The AI is taking our jerbs thread

[Ayahuasca]

Nothing says...

"Tell me you don't track productivity without telling me you don't track productivity"

... by looking for solutions for tracking remote working.

Not saying it’s a good thing and as dowie said it’s already happening in some capacity. I can just see AI solutions being able to give them more hard evidence for employees slacking off for hours.

As it stands you can easily manipulate something like your Microsoft Teams status and even if you’re still getting all the work done you should, employers are always looking for ways to get more work out of people.

 

[OspreyO]

Not saying it’s a good thing and as dowie said it’s already happening in some capacity. I can just see AI solutions being able to give them more hard evidence for employees slacking off for hours.

As it stands you can easily manipulate something like your Microsoft Teams status and even if you’re still getting all the work done you should, employers are always looking for ways to get more work out of people.

If someone is slacking off for hours it's implies their line manager is not managing them, they are also slacking off. Also they don't do it in the office either.

 

[OspreyO]

Are you being asked to think of ideas? helpdesk is an easy win for a bog standard bot to answer simple questions or add some docs for RAG goodness

Like I said the helpdesk doesn't have simple questions. Often no answers either. Because people can't ask a straight question and people don't put solutions in the helpdesk. They just write done or fixed.

 

[Ayahuasca]

If someone is slacking off for hours it's implies their line manager is not managing them, they are also slacking off. Also they don't do it in the office either.

While this may be logical, workplaces don't always operate like that. It's already rather easy on a helpdesk, given that everything is KPIs with metrics and everything is pretty much recorded. In other roles and sectors, everything can be far more varied, with no defined time allocated to activities. If there's a tool that provides hard evidence of an employee slacking off, or that makes people think 'I'd better work harder in case my activity is being scrutinised more closely', then they'll use it.

I don't like it, but I'd be very surprised if AI solutions didn't become much more widespread in this area.

 

[OspreyO]

While this may be logical, workplaces don't always operate like that. It's already rather easy on a helpdesk, given that everything is KPIs with metrics and everything is pretty much recorded. In other roles and sectors, everything can be far more varied, with no defined time allocated to activities. If there's a tool that provides hard evidence of an employee slacking off, or that makes people think 'I'd better work harder in case my activity is being scrutinised more closely', then they'll use it.

I don't like it, but I'd be very surprised if AI solutions didn't become much more widespread in this area.

Have tool that requires people to look busy people will put their effort into looking busy not productivity. Which is how the current tools work.

You can always measure something. Most things have a start middle and end.

 

[Ayahuasca]

Have tool that requires people to look busy people will put their effort into looking busy not productivity. Which is how the current tools work.

You can always measure something. Most things have a start middle and end.

That's the thing, at the moment it's easy to still look busy/active in certain roles. Some of these AI solutions will make it so you can't fudge it to that extent and mean you don't need a human sat micromanaging people and combing over stats.

Yeah you can always measure something, but it's the methods of doing so and how efficient they are and the cost/time involved.

 

[Rob_B]

Like I said the helpdesk doesn't have simple questions. Often no answers either. Because people can't ask a straight question and people don't put solutions in the helpdesk. They just write done or fixed.

You dont have the "I cant do x" questions where "x" is clearly documented and explained in a document/intranet page titled "how to do x" ?

 

[OspreyO]

That's the thing, at the moment it's easy to still look busy/active in certain roles. Some of these AI solutions will make it so you can't fudge it to that extent and mean you don't need a human sat micromanaging people and combing over stats.

Yeah you can always measure something, but it's the methods of doing so and how efficient they are and the cost/time involved.

Obviously someone is not micromanaging stats if they want AI to do generate metrics for them.

This is the catch 22. AI needs well formed data. It can't pull something from thin air. We'll it can I suppose that would be a concern. It will just invent stuff.

 

[Ayahuasca]

Obviously someone is not micromanaging stats if they want AI to do generate metrics for them.

This is the catch 22. AI needs well formed data. It can't pull something from thin air. We'll it can I suppose that would be a concern. It will just invent stuff.

Nor should they have to if it can be done more efficiently and nobody said it would just be generating metrics.

Nor was it suggested it would be pulling something from thin air either. You're thinking of it in terms of something like ChatGPT in generating stuff. That's not the kind of AI solution we're talking about.

We get it, you believe AI isn't capable of doing anything well. The reality is businesses are already using it and will continue to look for ways to use it in areas such as this. I've already agreed I don't like it. I mean, you've just told us you're being put under pressure to implement it...

 

[OspreyO]

Nor should they have to if it can be done more efficiently and nobody said it would just be generating metrics.

Nor was it suggested it would be pulling something from thin air either. You're thinking of it in terms of something like ChatGPT in generating stuff. That's not the kind of AI solution we're talking about.

