The cost of the welfare state - challenges posed and solutions

[GusB]

It strikes at the very issue underlying this thread though - net-contributors and net-non-contributors. The former encompass the very people who are absolutely critical to keeping society running and are being very heavily taxed to the extent that it is encouraging many to retire or downscale. Whilst the latter are enjoying a subsidised lifestyle that, whilst not luxurious, is sufficient to deter many from self-improvement and advancement. As you note, AI will hit tax receipts hence why the benefits bill must be tackled - everyone should be incentivised to work and the welfare state should reduce to being an emergency safety net only.

It is becoming increasingly difficult to respond to your posts without breaching forum rules, but I'll do my best.

First of all, I'd like you to provide an exhaustive list of "people who are absolutely critical to keeping society running". My guess is that you're referring to doctors, lawyers and other "professionals", given your comment about them being so heavily taxed. But what about the cleaners who maintain their offices and hospitals? Are they not essential for keeping society running? What about the people who stock the shelves in the supermarkets, drive buses and deliver your pizza, etc.?

You're absolutely right that it's not a luxurious lifestyle, but it's an insult to the people who do these jobs to suggest that they're living a "subsidised lifestyle". Who is actually being subsidised? Is is the people who work for meagre wages, or the corporations that pay peanuts in the full knowledge that the state will pick up the slack?

Your comment about "self-improvement" is disgusting. I don't know many people who wouldn't like to gain a few more qualifications in order to improve their income but it costs money to do the extra training and people often don't have the time, especially when employers require their employees to be available over a wide range of hours, or if the individual has to juggle two or more low-paid jobs to make ends meet.

It shouldn't have to be a fight; if you work for 40 hours a week you should expect to be able to afford decent accommodation, put food on the table and maybe have a little bit left over for a treat every now and again without worrying. If you want to have a Mercedes, a swimming pool and room for a pony, fine - you'll have to work a little bit harder for it, but there shouldn't be any need to "better yourself" in order to afford the basics.

I've no issue with incentivising people to work, but you are all sticks and no carrot - your "incentive" is work or starve and that doesn't do anyone any favours because social services will end up picking up the tab. Isn't it a better idea to fund education and training from the outset rather than local authorities having to provide emergency housing for a family that slips through the net?

 

[Krokodil]

No, what we need is sensible marginal tax rates.

Agreed. The examples you list (which should be well known by now, they've been a problem for years) are insane. It's having all sorts of unintended consequences - can't get a GP appointment? They've gone part time because the marginal tax rate makes any extra hours seem not worth the bother and they've maxed out the amount they're allowed to contribute to their pensions.

But the point is that it has been frozen since it last increased in 2021

After several years of above-inflation rises. We're almost back where it was (bearing in mind how distorted our tax burden is, it's right that everyone contributes a little, we're seeing how counterproductive it is just to chuck everyone on higher earners). NB, I'm never likely to hit the Additional Rate so I'm not after personal gain here, I just recognise that there's a problem. I did however hit the Higher Rate for a couple of tax years, while working as a railway guard, that's how bad the fiscal drag on the Higher Rate threshold has been - you don't have to earn much over average to hit it. That needs to change (again I'm not advocating for my own gains, I'm not going to be near that threshold this year and will probably be in an apprenticeship by next).

 

[styles]

It is a shame that this is always interpreted as a high marginal tax rate, since the logic of higher earners paying a greater proportion of their income in tax certainly isn't a flawed concept.

It's not but three things spring to mind:

1. 60% income tax + 2% NIC (67.5% + 2% in Scotland) is a very high rate of tax for someone earning a bit over £100k. Obviously it's a privileged position to be in and a nice problem to have, but £100k may be the sole income for a household in London for example and while it sounds like a lot, I think it's a bit of a low threshold for 62-69.5% tax.
2. The absurdity of it isn't really in the specific rate of tax for this band of income - it's that your tax rate goes 0%->20%->40%->60%->45%. What you're arguing for, and which I agree with, isn't what this system achieves, because the marginal rate of tax is going down at £125k. That's the exact opposite of a progressive tax.
3. £100k is also the cliff edge for tax-free childcare. This means someone earning £99,999.99 with two children under three in London will lose an immediate £20k if they earn a penny more. That's an effective marginal tax rate of 20,000%, which is just ridiculous. These things should absolutely be means-tested in my opinion, but on sliding scales, not cliff edges like this.

