When I worked in London they were a client of the accountancy firm i worked for. Mostly ocean going tugs. One of them rammed a nuclear sub up in Scotland. I don’t think it was deliberately.
Wales’s place tells me that “The first use of a signal box was by the London & Croydon Railway in 1843 to control the junction to Bricklayers Arms in London”
Semaphore of course, but same principle.
Stiglitz saying that regulation such as traffic lights increases the freedom of motor car users is right, but the private sector got there first.
Norman: « Right now the only way to get past some of these is to say “I want to close my account”. That gets you through to a decision-making human in short order. »
Vodafone (a curse upon them) now add a rider to their “security procedures” where they ask if one has any special issues, needs or requirements before proceeding and it’s helpful at this point to mention a regrettable tendancy to get very rude and angry when treated like an idiot by people reading off a screen.
Everyone needs to time their exit from NVIDIA. To when the current AI fad dies, as it will. And then take off again, and achieve something much less exciting than the bull we are being fed.
If anyone has an idea when that might be, I’d appreciate the tip.
@Grikath
I could see AI replacing a lot of manual labour. Care home workers would be a good choice. A robot with AI guidance could do the somewhat repetitive care and cleaning chores as well as providing entertainment and conversation for the residents. The main barrier to entry would be the initial and maintenance cost of robots compared to the staffing costs which will get higher and higher as the number of people needing care rises and the number willing to work in the care industry falls. The cost of the AI element could be comparatively negligible.
AndyF: That’s not AI, that’s robotics.
@AndyF
In your dreams, mate. That’s the thing about many manual jobs. A robot to do the same work would cost millions. Just take something like sweeping a floor. A machine to sweep the clear areas is simple. Now try & sweep a floor with tables & chairs on it. And the ledge on the top of the skirting. And in that tight corner. Now add people using the floor as being swept. Or a liquid spill. Or valuable items inadvertently dropped. So how much for a robot can do all that? When you can hire someone on minimum wage.
The jobs that will go are jobs handle information. Any job handles information. Virtually all desk jockey jobs are distributed, rules based, information processing.
At last. Middle classes.The machines are coming for you! After a century of supremacy, the demise of the administrator. Back to Bob Cratchett times when the most menial job in the firm was the clerk.
Jim,
“How did that theory work out in the post industrial parts of the UK then? Why is Merthyr Tyfil a shit hole and not one of the wealthiest parts of the UK right now? After all once the factories shut there were loads of people freed up to do other things and make the whole area wealthier, right?”
No, because areas may suit a particular industry and not another industry. Merthyr only ever had what it did because of its natural resources of iron ore and coal, so grew rapidly in the 19th century to produce steel. Then other places learned how to do it that were more coastal and could do it cheaper, so Merthyr’s steel works declined and they shifted to coal mining, again, because of the location. But does it makes sense to have a factory there rather than Bridgend or Port Talbot where you can quickly get them to port?
It’s like the abandoned whaling station at Grytviken in the Falklands. Used to employ a lot of people to get whales. We stopped whaling, what else are you going to do there?
But what happened is lots of people left Merthyr over time. Population declined from 90,000 in 1921 to 55,000 in 2001.
You also have to be really careful with Merthyr’s data. It has one of the highest rates of labour inactivity in Wales, but its unemployment rate is quite low. There’s lots of retired people there. You don’t need to go to work, why not live in a little place in the middle of nowhere?
“ Civil Service middle management has doubled in size since 2010, a report has found.”
From the Telegraph.
That’s a shit load of low hanging fruit for AI to get rid of.
“That’s a shit load of low hanging fruit for AI to get rid of.”
You can bet your bottom dollar that AI will not reduce the State employment headcount by one person.
“No, because areas may suit a particular industry and not another industry. ”
Ah, so what if the same thing we have seen in Merthyr also applies to the UK as a whole? We get rid of loads of the current jobs, but there’s nothing productive left for everyone to do afterwards? Can the UK as a whole end up a shithole like Merthyr? I mean we’re well on our way there now, AI might just tip us right over the edge. But hey we’ll be wealthier, our economics geniuses tell us so……..
BiND’s post at 9:50 am on squatting law seemed a good example of how thing may go. If you asked a lawyer the question BiND asked, they’d probably have to go look it up. So what are they doing for their money that you couldn’t do for yourself? They simply have the advantage of knowing where to look it up. Like AI does.
