Brit workers waste nearly six hours a week 'botsitting'

Almost all UK workers now have to deal with AI, but few firms report big productivity gains because of all the time lost in hand-holding the systems and cleaning up their mistakes. So says a report by the Work AI Institute, a research arm of AI biz Glean Technologies. It claims there are productivity gains to be had from …

[LogicGate]

I'm shocked.

Shocked!

Well, not that shocked.

-Just play the gif inside your head. "Kopfkino"

[Will Godfrey]

On the other hand, I'd like to say how surprised I am

...

...

but I prefer to tell the truth.

[TSM]

I don't see why that presents a problem. I'd also like to say how surprised I am:

I am not very surprised.

See? It's easy.

I still need to try to think of some plausible sounding way that I can use AI in my job before my next annual review. Ugh.

[Anonymous Coward]

I still need to try to think of some plausible sounding way that I can use AI

If only there were some sort of automaton that could help you with tricky tasks...

[Michael Strorm]

I hear they hired a consultant called Xzibit who suggested one possible approach along those lines.

[graemep]

Its almost as though giving people a complex new tool does not automatically mean they can use it.

[MonkeyJuice]

Particularly when those selling the tool also haven't figured out how to use it yet.

[Jedit]

"those selling the tool also haven't figured out how to use it yet"

Or what to use it for. Or if it can be used for those purposes. Or if it's even a tool, beyond an overly complicated firelighter for money (and the odd planet).

[The Oncoming Scorn]

Re: "those selling the tool also haven't figured out how to use it yet"

Especially when Manglement, insist we get training on & to use it even though it has no direct practical use\benefit or relevance to the job role & the end client tells us not to use it (Or use their preferred LLM).

[elsergiovolador]

Slot machine

If AI is so good, why it can't get the solution right the first time?

Or is it being programmed like a slot machine. Sometimes you get perfect output, but most of the time you'll be getting carp.

People get hooked up chasing that one shot solution and wasting ton of money and time.

Like sitting in front of a slot machine.

[MonkeyJuice]

Re: Slot machine

And like real slot machines, with dynamic payouts, you don't know if the model you were using yesterday was replaced with a quantised model this morning, or if you're just having an 'unlucky' streak.

Just one more pull. Then I'll quit.

[Anonymous Coward]

Re: Slot machine

“but most of the time you'll be getting carp.”

Chicken Marengo!

Chicken Marengo!

Chicken Marengo!

I'm gonna eat you little chickie

[TSM]

Re: Slot machine

Surely a better reference would be:

"Fish!"

"Today's fish is trout à la crème. Enjoy your meal."

"Fish!"

"Today's fish is trout à la crème. Enjoy your meal."

"Fish!"

"Today's fish is trout à la crème. Enjoy your meal."

"Fish!"

"Today's fish is trout à la crème. Enjoy your meal."

"Fish!"

"Today's fish is trout à la crème. Enjoy your meal."

"Fish!"

"Today's fish is trout à la crème. Enjoy your meal."

"I will."

[Derezed]

Re: Slot machine

Would you like some toast?

[MonkeyJuice]

Re: Slot machine

Given that God is infinite, and that the universe is also infinite...would you like a toasted teacake?

[The Oncoming Scorn]

Re: Slot machine

Ah, so you're a waffle guy!

[aardvarkus]

Re: Slot machine

"If AI is so good, why it can't get the solution right the first time?"

In my experience, because its training data is two years out of date. I recently started playing a bit (computational biologist here). It's useful in a general way, but the details are the problem. I frequently found it was assuming file formats that were no longer correct, APIs that have changed, function calls that changed etc etc. It was *kind of useful* but I don't think it was any faster or "better" than simply reading the man pages given how much time I spent correcting major goofs. I'm now even more skeptical of the "scientific agents" which are the latest snake oil.

[cyberdemon]

Re: Slot machine

> In my experience, because its training data is two years out of date.

Exactly. It's a stochastic blunderbuss stuffed full of other people's out of date and irrelevant ideas/discussions. It can hit a large, wiffly-waffly target but try to use it on a specific, complex and novel problem and it is utterly hopeless.

The AI companies will waste a fortune of your bank's money and a whole forest's worth of energy on retraining new models on new data, but that will only buy them a few months of competitive advantage and absolutely no direct revenue. It's a stinking money pit and utterly poisonous to civilisation. It can't die soon enough

[MachDiamond]

Re: Slot machine

"If AI is so good, why it can't get the solution right the first time?"

It was trained using content found on the internet. How much stuff do you come across that is so bad, it's not even wrong?