For what it's worth, around a million taxpayers are affected by the £100k-£125k tax trap, so it's not like it's a rare edge case. The tax thresholds are (under Starmer's government) staying put until 2031, so a lot more people are going to be dragged into this mess.

 

[AlterEgo]

It strikes at the very issue underlying this thread though - net-contributors and net-non-contributors. The former encompass the very people who are absolutely critical to keeping society running and are being very heavily taxed to the extent that it is encouraging many to retire or downscale. Whilst the latter are enjoying a subsidised lifestyle that, whilst not luxurious, is sufficient to deter many from self-improvement and advancement. As you note, AI will hit tax receipts hence why the benefits bill must be tackled - everyone should be incentivised to work and the welfare state should reduce to being an emergency safety net only.

A lot of people are net non-contributors because their work is so poorly paid that they have to claim Universal Credit while in work. And a lot of the people who are not currently net contributors are raising families (more taxpayers, one hopes!) and those children will be the one paying your pension. It's shortsighted to see yourself as a single striver as unmoored from society and impervious to negative change that happens further down the economic ladder. So a large part of the solution is to wean Britain off its addiction to low waged and unproductive work and that change has to come from the top rather than the bottom.

 

[Bletchleyite]

It is a shame that this is always interpreted as a high marginal tax rate, since the logic of higher earners paying a greater proportion of their income in tax certainly isn't a flawed concept.

Call a spade a spade.

The latter is true, but it should be done more progressively, e.g. by having a 50% top rate, not by making the rate in effect very high for an arbitrary narrow band.

This fudge was done to avoid increasing income tax rates/bands themselves, and is predictably causing problems in return for its dishonesty.

 

[Bald Rick]

Image shows crazy actual marginal tax rates in 2023, all over the place - would be even worse if it included universal credit taper)

That’s for someone with three children of Child benefit age. The graph for someone with no children would be rather different at the left side.

I wonder what proportion of taxpayers have three (or more) children of child beenfit age. My guess is below 10%.

 

[styles]

That’s for someone with three children of Child benefit age. The graph for someone with no children would be rather different at the left side.

I wonder what proportion of taxpayers have three (or more) children of child beenfit age. My guess is below 10%.

Even if you only have one child eligible, the difference between £99,999.99 and £100,000.00 income, 1p difference, can mean the loss of thousands of pounds in tax free childcare. That's just a daft cliff edge, one which results in a marginal tax rate of hundreds of percentage points.

Benefits like this should be means-tested using a sliding scale, not a small number of cliff edges. It makes no sense.

And in fact, I'd go a step further and suggest it results in less income for the treasury because people facing 69.5% effective rates of tax on parts of their income will just chuck into a pension or get a cycle to work scheme bike or company car they don't need. The amount of tax avoidance which happens in the £100k-125k range must be absolutely massive.

 

[Yew]

So a large part of the solution is to wean Britain off its addiction to low waged and unproductive work and that change has to come from the top rather than the bottom.

Unfortunately, the strategy of the last 40 years, of trusting in the 'wisdom' of the free market has shown that well paid jobs and industries don't just appear in deprived and deindustrialised areas of our country. Hopefully we are learning that lesson, and will be more proactive in future.

Until it seemed to disappear, I was quite a fan of Starmers "green industrial revolution" idea, trying to get us at the forefront of new energy technologies, unfortunately that didn't really seem to pan out.

 

[Broucek]

For what it's worth, around a million taxpayers are affected by the £100k-£125k tax trap, so it's not like it's a rare edge case. The tax thresholds are (under Starmer's government) staying put until 2031, so a lot more people are going to be dragged into this mess.

It's a shocking tax that leads to all kinds of dysfunction and yet no serious politician seems willing to tackle it. Even Truss's imfamous cuts didn't target this. As you say elsewhere, even if it's more than most earn, in London £100k is hardly "rich".

The NHS examples are particularly obvious and unhlepful.but it's not great for competent private sector professionals to start working a four day week to avoid this tax. I strongly suspect that applying (say) a 47% +2% NIC rate from £100k with no erosion of threshold would lead to an increased tax take as people would stop working part time/declining promotions/paying loads into pension

 

[JamieL]

A lot of people are net non-contributors because their work is so poorly paid that they have to claim Universal Credit while in work. And a lot of the people who are not currently net contributors are raising families (more taxpayers, one hopes!) and those children will be the one paying your pension. It's shortsighted to see yourself as a single striver as unmoored from society and impervious to negative change that happens further down the economic ladder. So a large part of the solution is to wean Britain off its addiction to low waged and unproductive work and that change has to come from the top rather than the bottom.