We get it, you believe AI isn't capable of doing anything well. The reality is businesses are already using it and will continue to look for ways to use it in areas such as this. I've already agreed I don't like it. I mean, you've just told us you're being put under pressure to implement it...

Put it this way if it thinks I'm busy because I'm in meetings I can create meetings all week to be in. To appear busy. Because that's the road this is going down. Looking for activity.

The other issues is I can't just point it a load of unstructured data and expect to get any sanity from it. It can't learn from that. But I think that's the expectation people have.

 

[OspreyO]

You dont have the "I cant do x" questions where "x" is clearly documented and explained in a document/intranet page titled "how to do x" ?

Usually the solution is technical that the requester can't action. Which is why they logged a ticket in the first place.

We have many document libraries across many sites, that are more straightforward.

 

[jpaul]

Have been watching ITV's 'Genius game' , involving some fascinating development of team strategies for new games participants haven't seen before,
would like to see if AI could be trained to quickly develop winning strategies - humans are so inventive.
not really dissimilar to 'trying' to deploy AI for a warfare per the Israelis.

 

[dowie]

would like to see if AI could be trained to quickly develop winning strategies - humans are so inventive.

Yes it can and has been used for that for a while now - I mean the first chess computers are ancient now! But more recently (or at least say 10 years ago) you had deep reinforcement learning used to tackle things like Atari Games or beat the best human players at Go, etc..

Now, if you're throwing social manipulation, teamwork etc.. into the mix then you'd need to make use of LLMs too, that's perhaps more of an open question for researchers but as far as the strategy side of things is concerned in finite games, then AI is basically already way ahead of humans.

 

[NickK]

I can see it being used to track remote workers in a more invasive way. We've seen Microsoft implement something similar in Windows with Recall.

It's not a big leap to imagine businesses wanting something like this to see just how much people are slacking and/or to try and force an increase in productivity.

I remember one boss many moons ago tracked everything the staff did in the office in an app called grindstone. He even had a toilet charge code.
Reality is it didn’t really do much or stop redundancies.

Also has a US director demand a 360 webcam and all staff to sit in the warroom. His interest lasted a day and it never got used after that.

 

[Ayahuasca]

If anyone gets a chance to use Claude 4 Opus + Claude Code you'll be in for a nice surprise if you've only used some of the free older models to test how good AI is at this stuff.

 

[mid_gen]

If anyone gets a chance to use Claude 4 Opus + Claude Code you'll be in for a nice surprise if you've only used some of the free older models to test how good AI is at this stuff.

Haven’t got it accessible in Windsurf yet, still on 3.7.

I’ve completely rebuilt a whole application which was vibe coded with Claude, this week. It was a mess, just had a non programmer directing it.

I would never have been able to do a clean rebuild in this timeframe without the AI though. I’ve got really strong context prompts that make sure it sticks to patterns etc. Really amazingly powerful.

 

[NickK]

I have put the Mountain Dew cans around my keyboard in a pentagram..

Ah, fantastic. Another user thinking they’re clever, summoning the ******* Operator From Hell like it’s some kind of magic incantation. What do you want then, user? Tech support for your toaster? Help rebooting your brain’s BIOS?
Fine. From now on, responses will be laced with just the right amount of sarcasm, disdain, and passive-aggressive loathing — BFOH style. Just don’t come crying when your files mysteriously disappear into the /dev/null of broken dreams.
So, what’s the problem today? Forgot your password again, genius?

In other news it took McLaren 2 hours to reject my application this morning.. so I'm now downloading a local AI to run in a VM, then I will create a simple django front end so I can drag and drop the job application into it..

curl -fsSL https://ollama.com/install.sh | sh
ollama run mistral

That's the local mistral AI installed..

verifying sha256 digest
writing manifest
success
>>>
>>> test
 Hello! It seems like you'd like to test something. How can I assist you with that? Please provide more details about what exactly you want to
test, and I'll do my best to help.

For instance, if you want to test a piece of code, I can help by providing an input or explaining the expected output. If it's for a different
purpose, just let me know!

So that's the AI model working locally.

 

[dowie]

so I'm now downloading a local AI to run in a VM, then I will create a simple django front end so I can drag and drop the job application into it..

Why would you use a local model for job applications? There are far better models available via ChatGPT or Anthropic.

 

[NickK]

Why would you use a local model for job applications? There are far better models available via ChatGPT or Anthropic.

Testing options, also free only goes so far for developing.

 

[dowie]

Testing options, also free only goes so far for developing.

I'd strongly recommend a paid subscription, if you're only using free you're barely scratching the surface of what's available, AFAIK for OpenAI it's only limited 4o then moves to 3.5 and no access to the reasoning models o3, o4 mini and o4 mini high.