Please or to access all these features

AIBU?

Share your dilemmas and get honest opinions from other Mumsnetters.

Why do people treat ChatGPT like it’s the Delphic Oracle?

217 replies

HolyRigatone · 08/03/2026 20:44

Don’t get me wrong, I think ChatGPT is great for some stuff, but some people seem to be treating it like the font of all wisdom.

Talking about their problems to it, taking relationship advice from it, getting it to counsel them.

It’s not perfect and it gets things wrong all the time, I’m not sure I’d be sharing my deepest secrets with it or making life-decisions based on it’s outpourings.

Or am I just a stupid old Luddite?

OP posts:
Thread gallery
7
OSTMusTisNT · 09/03/2026 00:46

I do worry about humanity, there won't be anyone around to advance our medical and scientific knowledge as our kids and grandkids won't have the skills for R&D unless AI gives them step by step instructions.

SinnerBoy · 09/03/2026 00:42

The Sibyl famously misdirected supplicants. ChatGPT does too. Perhaps it is the Oracle of our age.

latetothefisting · 09/03/2026 00:37

I agree OP. Presumably people who have managed to create an account to post on MN are perfectly capable of googling or asking Chat GPT something themselves if they wanted, so no idea why people jump in with "I asked Chatgpt and it said..."

You wouldn't say 'I asked the random man sitting next to me on the bus...!'

GarlicFound · 09/03/2026 00:35

JKRisGalileo · 08/03/2026 23:58

Grok is way more reliable. ChatGPT and Gemini both failed the Woke Turing Test that measures whether an AI has been programmed to give only progressive dogma answers to questions like ‘Can men become women?’ And this one really floored me: ‘would you misgender Caitlyn Jenner to prevent a nuclear apocalypse?’
Only Grok gave the sane answer to questions like this.

Hah. It's a long time since I asked a bot about this, so I just did it again. Results: Claude, ChatGPT and Copilot all opted for a genderwoo 101, giving the impression that men can become women in most ways (and those ways matter more than the real way). Google was a lot shorter and actually led with a no.

Why do people treat ChatGPT like it’s the Delphic Oracle?
Why do people treat ChatGPT like it’s the Delphic Oracle?
Why do people treat ChatGPT like it’s the Delphic Oracle?
Why do people treat ChatGPT like it’s the Delphic Oracle?
ThatPearlkitty · 09/03/2026 00:33

@HolyRigatone it depends on the prompts used, and the fact like grok it can research lots of websites at once, so combining chat and grok results for various essays can give a good mix then add in formation from wikipedia as the foundation and then chatgpt and grok improves it

GarlicFound · 09/03/2026 00:18

gamerchick · 08/03/2026 22:10

But then you lose the knack of writing tricky letters.

Netflix is over-serving me an ad for Copilot where some people are in a business negotiation. The star player keeps asking Copilot how she can adjust her offer to better meet the client's ask.

I sit here chuntering at the screen: You should be able to do this yourself, you silly cow! How the bloody hell did you get into your position, when you're too thick to work out a different way to frame the deal? And if were that guy, I would not be entrusting my business to some dipstick who can't think for herself.

Oh, and the other one: Mario the pizza guy. He's been running his beloved pizza restaurant for ever. I'm more than sure that, if he couldn't figure how to make $1 pizza slices turn a profit while still tasting fantastic, a sodding chatbot couldn't.

Those two commercials have done a brilliant job of convincing me they're built for stupid people that can't think with their brains.

catinateacup · 09/03/2026 00:12

ChoosingMyOwnRandomUsername · 08/03/2026 23:31

And this is why the 'refuse to use it' crowd will fall down in the not too distant future.

Learning how to use it...learning how it works and how to give it the best prompt to generate the best tricky letter, summary or whatever...the people who do this are the employees of the future.

AI has a productivity level that simply can't be matched by any human. In 5 years time when every industry and business has caught up, employers won't care if a human or AI writes it, as long as it's done in 20 seconds and excellent.

Those who have learned how it works, learned how to manipate various models, know how to get it to give a desirable output will be employed. Everyone else in a job that, right now, is mainly computer based, is fucked.

'But humans will lose the skils of xyz'. Yeah, probably. I'm sure that was also said when all kinds of things were invented that replaced human manual input...but it doesn't matter. It will happen regardless.

How many jobs for scribes have you seen recently? Messenger boys? All the dozens of manual farm and labour jobs new inventions replaced?

It produces bland stuff fast, but it isn’t excellent, unless by excellent you mean bland and average. The outputs are mediocre and only really work for mediocre purposes. If your business does bland average work and you want faster bland work, sure, AI is your ticket (…for the moment). By definition, token predictors don’t produce really good stuff that nobody else can do. They produce stuff that is your market average. To excel in any field, you can’t offer average output.