I think the trend will go the other way. AI will eliminate the requirement for many of the middle managers and white collar work which will force many of those individuals into low paid employment doing tasks that it would be uneconomical to commission robots and AI for. Accordingly the need for clear blue water between a low paid role and benefits is essential.

 

[ainsworth74]

It's a shocking tax that leads to all kinds of dysfunction and yet no serious politician seems willing to tackle it. Even Truss's imfamous cuts didn't target this. As you say elsewhere, even if it's more than most earn, in London £100k is hardly "rich".

I think part of the issue is that we seem be locked into thinking of salaries as if its still 2010 or something. £100k in 2010 was a very good salary and as most people still seem to think in those terms lots of voters get up in arms at the idea that we need to ease the burden of people in that £100k to £125k tax trap. "Oh no how said the well paid are having to pay more tax boo hoo" etc etc. But actually £100k in 2026 after sixteen years of inflation and various other changes is still a good salary don't get me wrong. But it's nothing like as good as it was.

I don't quite know what the solution is there but it does seem like a raft of issues are being caused by people thinking that what was a good very good salary sixteen odd years ago is still a very good salary now.

I suppose it's probably linked to the wider issues we have whereby there's a huge chunk of the working population stuck in the minimum wage and just above band (wage compression I think is the concept?) for whom £100k is still a fantasy salary as the UK seems to have become a low wage economy outside of a few industries/areas.

 

[AlterEgo]

I think the trend will go the other way. AI will eliminate the requirement for many of the middle managers and white collar work which will force many of those individuals into low paid employment doing tasks that it would be uneconomical to commission robots and AI for. Accordingly the need for clear blue water between a low paid role and benefits is essential.

I don't share this view. AI will not result in mass unemployment or underemployment and the functions it does best do not require executive function like being someone's line manager. Anyone who has a low level white collar job where they do not need to sign off on work done or supervise it is most at risk, not middle management.

Even so, we still haven't seen a lot of workplace upheaval despite massive advances in AI. The idea we need to make a barrier between benefits and work that starts off by saying poor and vulnerable people should get even less money is a non starter. You will feel the effects of this at some point because you will live in a worse society with greater intractable problems that may manifest in unpleasant ways.

 

[Bantamzen]

I don't share this view. AI will not result in mass unemployment or underemployment and the functions it does best do not require executive function like being someone's line manager. Anyone who has a low level white collar job where they do not need to sign off on work done or supervise it is most at risk, not middle management.

Even so, we still haven't seen a lot of workplace upheaval despite massive advances in AI. The idea we need to make a barrier between benefits and work that starts off by saying poor and vulnerable people should get even less money is a non starter. You will feel the effects of this at some point because you will live in a worse society with greater intractable problems that may manifest in unpleasant ways.

AI is moving at a pace probably unapparelled in industry. What was little more than a useless gimmick just a few months ago, is rapidly being looked at as a serious solution not just for automating low level jobs, but for automating a lot of that "middle management" work too. I'm currently part of an evaluation of some models, that have gone from quirky but inaccurate search engines, to full systems capable not just of data engineering, but analysis, evaluation and decision making, and this evaluation has just been massively scaled up. Even in a relatively small sandboxed environment, AI models can show very rapid improvements as they start to learn from the data and processes we expose them to. It's impressive and frightening in equal measures.

Being quite frank, I can see my team's current roles in data engineering, analysis and development disappearing down the AI rabbit hole, leaving only the need for some AI modellers and a few senior leaders to sign off on decisions. Projects that might take weeks or months currently can be accelerated to days or even hours, especially once models are given the full capacity of data centres and not just sandboxed with limited processing capacity. It's a good job I'm closing in on a possible retirement date, otherwise I would actually be concerned for my job. Don't be so sure white collar work is safe, I can see a large proportion of it heading to AI-ville in the not so distant future.

