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

[HolyRigatone]

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?

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.

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.

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?

Not too bright and gullible

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

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.

That was Boris who called it that

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.

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.

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?

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.

@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

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.

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...!'

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

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.

but then asking on mumsnet is kinda like : ' asking the random person sitting next to me on the bus.

And even when people do that, it often still kisses your arse. It’s a topic I’ve been interested in for years and I’ve followed the sub on Reddit and I’ve seen so many people blindly trusting that because they have used that prompt then it’s not kissing their arse sometimes.

dh and I experiment with it on our iPads, we make up a bullshit argument, and even using the prompt you describe, it still kisses arse. If we were both using it to resolve genuine conflict the logic it uses would have us divorced.

There was a car wash prompt doing the rounds a couple of weeks ago where people were saying their car is dirty, the car wash is 50 metered away and asked chat got if they show wall or drive, it’s told everyone he seen who did that prompt to walk. 🤦🏻‍♀️. I tried it myself and it told me to walk, even when I pointed out I need the car with me to wash it, it still reasoned walking is more efficient and I can walk to car wash, return home for car and go back 😂😂. I’m not going to trust it to point out flaws in my logic when it can’t even spot its own.

The thing I find both strange and oddly upsetting is when one person asks a question/ starts a thread/ etc. and someone responds with "I asked ChatGPT and it said..."

I don't know why, I think it bothers me that someone is asking for actual human advice and people take it upon themselves to ask a computer instead. It feels impersonal, presumptive and almost rude. If people wanted to ask AI, they'd do so; there's no need to bring it into spaces intended for real connection. And that's not even going into the environmental impact of every single input into AI chatbots, which is something I would never want done on my behalf, without my consent.

Absolutely this!

I find this annoying too. People post here to get answers and opinions from actual people. Not asking someone to relay what ChatGPT has told them. The way they post replies, they are so proud and think they’re super clever for doing this too.

It dumbs you down. So when you were not stupid once you will end up stupid and relying on a load of pixels to keep you right. Then you'll believe anything it tells you because you've lost your critical thinking skills.

Years ago, back in the texting era where we shortened words because texts cost money. I eventually lost the ability to spell properly and it leaked out into other areas. I had to force myself to reverse it.

You don't want to lose abilities. It will happen.9

Did you start a new chat? If you don't, it will refer to your earlier convo as you haven't told it there's a change.

I find it's happy to give a definite answer to questions that google, research, you tube and real people won't give. So it brings an initial sense of relief and calm - finally! An answer, something that understands! But the answer is just bullshit asserted convincingly. If you query it, it immediately changes its certain stance to a completely different answer! It's absolute rubbish. But we all love certainty over uncertainty, it makes us feel good and takes away the awful feeling of not knowing or having to make difficult choices.