Chatgpt - Seriously good potential (or just some Internet fun)

seems like it was just good fun - jeremy diamond giving it the thumbs down, the in-breeding share holding, like usoft buying into AMD - market correction coming in.
just as starmer betted on red.

the mkinssey report is interesting

KN1lYzu.png


despite

Many millennials aged 35 to 44 are managers and team leaders in their companies. In our survey, they self-report having the most experience and enthusiasm about AI, making them natural champions of transformational change. Millennials are the most active generation of AI users. Some 62 percent of 35- to 44-year-old employees report high levels of expertise with AI, compared with 50 percent of 18- to 24-year-old Gen Zers and 22 percent of baby boomers over 65 (Exhibit 4). By tapping into that enthusiasm and expertise, leaders can help millennials play a crucial role in AI adoption.
Click to expand...
 
An old report though :o The interest is huge because the returns are as well if they manage to get it right. I suspect it will remain as it is today a useful tool to help, assist and guide people in the given subject
 
I make print and play board games/card games for various groups and using photoshop to create card/templates. Spreadhseets to create text data to add onto these cards via an app called nandeck and it generates any number of cards/boards whatever you need to prototype and print and test games out.

I felt nandeck was too all over the place altough it is great to prototype. I went to chat today and asked if it could write me a python script to do a similar thing to nandeck.

A couple of hours later I have gone from adding an image under another image that had a transparent section (card template at top layer and card image -monster/item etc as the underneath so only show on the transparent part of the top template) to a near enough fully functioning card editing prototype bit of software. select size of canvas ie poker card 2.5x3.5in, loads and reads csv file for column header (row is individual cards and columns are elements that are added to cards like title, cost, power, health etc). draw a box on poker card sized canva add it as image and elect the cvs file header that contain link to image. repeat for other layered images. a layer box to rearrange layer up or down. create a box select text and point it towards csv header with text. etc, and generate. It creates all my cards with the correct text and images all 300dpi and sized correctly.

Still have a way to go to add new things but. GOD DAMN. I cant code for **** and have no idea what chat is churning out at me.

yeah, so. I am impressed.,
 
Jimbeam3678 said:
An old report though :o The interest is huge because the returns are as well if they manage to get it right. I suspect it will remain as it is today a useful tool to help, assist and guide people in the given subject
Click to expand...
it was referenced associated with diamond's comments ..
and there is a kings no clothes aspect for many companies, effectively subsidising others that have some success.

high levels of experise with AI ? .. despite seeing increasingly posts of copilot/chatgpt misinformation, partial truths.

recent interview I heard with social care company claiming to have saved NHS 1.8BN, I thought, by analysis of seemingly incidental medical symptoms from carers to predict what cared folks might have undiscovered risks.
[ I'm waiting to hear more disclosure of inappropriate AI use in Gaza situation too - irresponsibly causing deaths ]
 
jpaul said:
it was referenced associated with diamond's comments ..
and there is a kings no clothes aspect for many companies, effectively subsidising others that have some success.

high levels of experise with AI ? .. despite seeing increasingly posts of copilot/chatgpt misinformation, partial truths.

recent interview I heard with social care company claiming to have saved NHS 1.8BN, I thought, by analysis of seemingly incidental medical symptoms from carers to predict what cared folks might have undiscovered risks.
[ I'm waiting to hear more disclosure of inappropriate AI use in Gaza situation too - irresponsibly causing deaths ]
Click to expand...
I don't even know what this is meant to mean - I knew I shouldn't have replied
 
I do love chatgpt. I'm facing a family lunch which I don't want to deal with so asked chatgpt for advice, it gives me a thorough 8 pointer guide. And then I ask it 'can you turn this into an easy to remember poem?'

Sunday Lunch Truce

Arrive on time, bring something small,
Keep your hopes gentle, don’t expect all.
Talk light and easy, let heavy things slide,
Let calm be your armour, your quiet your guide.

If sharp little comments come stinging your way,
Just breathe and go neutral — don’t join in the fray.
Sit near the kind ones, stay steady, stay true,
And leave when you need to — that’s caring for you.

When home once again, let the tension unwind;
A walk, tea, or music brings peace back to mind.

A lunch isn’t forever — just get through the day,
With smallness, with softness, and stepping away.
 
A friend of mine called me today in a pinch because he runs a workshop with super heavy duty old industrial waterjet machines and they'd dropped off the network so he couldn't get a job done. I'm not a networking pro, but I popped over and did what I can to try and see what devices are actually connected to the router, see if we can figure out their IP, get them mapped, make sure theirs no conficts, we checked all the wire routes, he'd recently moved his computer to another office, so I figured it was a connection issue, but all cables seemed okay.