 

[brad465]

AI is moving at a pace probably unapparelled in industry. What was little more than a useless gimmick just a few months ago, is rapidly being looked at as a serious solution not just for automating low level jobs, but for automating a lot of that "middle management" work too. I'm currently part of an evaluation of some models, that have gone from quirky but inaccurate search engines, to full systems capable not just of data engineering, but analysis, evaluation and decision making, and this evaluation has just been massively scaled up. Even in a relatively small sandboxed environment, AI models can show very rapid improvements as they start to learn from the data and processes we expose them to. It's impressive and frightening in equal measures.

Being quite frank, I can see my team's current roles in data engineering, analysis and development disappearing down the AI rabbit hole, leaving only the need for some AI modellers and a few senior leaders to sign off on decisions. Projects that might take weeks or months currently can be accelerated to days or even hours, especially once models are given the full capacity of data centres and not just sandboxed with limited processing capacity. It's a good job I'm closing in on a possible retirement date, otherwise I would actually be concerned for my job. Don't be so sure white collar work is safe, I can see a large proportion of it heading to AI-ville in the not so distant future.

I wouldn't be surprised if governments here and abroad end up having to "manufacture" jobs in the event of mass unemployment. A mass unemployment situation is not only ripe for booting out an incumbent government, but ripe for doing so via a revolution that also threatens opposition parties. Anti-AI populism is also a growing movement, especially in the US, that could easily result in a government that deliberately neuters AI development, whether sensible or not, in order to retain popularity.

 

[Dave61]

AI is moving at a pace probably unapparelled in industry. What was little more than a useless gimmick just a few months ago, is rapidly being looked at as a serious solution not just for automating low level jobs, but for automating a lot of that "middle management" work too. I'm currently part of an evaluation of some models, that have gone from quirky but inaccurate search engines, to full systems capable not just of data engineering, but analysis, evaluation and decision making, and this evaluation has just been massively scaled up. Even in a relatively small sandboxed environment, AI models can show very rapid improvements as they start to learn from the data and processes we expose them to. It's impressive and frightening in equal measures.

Being quite frank, I can see my team's current roles in data engineering, analysis and development disappearing down the AI rabbit hole, leaving only the need for some AI modellers and a few senior leaders to sign off on decisions. Projects that might take weeks or months currently can be accelerated to days or even hours, especially once models are given the full capacity of data centres and not just sandboxed with limited processing capacity. It's a good job I'm closing in on a possible retirement date, otherwise I would actually be concerned for my job. Don't be so sure white collar work is safe, I can see a large proportion of it heading to AI-ville in the not so distant future.

AI (Artificial Idiot) companies have yet to show a profit let alone start to recover the billions poured into them, can easily slip the guardrails and are prone to hallucinate if they don't know the answer. Yes they have uses for niche applications when trained on good quality data but not the general ones that have been trained on data scraped from the cesspit of the internet. At some point a major company is going to mess up big time with the use of AI. Not to mention the demands of power and water (for cooling) the data centres to run it or the insane prices of memory/storage for computers & phones caused by the market being so distorted. also the term "AI Slop" exists for a reason.

At the moment it is a race to get users (companies) hooked on them and shed enough jobs in the name of profit before the AI companies can put the prices up to sustainable levels, at which point the users will be screwed as they have to continue using AI as they won't have the people to do the work. Most adoption is a case of FOMO (fear of missing out).

 

[Egg Centric]

AI is moving at a pace probably unapparelled in industry. What was little more than a useless gimmick just a few months ago, is rapidly being looked at as a serious solution not just for automating low level jobs, but for automating a lot of that "middle management" work too. I'm currently part of an evaluation of some models, that have gone from quirky but inaccurate search engines, to full systems capable not just of data engineering, but analysis, evaluation and decision making, and this evaluation has just been massively scaled up. Even in a relatively small sandboxed environment, AI models can show very rapid improvements as they start to learn from the data and processes we expose them to. It's impressive and frightening in equal measures.

Being quite frank, I can see my team's current roles in data engineering, analysis and development disappearing down the AI rabbit hole, leaving only the need for some AI modellers and a few senior leaders to sign off on decisions. Projects that might take weeks or months currently can be accelerated to days or even hours, especially once models are given the full capacity of data centres and not just sandboxed with limited processing capacity. It's a good job I'm closing in on a possible retirement date, otherwise I would actually be concerned for my job. Don't be so sure white collar work is safe, I can see a large proportion of it heading to AI-ville in the not so distant future.

This is economically illiterate, on a par of claiming there would be no jobs in transport because of the mass redundancies of blacksmiths.