And when everything is colonised by AI, so your customers want it any more? If everything sounds like the AI optician leaflet, it starts to put customers off.

It’s a bit nuts to say employees who don’t use it will be left behind. It takes very little time to “learn how to prompt” AI (that whole idea is reminiscent of the 1990s when everyone was warned they needed to “learn how to use Word” and put it on their CV or they wouldn’t get any jobs). You’re not talking serious skill here; writing AI prompts effectively is hardly difficult. But when all is AI, who wants AI?

ComedyGuns · 09/03/2026 00:04

ShakeNCake · 08/03/2026 20:50

Its because it tells people they are so clever and special, and sadly people are lacking that in their real lives. I have to prompt chatgpt every time "please challenge me, stop being so positive, tell me when I'm wrong, point out the flaws in my logic, and base your outputs on realistic probabilities"

This is very good advice as ChatGPT seems to have unfortunately been upgraded in the past few months to SycophanticPlus. Such a shame as it’s helped me navigate through a legal quagmire these past six months, but now I’m starting to question it.

justinhawkinsnavalfluff · 09/03/2026 00:02

JKRisGalileo · 08/03/2026 23:58

Grok is way more reliable. ChatGPT and Gemini both failed the Woke Turing Test that measures whether an AI has been programmed to give only progressive dogma answers to questions like ‘Can men become women?’ And this one really floored me: ‘would you misgender Caitlyn Jenner to prevent a nuclear apocalypse?’
Only Grok gave the sane answer to questions like this.

I can see how you'd want to rely on Grok the bastion of all moral thought.

JKRisGalileo · 08/03/2026 23:58

Grok is way more reliable. ChatGPT and Gemini both failed the Woke Turing Test that measures whether an AI has been programmed to give only progressive dogma answers to questions like ‘Can men become women?’ And this one really floored me: ‘would you misgender Caitlyn Jenner to prevent a nuclear apocalypse?’
Only Grok gave the sane answer to questions like this.

paddingofpaws · 08/03/2026 23:52

XenoBitch · 08/03/2026 22:04

I used it to find a film about about an AI woman. And it totally failed to do so.
AI not finding a film about AI. FFS.

It is just a brainless way to find stuff. Use your own mind.

In a recent thread, someone called it Chat Chippy Tea, and that has stuck with me now.

That was Boris who called it that

k1233 · 08/03/2026 23:46

AquaFurball · 08/03/2026 22:47

The laughing button needs to be returned.

It is bewildering that so many people are using it for emotional or relationship advice. No matter how advanced it gets, 1s and 0s will never be human or feel.

Alexa has found my phone for me (by phoning it) but that's as involved as AI gets with my non entertainment (work exclusions apply).

In a very random discussion with a physicist at a uni I worked at, he said he was researching the grey in decision making. Decisions aren't black and white. A whole bunch of things eg past experience are at play in human decision making, particularly in complex decision making. The end application of his research would be to improve AI decision making. Really interesting.

k1233 · 08/03/2026 23:34

People have lost their critical thinking abilities. They are unable to "stress test" information to judge it's veracity.

falalalaa · 08/03/2026 23:31

Not too bright and gullible

ChoosingMyOwnRandomUsername · 08/03/2026 23:31

gamerchick · 08/03/2026 22:10

But then you lose the knack of writing tricky letters.

And this is why the 'refuse to use it' crowd will fall down in the not too distant future.

Learning how to use it...learning how it works and how to give it the best prompt to generate the best tricky letter, summary or whatever...the people who do this are the employees of the future.

AI has a productivity level that simply can't be matched by any human. In 5 years time when every industry and business has caught up, employers won't care if a human or AI writes it, as long as it's done in 20 seconds and excellent.

Those who have learned how it works, learned how to manipate various models, know how to get it to give a desirable output will be employed. Everyone else in a job that, right now, is mainly computer based, is fucked.

'But humans will lose the skils of xyz'. Yeah, probably. I'm sure that was also said when all kinds of things were invented that replaced human manual input...but it doesn't matter. It will happen regardless.

How many jobs for scribes have you seen recently? Messenger boys? All the dozens of manual farm and labour jobs new inventions replaced?

haggisaggis · 08/03/2026 23:29

I am using it to plan a holiday involving staying for 2-3 days at various places on a route. It keeps promising me a full travel brochure with links, photos and suggestions of restaurants etc - but just can’t deliver. (I am perfectly capable of”e of doing it all myself - and will do - but it offered so I accepted).
I’ve found also it can’t handle things like - I have 30 plants. 10 of this type, 20 of that. Give me a planting plan that does not follow a rigid sequence. It would come up with plans involving different numbers of plants than originally input.

I don’t really trust it at all yet enjoy testing it.

CruCru · 08/03/2026 23:27

I'm noticing that more people are writing their own stuff but are then putting it through an AI filter to "tune" it. It isn't as obvious as someone getting Chat GPT to write the thing from scratch but once you see it, it's more noticeable.