I've managed to fix the odd issue in the past for him, but couldn't figure this out. TBH I was about to give up, but literally just took some photos on my phone of the router admin screen (I could see one of the machines, but not the other), and photos of the tooltips of the old shortcuts he had on the desktop with their IP addresses and asked chatgpt about it with some context.

First try, it guessed that due to the nature of the machines it was likely that it was an SMB issue and told me it was likely due to a windows update and how to re-enable it. 2 mins later and a windows restart he was back in action. Was incredible, as otherwise I'd have felt terrible for not being able to help and he was so relieved to not let the customer down. It really came through for us today.
 
hearing a bbc podacst on algorithmic wage discrimination in usa, basically setting uber/bank nurses/fds wage rates based on their financial situation from credit-rating,
or, for uber customers price based on what they might be willing to pay ,
such a strategy can be extended to give endless fun for personalised/inflated insurance quotes, or car pricing.
 
interesting article on r4 pm today - schools trust head complaining about (low intelligence) parents writing legalistic complaints about aspects of their childrens schooling using chatgpt,
and being unprepared(incapable) to actually visit school and discuss the problem efficiently face to face;
director predicated that soon their AI might be talking to his AI.

--------------

meanwhile - Google customers want some kick-back from lost clicks on their web-sites because the google AI response had plagiarized their sites.
 
jpaul said:
hearing a bbc podacst on algorithmic wage discrimination in usa, basically setting uber/bank nurses/fds wage rates based on their financial situation from credit-rating,
or, for uber customers price based on what they might be willing to pay ,
such a strategy can be extended to give endless fun for personalised/inflated insurance quotes, or car pricing.
Click to expand...
I can't wait for the day I walk into a shop, my face is matached, and the prices adjusted to the maximum I'm prepared to pay.
 
Last edited:
American article on misuse of AI by parents in comms with schools, paralleling the uk situation

In the past, an email that might have simply expressed frustration with a disciplinary action or bad grade is now a catalog of serious legal allegations, said Rodrequez.

“A lot of times, it is going to be just a kitchen-sink approach. It’s going to have anything and everything that AI determined that the district could have possibly violated,” Rodrequez said, ranging from accusations of civil rights violations to IDEA infractions. AI “doesn’t know the specifics [of the situation] because that wasn’t put into the system.”

Those are typically the first clues for Rodrequez that he might be dealing with a complaint written by a chatbot. He’ll take several of the same steps he would with any other complaint, like reviewing it and checking if anyone else in the district has already had contact with the parent. But these complaints can be very long and detailed, and responding to every issue raised can take a while.
Click to expand...
 
Claimed AbilityEvidence-based RealityVerdict
Context-aware reasoning across steps~35% successInaccurate
Code generation & environment integration~45% successOverstated
Large-scale workflow automation~30% successOverstated
Multi-source decision making~25% successMisleading
Fit for homelab automationRepeated failuresInaccurate
Multi-file, environment-specific automation<50% successOverstated

I went into using ChatGPT expecting it to handle homelab tasks — things like Unraid scripting, DokuWiki automation, cron jobs, VM passthrough configs, and multi‑step environment‑specific workflows. In theory this AI claims to be capable of context‑aware reasoning, multi‑step automation, and real‑environment script generation.


In practice, though, the reality was very different:
• Repeated path errors and misapplied commands even when exact paths and environment details were provided.
• Scripts that were suggested did not work on first pass and required constant correction.
• Repeated violation of clear user rules (e.g., using the wrong tools or paths).
• Failures compound over multiple steps, leading to long, frustrating correction loops.

Based on my experience, for complex, multi‑step, environment‑specific automation tasks, ChatGPT’s reliability was low (around ~30‑40% success rate). It is useful for simple facts, one‑off commands, definitions, and writing tasks, but it’s not dependable for real homelab automation or system workflows without heavy manual correction.

In short:
Cool for quick info and simple tasks — but not ready to replace a human for real configuration and automation work yet.

Cheers

HEADRAT

PS Chat GTP wrote the above ;)
 
Last edited:
I was looking through a patients notes on Wednesday when at the bottom of the Medical Letter was "This letter has been wrote using AI taken from a Clinicians voice notes however it has been checked by a staff member" or words to that effect.
I await the first claim.
 
I used ChatGPT twice this week

1 - Asking it to suggest a gym routine to me for a beginner.
2 - Turn a photo of someone into a movie poster.
 
You must log in or register to reply here.
Back
Top Bottom