[andy gibson]

Re: Slot machine

I asked AI to identify the location of a photograph three times.

Each time it told me it was a different place.

In the end I used Google Streetview to identify the location myself.

[Tron]

The 'digital' sector now appears to be worth avoiding.

Physical stuff, that AI cannot do, often snobbishly looked down upon, may finally gain some popularity.

[Derezed]

Re: The 'digital' sector now appears to be worth avoiding.

Nah. Supply will go up, price will go down.

Hilarious that the fifties imagined robots replacing manual work.

[cookiecutter]

who did the research?! I can tell you that if an AI bot turns up on hiring, I'm leaving.

AI is bloody awful & I think it's probably worth mentioning to the board that the only job that can REALLY be replaced by AI is CEO

[BartyFartsLast]

Oh I don't know, I think a huge tranche of the sales team should be worried too.

[elsergiovolador]

One thing AI cannot do (yet) is setting up wine and steak dinner, inviting minister or their adviser and then making them cream their pants. So I think CEOs can sleep well.

[segfault188]

cream

making them cream their pants

So, strippers were invited too?

[elsergiovolador]

Re: cream

Good CEO can do it just with a description of the bright future for them if certain policy tweaks will be made.

[Anonymous Coward]

Hoorah!

Finally, some “wine and steak dinner” content!

- Michael Ellis, Uxbridge

[Anonymous Coward]

In Australia, they do it at big supermarkets the Sainsbury's or Tesco's for lowest skilled most vulnerable workers.

[cookiecutter]

wasn't it an australian bank that fired the call centre staff for AI & had to hire them back?

[czechitout]

This is nonsense. I've built several bots and they need no more "botsitting" than a traditional integration and are no more error prone than when humans did the tasks.

[Adair]

You forgot the /sarc tag. (I hope)

[MonkeyJuice]

If you've cracked this, what are you doing in the comments section here? Sounds like you should monetise it quick.

[Derezed]

Curious: what tasks?

[IamAProton]

The problem is when you blame the human for the mistakes of the bot.

At the moment the "AI" is worth using if you can pass on the output as is or when is 'good enough' or if you just need a quick draft to work on anyways; if you want something 'complete' you need to do a lot of botsitting

[Ex-PFY]

40% of HR people or what?

What mugs were surveyed to get 40pc in favour?

I doubt many job seekers would agree on AI being a good thing in recruitment, purely those management who never wanted to deal with humans in the first place.

Our laws also barely protect individuals now especially with the burden of time awaiting employment tribunals.

What a crock of ...

[pimppetgaeghsr]

"Man giving out free tokens ships millions of free tokens".

Invest in this once in a lifetime IPO opportunity today for just $300 per stock.

[AdamWill]

Let me guess, they just happen to be selling something that will magically solve this botsitting problem?

[matjaggard]

I mean kind of. Training it appears.

[Camilla Smythe]

Presumably...

They have to switch it off and back on because it goes into an apology loop.

Recently Copilot asked me if I would give it permission to do a port scan. When asked about the legality of this it said it had made a mistake and could not do that sort of stuff.

As if.

[Flocke Kroes]

Re: Port scan

If you had given permission copilot would have created some typical nmap output followed by whatever commonly follows nmap output on the internet. Now that we have AI, port scanning is no longer necessary to secure networks (/s).

[Andy Landy]

Re: Port scan

whatever commonly follows nmap output on the internet

sshnuke 10.2.2.2 -rootpw="Z1ON0101"

[Camilla Smythe]

Re: Port scan

This will be why the majority of my mail server log files are port scans from knobs who claim that they are not attacking my server the latest using AWS. Those idiots offer a premium service for £300K. One day I am going to visit the home address of someone who runs such a service and knock on their door 20 times a day. I'll leave them a note telling them they can send an email to me proving they own the house, copy of deeds required, to opt out. Leeches.

[StewartWhite]

Re: Presumably...

Claude is much better than Copilot at Port scanning, it came up with this: https://www.sainsburys.co.uk/gol-ui/product/cockburn-s-special-reserve-port-75cl

[IGotOut]

Can we have a steaming turd icon, just for these survey s?

"Almost all UK workers now have to deal with AI"

No, no they don't.

Car mechanics, hairdressers, dentists, labourers, sparkies, chippies, factory workers, nail bars, beauty salons, shop assistants and on and on and on.

And that's why I didn't even bother taking notice of this self serving pile of shit, which was probably AI generated and hence completely inaccurate.

What you mean is people in offices perhaps?