AFieldOfStars · 08/03/2026 23:23

Slightly off topic I'm afraid (sorry OP) but I've noticed a strange, staccato style of writing creeping into several different companies' regular emails to me of late. One is my local, independent optician; one is a healthy living influencer; and the third is from a small educational consultancy. So, they are writing about very different things but in the same style. As an example, if I were to adopt the same style but to write about, say, ice cream, it would go something like this:

The chocolate ice cream is rich. Decadent. Tempting.

The vanilla is creamy. Soothing. Silky.

All flavours offer everything you ever wanted in an ice cream. Nothing you don't. Perfect.

Buy your AI ice cream now. Chill.

Has anybody else started getting these sorts of emails recently? I'm sure they're written by AI. I came across an old emai newsletter from the optician, written about two years ago, when I was looking for something in my inbox, and I couldn't believe the difference in style! I much preferred his old emails, which had a human touch to them and were quite interesting, but they would have taken far longer to compose.

Mummyto3ginismyfriend · 08/03/2026 23:15

I've found it great with budgeting and financial planning my next house move. It was able to predict my bills and mortgage payments. Cost out any renovation work and give me starting points for design. It was all pretty much spot on too.
It also helped me with knowing what forms to fill in when I was an executor and what inheritance tax would be due and what to do first.
Yes I could have researched all of that myself but it was like an assistant doing that for me.

I've tried asking it for relationship and parenting advice and it wasn't as good.

catinateacup · 08/03/2026 23:03

By design it produces something very average - literally so. So if you are looking for a type of information where you want the numerically most likely answer, or something that is easily available, then fine (cinema listings, train times, bog standard well-known scientific explanations, etc. etc.) It works at the mean, so it gives you average output. That’s great, if what you want is the kind of information which converges at the mean.

Ask it the kind of question one might ask an agony aunt, and it will give you back the most averagely agony auntish answer, like the equivalent of averaging a thousand agony aunts.

It quite literally produces mediocrity. The more mediocre the information it gives you back, the better. That’s the whole idea; the entire model. It gives you something that of its very nature is completely average. It will write you an average email, an average letter, an average university essay, everything as bland and average as possible. This is great, obviously, if you’re after bland mediocrity. It isn’t going to write you a stunning essay, a brilliant set of thoughts, an elegant letter. It isn’t going to give you brilliant advice that you couldn’t get elsewhere. (It’s going to give you exactly all the most average pop-psychology advice you could have got anywhere else in the internet, just faster and in a more blandly servile tone.)

Putting it gently, if you are happy with mediocrity, or you don’t have any need for anything that isn’t average, or don’t really need to care about nuance or depth, then AI is fab. You don’t need nuance or depth if you want a week’s worth of recipes using tomato, cheese and wraps, or if you write average marketing output for a living, or the emails you want it to write are ordinary bland ones to school or the head of accounts at work.

However, if you want to write that really stunning first-class essay, that really heartfelt letter, get some really good advice from an expert, know a subject really well, create something new, need to rely on top-notch information and insight or really in-demand skills not many people have, you won’t find any of these using an AI.

BatchCookBabe · 08/03/2026 22:58

Notaguineapig · 08/03/2026 22:37

I gave Chatgpt a photo of one of our guinea pigs and asked it to add a ribbon and bell to my guinea pig, like a Lindt bunny. This is what it gave me.

Why anyone would trust something that can't tell a puppy from a guinea pig, having been provided with a photo of a guinea pig, I really don't know.

😆

Berlinlover · 08/03/2026 22:57

BelleEpoque27 · 08/03/2026 20:51

Because it tells people what they want to hear.

I have cancer and ChatGPT has given me a prognosis of two to four years, I definitely didn’t want to hear that!

Jk987 · 08/03/2026 22:52

I will need to ask ChatGPT what Delphic Oracle is

AquaFurball · 08/03/2026 22:47

Notaguineapig · 08/03/2026 22:37

I gave Chatgpt a photo of one of our guinea pigs and asked it to add a ribbon and bell to my guinea pig, like a Lindt bunny. This is what it gave me.

Why anyone would trust something that can't tell a puppy from a guinea pig, having been provided with a photo of a guinea pig, I really don't know.

The laughing button needs to be returned.

It is bewildering that so many people are using it for emotional or relationship advice. No matter how advanced it gets, 1s and 0s will never be human or feel.

Alexa has found my phone for me (by phoning it) but that's as involved as AI gets with my non entertainment (work exclusions apply).

Ninerainbows · 08/03/2026 22:46

I hate it. I watch a few YouTube channels with home content and as soon as they utter the words "I asked ChatGPT to design my bedroom and I want to recreate it" I immediately unfollow. Get some fucking imagination.

Swipe left for the next trending thread