One of the few desk jockey jobs I’ve had was writing share portfolio reviews. So what was I actually doing? Taking a handful of data points & putting them into a form someone not experienced in portfolio management could vaguely understand. Even the data points were public knowledge if you knew where to look. Essentially, it was a non-job. Just a capability to string some words together. The management decisions weren’t much different. The data points produced the decisions. The broker I was dealing with a while back produced similar & charged two grand annually for the privilege. Until I fired him.
‘When do you think we should have pressed the pause button? Steam engines? The ‘spinning jenny’?’
Obviously when I was pampered by hot and cold running servants catering to my every whim.
Odd isn’t it Andrew C, how my answer to every question is, ‘Pamper me more!!!!’
I had occasion to ask co-pilot two questions this past couple of days – the first one required sever tries to get the response to be sufficiently detailed (allowed rates of return on equity in a regulated utility segment), and I don’t think that’s a problem. When I looked at the citations / links it provided, they were a acattershot of years – 2016, 1995, 2003 – none current, and all quite useless.
The second question was when GE sold their hydro turbne business (which I remember was in the late 90’s or early 00’s). I was told that GE didn’t sell their turbine business, and it continues as GEVernova. I had to correct co-pilot that GEVernova was the continuation of the turbine business that GE bought from Alstom in 2013 (IIRC), and that GE had a turbine business that they sold in the late 90’s. The bot said, ‘gee. sorry I was wrong – GE sold that business to XXX Energy – but I knew that was wrong.
I score it as a 0.5 out of 2, and I think I am being generous. I wouldn’t trust AI with anything important. Sure – knocking out a rejection e-mail is fine – but nothing I really cared about.
By the way – if anyone remembers the details on that sale of the GE turbine business, I would appreciate a pointer.
I can see an attempted to sale to Pescarmona in around 2007
https://www.reuters.com/article/business/sale-of-ge-hydro-unit-to-pescarmona-falls-through-idUSN22460163/
Not clear at first search if it did eventually get sold
And following the rabbit down a hole we have
https://www.cbc.ca/news/canada/montreal/ge-hydro-plant-closure-to-cost-450-montreal-jobs-1.683915
Interesting bit about the GE turbine plant closure in Canada. I’m sure all the 450 engineers and machinists who lost their jobs all found wonderful new lucrative employment opportunities in place such as Starbucks and KFC, or offering dog walking services and manicures. And the rest of Canada was made considerably wealthier as a result.
But the question always is, dcardno, if you asked a human “expert” would you get a better result? I’ve had far too many claimed experts tell me things I’ve known to be incorrect to be definitive on that.
The thing about AI is not whether it’s perfect but whether it produces better results than a human. And of course the cost of producing those results.
It was like my supposed “investment advisor”. After mere half hour’s experience of him he was obviously an empty suit only capable of parroting his paperwork.
Asking AI is like Google searching, it depends on the questions asked. Here’s my conversation with ChatGPTm
Me: when did GE sell their hydro turbine business?
ChatGPT: In 2006, General Electric (GE) agreed to sell its hydro business to the Pescarmona Group of Companies (PGC). However, in 2015, GE acquired Alstom’s power generation and electricity transmission business, which included hydro and wind assets. These were integrated into GE Power & Water and later spun off to create GE Renewable Energy. As of April 2, 2024, GE’s energy businesses, including hydro, were consolidated under GE Vernova. Therefore, GE has not sold its hydro turbine business; instead, it has restructured and integrated it into different divisions over time.
ME: (Hmm, that looks suspicious) Did the sale to Pescarmona get completed?
ChatGPT: No, the sale of GE’s hydro business to Pescarmona Group of Companies (PGC) was not completed. Although GE announced an agreement to sell its hydro business to PGC in December 2006, the deal fell through in June 2007, and GE continued to operate the hydro business.
Jim,
“Ah, so what if the same thing we have seen in Merthyr also applies to the UK as a whole? We get rid of loads of the current jobs, but there’s nothing productive left for everyone to do afterwards? Can the UK as a whole end up a shithole like Merthyr? I mean we’re well on our way there now, AI might just tip us right over the edge. But hey we’ll be wealthier, our economics geniuses tell us so……..”
It’s possible. But there’s no “we” in this. If AI eradicates a large chunk of the services sector, it will eradicate a large chunk of the services sector. There’s nothing you can do about it. All you can really ever do is have the sort of government that does the right things to bring about economic growth and hope if works out.