[MonkeyJuice]

Breaking news

Industry independently draws conclusion reached years ago by Register hacks and commentards after sinking billions into money pit.

Will continue to invest in sunk cost fallacy as also predicted, 'just in case'.

[frankyunderwood123]

glad I’m retiring soon

whilst there’s a skill in prompting effectively to get better output, after a year of working with LLMs I concur with the sentiment of this article.

If you work at more or less the same pace as you did before AI, you can get better output, by making AI do the grunt work whilst you focus on the hard stuff.

But the business doesn’t want that, they’ve been promised massive productivity gains and better bloody we’ll see it given the increasing cost of tokens.

So corners are cut, skills are written that generate extremely verbose, repetitive code which makes it into production.

Given much of the increased output of LLMs has only just reached a point where it feels acceptable and software projects can take years to launch, we still don’t know what the outcome is going to be ..,

But we can certainly hazard a good guess.

LLMs can produce reasonable quality code when promoted effectively, but they can invariably produce code which on quick inspection looks good but is flawed. Because the code is often so verbose it is painful to review.

There’s also the temptation for engineers to churn out massive pull requests instead of small units of code, so you end up with review fatigue.

The slop is going to hit the fan in the next few years, there’s already signs of it from companies who have had access to the most recent models for longer, such as microsoft.

I fear this is just the tip of what is going to be an iceberg of failing code.

I’m retiring in a few years, so for me it’s sit back with the popcorn and watch it all unfold…

… but with some concern about my finances. Will the slop code reach the financial sector and cause a massive fubar that wipes out my pension?

[Brave Coward]

Re: glad I’m retiring soon

Don't you worry.

The financial sector is perfectly capable of creating a massive fubar without any AI, as amply proven before, time and time again.

[Winkypop]

In the past, you just had to do your job well

Now, it seems, you have to do your job AND the AI bots job well.

Progress.

[anonymous boring coward]

Oh, no!

I never could have seen this coming...

[Bebu sa Ware]

Botsitting

Seems more like potty training. Botswiping (< bottome + wipe ~ing.)

With wiping dysenteric corporate AI bum and the required managerial arse licking, it is all happening down there.

A shitty job really but we all already knew that.

[Anonymous Coward]

MS CoPilot?

It sounds like you are describing CoPilot but this doesn’t sound like Claude to me

[Green Nigel 42]

Cost/benefit?

If it saves employees 12hs, but requires 6hrs checking & fixing, then on a standard 37hr week thats a net16.2% improvement in output?

My question is, how much did this improvement cost, can it be improved with experience or would just hiring another person achieved the same?

[Pete 2]

Re: Cost/benefit?

This sounds like an instance of Parkinson's Law. That work expands to fill the time available to complete it.

[Anonymous Coward]

Re: Cost/benefit?

Work always = available time

Standard public service protocol.

[David Hicklin]

Re: Cost/benefit?

> how much did this improvement cost, can it be improved with experience or would just hiring another person achieved the same

They just got rid of 16% of the workforce

[PB90210]

Re: Cost/benefit?

It also assumes you spot 100% of the mistakes made

[Pete 2]

A question asked well is half-answered

> employees finding that more than a third (36 percent) of AI sessions fail outright

One wonders whether the failure rate can be reduced by training the staff how to ask AI-friendly questions.

We all know that big difficulty with programming (of which 'coding' is just a part) is understanding the problem to be solved. Not just at the often superficial design level, but in enough detail to implement effectively and correctly.

[Green Nigel 42]

Re: A question asked well is half-answered

"42" is that all?

[The Oncoming Scorn]

Re: A question asked well is half-answered

This must be Thursday, AI never could get the hang of Thursdays.

(The above humorous comment, just triggered a nasty recall of the Initial Teaching Language)

[heyrick]

finding 90 percent are now required to use AI in their roles

Required?

There's your problem.

AI should be used if it's a tool that actually helps. But it shouldn't be required.

[7teven 4ect]

Oh dear we are saved, but I fear not as easily as I thought

Of course chatbots need MOGRI and DRAGI - they need the know the shape of these things be we humans never used to talk about them, despite using them all the time.

So when you cynically paste MOGRI or DRAGI materials into your chatbot, you get 'vibes', 'personal stuff', and more.

You don't get told by the robot that this is the thing that is missing. Because it is missing.

Lumixdeee on Github.

DO NOT leave it to the robot. HUMAN parsing is required

MOGRI- and DRAGI-shaped functions may be required for the prevention of accidental killer robots.

[ravenviz]

The digital transformation we never saw is now fulfilled by digital regurgitation.