Best Of

“This is a shocker. James Reed , CEO of the recruiter Reed, said on Times Radio the UK is in the grip of a graduate jobs crisis noting three years ago he had 188,000 graduate jobs on his books and today it was 55,000.

He encouraged middle class families to encourage their student kids to take up manual labour”

*cough* universities

*cough* doomed

There are still going to be university courses needed for those 55,000 who will do graduate jobs and the amount of manual labour needed isn't increasing much, so that is just a recipe for more unemployed and a higher benefits bill unless AI creates new jobs

Has anybody set out how AI actually creates new jobs? Unless it is in fact-checking the bullshit it spews out?

Well it's certainly having an effect on current jobs. I've a friend on a decent (professional) salary who has just been given the nod his job might go (see BBC Scotland news for a hint).

I've escaped into middle management but certainly won't be expanding/replacing my team in the next few years with a lot of pressure to use AI to take up the slack. I'm not sure how - part of the reason I can post on PB is we've automated most of our stuff already.

A friend of mine, who works for a leading law firm, has said his employer has been making redundancies because of AI. However, when he described the jobs in question (which sounded like just buggering around with PowerPoint) I was surprised you'd employ someone just to do that anyway. And wasn't there the 'Lazy Girl Jobs' phenomenon a few years ago?

https://www.theguardian.com/commentisfree/2023/jul/09/gen-z-lazy-girl-jobs-tiktok-work

I suspect the Lazy Girls are now stuffed.

Isn’t the problem with law firms specifically, that the clients aren’t willing to pick up bills of hundreds per hour for lawyers to write presentations.

Or is it the other way around, law firms realising they could pay the graduate £30k/year while billing her out at £250/hour?

As an executor I found that a solicitor's firm was wanting a non-trivial percentage of the total value of stocks and shares held by the deceased to work out the probate valuation on the day of death. That's a very simple job if one has access to the database.

I got the whole lot done by Sharedata for a very small percentage of the firm's quotation - including an insanely complex work-back to primary capital value of one shareholding through about half a dozen mergers, splits and share issues.

Lawyers offer two (sometimes three) things: 1) the knowledge of how to do things like this and 2) taking on liability for when they get things wrong. (You might have to sue them to take advantage of said liability insurance, but it is there.) Sometimes also 3) the imprint of a neutral & trusted third party on a potentially contentious transaction.

It’s that combination that you’re buying. Often at great expense admittedly, but you are getting more for your money than if you do it yourself.

“This is a shocker. James Reed , CEO of the recruiter Reed, said on Times Radio the UK is in the grip of a graduate jobs crisis noting three years ago he had 188,000 graduate jobs on his books and today it was 55,000.

He encouraged middle class families to encourage their student kids to take up manual labour”

*cough* universities

*cough* doomed

There are still going to be university courses needed for those 55,000 who will do graduate jobs and the amount of manual labour needed isn't increasing much, so that is just a recipe for more unemployed and a higher benefits bill unless AI creates new jobs

Has anybody set out how AI actually creates new jobs? Unless it is in fact-checking the bullshit it spews out?

Well it's certainly having an effect on current jobs. I've a friend on a decent (professional) salary who has just been given the nod his job might go (see BBC Scotland news for a hint).

I've escaped into middle management but certainly won't be expanding/replacing my team in the next few years with a lot of pressure to use AI to take up the slack. I'm not sure how - part of the reason I can post on PB is we've automated most of our stuff already.

If LLMs increase productivity, then we have the classic virtuous circle of cheaper goods and services enabling higher economic activity.

Used judiciously, in some cases they can. At current pricing.

I’m not aware of any LLM being charged for at the full cost - they are all running at a loss.

The real “is AI a bubble?” question is “how much real benefit vs the real cost”

Oh I agree that it could be an Agricultural/Industrial style revolution epochal transformation, and the economy will grow as a result. But those periods also had some terrible side effects on human welfare - at the very least, we're going to have increased levels of frictional unemployment.

I don’t see it being that. A number of analyses have been done - many jobs find a few percent improvement from LLM, not orders of magnitude stuff.

The non-LLM stuff of joining up disparate systems and creating flows, has more potential. The amount of manual patching and fiddling that goes on in systems that should be completely automated..

They are still so incredibly brittle. I was using them in the past week to design a load of slides for me. It certainly made a good stab at it, but then kept really wanting to place bullet point text over the title and also not realising it was over-shooting the page with bullet points that were too long. I had to spent several hours fighting with it to improve placement and in the end I just did it manually.

And ChatGPT5 is definitely a step back in easy tasks like summarise this paragraph and rewrite in x words. It does the classic under-grad trick rather than understand and paraphrase, it tries to just lop off words that breaks the meaning. And even when much more prompt engineers the result is very dry now. Funnily I checked the "personality" setting ChatGPT5 has and it wasn't even set to nerd, it was set to neutral that is supposed to more fun and engaging.

“This is a shocker. James Reed , CEO of the recruiter Reed, said on Times Radio the UK is in the grip of a graduate jobs crisis noting three years ago he had 188,000 graduate jobs on his books and today it was 55,000.

He encouraged middle class families to encourage their student kids to take up manual labour”

*cough* universities

*cough* doomed

There are still going to be university courses needed for those 55,000 who will do graduate jobs and the amount of manual labour needed isn't increasing much, so that is just a recipe for more unemployed and a higher benefits bill unless AI creates new jobs

Has anybody set out how AI actually creates new jobs? Unless it is in fact-checking the bullshit it spews out?

Well it's certainly having an effect on current jobs. I've a friend on a decent (professional) salary who has just been given the nod his job might go (see BBC Scotland news for a hint).

I've escaped into middle management but certainly won't be expanding/replacing my team in the next few years with a lot of pressure to use AI to take up the slack. I'm not sure how - part of the reason I can post on PB is we've automated most of our stuff already.

If LLMs increase productivity, then we have the classic virtuous circle of cheaper goods and services enabling higher economic activity.

Used judiciously, in some cases they can. At current pricing.

I’m not aware of any LLM being charged for at the full cost - they are all running at a loss.

The real “is AI a bubble?” question is “how much real benefit vs the real cost”

Oh I agree that it could be an Agricultural/Industrial style revolution epochal transformation, and the economy will grow as a result. But those periods also had some terrible side effects on human welfare - at the very least, we're going to have increased levels of frictional unemployment.

I don’t see it being that. A number of analyses have been done - many jobs find a few percent improvement from LLM, not orders of magnitude stuff.

The non-LLM stuff of joining up disparate systems and creating flows, has more potential. The amount of manual patching and fiddling that goes on in systems that should be completely automated..

https://x.com/RupertLowe10/status/1968195697943208102?s=19

Rupert's deportation petition heatmap is a useful guide to voting patterns in the next election. Discuss

Before I clicked on it, I thought it was going to be a heatmap on how popular deporting Rupert Lowe was....

Lol.
It is quite interesting though. Whilst Lowe is ex Reform i think it shows where they are going to pile up votes and where Lab or Tories might hang on - for the latter interesting that Harrogate (corrected) and Wetherby look less hard-core anti immigration as an example, or Derby Dales which might remain a LabCon marginal.

Its not a prediction of course but its the sort of supplemental data that would help more with constituency betting (nearer the time) than Baxter.

In his own constituency he has under 2% signatures. I don't have a yardstick for that.

But what is it with Carlisle?

The geographic spread of it 300,000 people's views on the joint top issue du jour and the leading party in the polls main draw is useful data.
In his constituency 5% of the 2024 turnout have signed. Thats fairly significant for a petition on mass deportation