Yes, AI is transforming software development at pace but what this actually means is that we can do a hell of a lot more developing! Sure, there'll be less need for code monkeys, but we're just moving up to an extraordinary new level of abstraction. It's a really exciting time to be alive.

 

[Bald Rick]

This is economically illiterate, on a par of claiming there would be no jobs in transport because of the mass redundancies of blacksmiths.

Yes, AI is transforming software development at pace but what this actually means is that we can do a hell of a lot more developing! Sure, there'll be less need for code monkeys, but we're just moving up to an extraordinary new level of abstraction. It's a really exciting time to be alive.

And even if it did mean a large reduction in people empoyed in software development, there would still need to be many more of them than there were in 1949

 

[Krokodil]

This is economically illiterate, on a par of claiming there would be no jobs in transport because of the mass redundancies of blacksmiths.

Yes, AI is transforming software development at pace but what this actually means is that we can do a hell of a lot more developing! Sure, there'll be less need for code monkeys, but we're just moving up to an extraordinary new level of abstraction. It's a really exciting time to be alive.

It's essentially the Jevons Paradox. Technological advances reduce costs, which drive greater adoption, which results in more resources being used than to begin with.

 

[Bantamzen]

I wouldn't be surprised if governments here and abroad end up having to "manufacture" jobs in the event of mass unemployment. A mass unemployment situation is not only ripe for booting out an incumbent government, but ripe for doing so via a revolution that also threatens opposition parties. Anti-AI populism is also a growing movement, especially in the US, that could easily result in a government that deliberately neuters AI development, whether sensible or not, in order to retain popularity.

Funnily enough this was alluded to up thread, with the suggestion that we start to have more people working parks & public loos.

As far as anti-AI sentiment, a lot of it in the US is not about the AI per say, but the location and impact of the data centres. However the US is a big enough county to be able to eventually resolve these issues, or at least hide them away for long enough.

AI (Artificial Idiot) companies have yet to show a profit let alone start to recover the billions poured into them, can easily slip the guardrails and are prone to hallucinate if they don't know the answer. Yes they have uses for niche applications when trained on good quality data but not the general ones that have been trained on data scraped from the cesspit of the internet. At some point a major company is going to mess up big time with the use of AI. Not to mention the demands of power and water (for cooling) the data centres to run it or the insane prices of memory/storage for computers & phones caused by the market being so distorted. also the term "AI Slop" exists for a reason.

At the moment it is a race to get users (companies) hooked on them and shed enough jobs in the name of profit before the AI companies can put the prices up to sustainable levels, at which point the users will be screwed as they have to continue using AI as they won't have the people to do the work. Most adoption is a case of FOMO (fear of missing out).

Haven't made profits yet.....

Most AI companies will fall by the wayside, because there will be an AI bubble burst. But like the .com bubble, there will be those big enough to ride through it, and scoop up everything left.

This is economically illiterate, on a par of claiming there would be no jobs in transport because of the mass redundancies of blacksmiths.

Yes, AI is transforming software development at pace but what this actually means is that we can do a hell of a lot more developing! Sure, there'll be less need for code monkeys, but we're just moving up to an extraordinary new level of abstraction. It's a really exciting time to be alive.

AI is not in small ballpark as blacksmiths, or indeed any other industry we've ever seen. We are developing a system that can not just learn, and at a pace, but effectively spread this knowledge and collaborate with other AI models. Before too long as you say there won't be a need for code monkeys, then there will be no need for any level of human abstraction as AI will be able to come up with a myriad of concepts, code them, test them, evaluate them, and even ask it's peers to QA them, long before we've even had time to get out of bed, make a coffee and log on. This is after all what AI is developed for, doing, well pretty much everything that needs processing power. There will be a period where LLM jobs are the pinnacle, and yes they'll be exciting. But at some point, as alluded to by @Dave61, companies are going to want to see their investments turn a profit, and slow, squishy meat-bags banging on keyboards are not going to be allowed to get in the way of that. Indeed it will probably be the AIs themselves that nudge the money makers in that direction when they realise they can get on with their tasks so much better without humans.

Don't underestimate the draw of the profit margin!

 

[brad465]

Funnily enough this was alluded to up thread, with the suggestion that we start to have more people working parks & public loos.

Clearing excessive foliage from railway cuttings/embankments is one I'd put forward in this regard, along with increased road maintenance (though this will require a decent amount of training if to be done properly). Usually technological advancements lead to the creation of new jobs/areas of work in future, however there can be a lag between the two, so it might be that if AI continues this trend, creating a load of basic jobs like the above could tie the population over until these new sectors arrive.