BIS,
“The thing about AI is not whether it’s perfect but whether it produces better results than a human. And of course the cost of producing those results.
It was like my supposed “investment advisor”. After mere half hour’s experience of him he was obviously an empty suit only capable of parroting his paperwork.”
It’s the sort of industry where I say “I can replace them with a flowchart”. And Vanguard have. You can log on, you put in age, risk etc and it suggests some ETFs, bonds etc. And you pay a tiny percentage for Vanguard to manage it. There are people I know handing over 2% per year to “wealth management” and when you look at the investments, it’s just average stuff. Even if there’s a slightly better performance from various managed funds over an ETF, the 2% fees kill it. Because the funds are mostly bullshit too, just spreading money around enough companies so you get an average.
GPs are like this. I can spend 5 minutes on Google and figure out my problem before I go to a GP. And have not yet been wrong. And I didn’t even take biology CSE. They are pretty much worthless. I use them to either getting me drugs I’m not allowed to buy, or referring me to a specialist.
And we pay them wildly to do this. People are like “we need more GPs”. No, we need to create a new job that takes less than 6 months of education to handle the top 10 routine things GPs do, like strep throat, stomach aches etc, so we can have a load of relatively cheap people doing it. Then spend more money on more specialists for the hard stuff, for the surgical side.
Thanks Bongo – I thought their attempted sale was earlier, but as I think of the contracts we had with them, it couldn’t have been. I think I was conflating closure of the Montreal plant with sale of the entire business.
BiS
Hmm – my test is whether it does a better job than me (or the guys on my team that I might assign to do that sort of survey) – and the answer for co-pilot was ‘no’ (I’ll admit ChatGPT seemed better). I wouldn’t have bothered with the financial data from 1995 (or immediately discarded it), would’ve made a note that they applied for a delay in filings subsequent to 2022, and so on. I probably would’ve had problems with the GE Hydro sale, becasue I had the wrong date in mind (although if I had run into Bongo’s second link that might’ve set me straight). Co-pilot made up a non-existent sale to match my erroneous memory. If it had mattered (it didn’t – that question was just for personal curiosity), that would’ve been a problem.
I asked Perplexity the same question: when did GE sell their hydro turbine business? – and received (the numbers are links to sources):
GE sold its hydro turbine business in multiple stages between 2006 and 2008:
1. In December 2006, GE announced an agreement to sell its worldwide hydro business to Pescarmona Group of Companies (PGC) 2.
2. However, this sale did not fully materialize. Instead, GE restructured its hydro business and moved its global headquarters to Brazil 1.
3. By mid-2008, GE had decided to divest its restructured hydro interests and sold its majority share in the joint venture GE Hydro Inepar (GEHI) to Andritz 1.
4. The sale to Andritz was completed in two parts, with the purchases of the hydropower technology and GEHI stake closing separately, but both within months of each other in 2008 1.
This series of transactions marked GE’s strategic decision to exit the hydro turbine business, which it had made less than 10 years prior to 2015 1.
Related
+ Why did GE decide to exit the hydro business
+ What happened to GE’s hydro business after it was sold
+ How did the sale of GE’s hydro business impact its renewable energy focus
+ What other companies were involved in the acquisition of GE’s hydro business
+ How did the sale of GE’s hydro business affect its employees
So according to Perplexity GE is out of the business, but according to ChatGPT it isn’t.
GE sold its hydro turbine business in India in 2024 and parts of its hydro turbine business to Andritz (Austria) in 2022.
Which also came from Perpoxity and is bollox. Don’t trust any AI output.
Seems like AI is going to be like the press – when you actually know something about the matter at hand you realise it doesn’t have a clue what its talking about, but when you know nothing yourself you will blindly believe anything it says.
Norman – thanks; it was the sale to Andritz that I was thinking of. Took quite while to find it, though!
Gen AI is a statistical parrot and the info will only be as good as the quality of the info available to it. GIGO.
The output is hard to test as it’s text so mistakes are more likely.
Machine learning on quantified data, or even qual data, that one can use for short term predictions is much more useful as it’s testable. And that means model tuning is possible.
GenAI is like a knowledgeable moron. Credible until you know it’s wrong. My boss loves GenAI. Tells you a lot about my boss.