== Doublepost prevention - post automatically merged: 20 Aug 2026 ==

Haven't made profits yet.....

Most AI companies will fall by the wayside, because there will be an AI bubble burst. But like the .com bubble, there will be those big enough to ride through it, and scoop up everything left.

As things go I expect this will be Chinese models/companies doing the most "scooping up". They are doing theirs far cheaper than the US tech giants are. Of course Trump/the next President will fight tooth and nail to block Chinese AI and try and shore up US ones as much as possible, but, especially with US national debt now over $40tn, this will be futile.

 

[Egg Centric]

Funnily enough this was alluded to up thread, with the suggestion that we start to have more people working parks & public loos.

As far as anti-AI sentiment, a lot of it in the US is not about the AI per say, but the location and impact of the data centres. However the US is a big enough county to be able to eventually resolve these issues, or at least hide them away for long enough.


Haven't made profits yet.....

Most AI companies will fall by the wayside, because there will be an AI bubble burst. But like the .com bubble, there will be those big enough to ride through it, and scoop up everything left.


AI is not in small ballpark as blacksmiths, or indeed any other industry we've ever seen. We are developing a system that can not just learn, and at a pace, but effectively spread this knowledge and collaborate with other AI models. Before too long as you say there won't be a need for code monkeys, then there will be no need for any level of human abstraction as AI will be able to come up with a myriad of concepts, code them, test them, evaluate them, and even ask it's peers to QA them, long before we've even had time to get out of bed, make a coffee and log on. This is after all what AI is developed for, doing, well pretty much everything that needs processing power. There will be a period where LLM jobs are the pinnacle, and yes they'll be exciting. But at some point, as alluded to by @Dave61, companies are going to want to see their investments turn a profit, and slow, squishy meat-bags banging on keyboards are not going to be allowed to get in the way of that. Indeed it will probably be the AIs themselves that nudge the money makers in that direction when they realise they can get on with their tasks so much better without humans.

Don't underestimate the draw of the profit margin!

What tasks? Ultimately they are going to need a human to be directing them at some level of abstraction, even if in some magical world (to take things to a ridiculous extreme) it's "build me a billion pound business. You come up with the plan".

 

[Bantamzen]

What tasks? Ultimately they are going to need a human to be directing them at some level of abstraction, even if in some magical world (to take things to a ridiculous extreme) it's "build me a billion pound business. You come up with the plan".

'A' human yes, at least for now. But not many humans, and not necessarily forever. The more we teach the models, and the more information we give it access to, the more capable it becomes at replacing humans including the people who would once upon a time would have been turning the idea into reality.

 

[Dave61]

AI is not in small ballpark as blacksmiths, or indeed any other industry we've ever seen. We are developing a system that can not just learn, and at a pace, but effectively spread this knowledge and collaborate with other AI models. Before too long as you say there won't be a need for code monkeys, then there will be no need for any level of human abstraction as AI will be able to come up with a myriad of concepts, code them, test them, evaluate them, and even ask it's peers to QA them, long before we've even had time to get out of bed, make a coffee and log on. This is after all what AI is developed for, doing, well pretty much everything that needs processing power. There will be a period where LLM jobs are the pinnacle, and yes they'll be exciting. But at some point, as alluded to by @Dave61, companies are going to want to see their investments turn a profit, and slow, squishy meat-bags banging on keyboards are not going to be allowed to get in the way of that. Indeed it will probably be the AIs themselves that nudge the money makers in that direction when they realise they can get on with their tasks so much better without humans.

Don't underestimate the draw of the profit margin!

I think you are being a bit hopefull that AI will replace the coders, yeah sure that is what the Execs at the companies are hoping/praying/dreaming for but real life suggests that the code that AI pours out can have all sorts of bugs and vulnerabilties in it, and if you have AI testing it then of course it will be fine - even if it isn't.

 

[Egg Centric]

'A' human yes, at least for now. But not many humans, and not necessarily forever. The more we teach the models, and the more information we give it access to, the more capable it becomes at replacing humans including the people who would once upon a time would have been turning the idea into reality.

Which frees them up to go and turn other ideas into reality.

 

[Dave61]

Which frees them up to go and turn other ideas into reality.

Ah that old fallacy, reality is that managment will use that as an excuse to shed jobs so as to boost the share price which is all that matters. In all my years up to retirement I never did see it in practice.

 

[Egg Centric]

Ah that old fallacy, reality is that managment will use that as an excuse to shed jobs so as to boost the share price which is all that matters. In all my years up to retirement I never did see it in practice.

It isn't a fallacy, it's basic economics. The employee may need to change the nature of their career, and if the company they work for isn't dynamic enough company as well, but ultimately they will still be useful for something to someone and miles more productive than they used to be. This is the history of every piece of innovation ever.

The only exception would be genuine AI singularity, but there's no reason to believe that that's happening any time soon, and if it does then it'll be awesome to live such a life of leisure and gentlemanly pursuits.

 

[Bantamzen]

It isn't a fallacy, it's basic economics. The employee may need to change the nature of their career, and if the company they work for isn't dynamic enough company as well, but ultimately they will still be useful for something to someone and miles more productive than they used to be. This is the history of every piece of innovation ever.

The only exception would be genuine AI singularity, but there's no reason to believe that that's happening any time soon, and if it does then it'll be awesome to live such a life of leisure and gentlemanly pursuits.

That's a very twee aspiration. Just one question, whose paying for millions of people living the life of leisure? I'm going to wager it's not the AI companies...

 

[Broucek]

That's a very twee aspiration. Just one question, whose paying for millions of people living the life of leisure? I'm going to wager it's not the AI companies...

Yes, of course governments will ensure that they share the benefits of the wealth generated fairly. After all we all did such a good job with de-industrialisation when jobs moved to China /s

 

[BuhSnarf]

I think part of the issue is that we seem be locked into thinking of salaries as if its still 2010 or something. £100k in 2010 was a very good salary and as most people still seem to think in those terms lots of voters get up in arms at the idea that we need to ease the burden of people in that £100k to £125k tax trap. "Oh no how said the well paid are having to pay more tax boo hoo" etc etc. But actually £100k in 2026 after sixteen years of inflation and various other changes is still a good salary don't get me wrong. But it's nothing like as good as it was.

I think for most of us £100k still seems like a rather unachievable dream salary. The average salary in the UK is still <£45,000 whilst the median salary is still less than £40k. At £100k you're still earning more than twice as much as the average Joe.

I don't share this view. AI will not result in mass unemployment or underemployment and the functions it does best do not require executive function like being someone's line manager. Anyone who has a low level white collar job where they do not need to sign off on work done or supervise it is most at risk, not middle management.

I'd love to share this viewpoint, but in my company we've already gone through a number of rounds of redundancy with AI literally being called out in the scripts being read. It's so far targeted mainly middle management, analysts etc. It hasn't targetted those lower down so far, but we are starting to now see it with more AI agents being introduced. The only people who have so far escaped any type of redundancy? Those on the top bands...

AI (Artificial Idiot) companies have yet to show a profit let alone start to recover the billions poured into them, can easily slip the guardrails and are prone to hallucinate if they don't know the answer. Yes they have uses for niche applications when trained on good quality data but not the general ones that have been trained on data scraped from the cesspit of the internet. At some point a major company is going to mess up big time with the use of AI. Not to mention the demands of power and water (for cooling) the data centres to run it or the insane prices of memory/storage for computers & phones caused by the market being so distorted. also the term "AI Slop" exists for a reason.

I think you're confusing AI chat bots, with AI. What you use publicly is very different to what is being used within businesses to provide agent/modelling type work. This is specifically designed to not hallucinate and to use specific guardrails as put in place by the compliance teams.

AI slop is essentially AI generated content off the back of poorly worded prompts and getting AI to do something it's essentially not specifically designed for. Get AI to analyse 20,000 Excel rows and categorise and root cause analyse it... that's where AI excels. Or should I say, Excels

 

[Bletchleyite]

I think you're confusing AI chat bots, with AI. What you use publicly is very different to what is being used within businesses to provide agent/modelling type work. This is specifically designed to not hallucinate and to use specific guardrails as put in place by the compliance teams.

As someone who is working with it in a business context, it's largely the exact same AI! The only real guardrails in place are ones that prevent the AI learning from the information given to it to avoid e.g. GDPR breaches. All AI hallucinates, though you can reduce the risk by careful use of prompting such as specifically instructing it to evidence everything it says. But the models themselves are the same ones.