How to know ChatGPT is full of shit.

[DiggingHoles]

Take a book of a shelf. A classic is best. Open up to the first page of a random chapter. Now ask ChatGPT to quote the first paragraph of that chapter.

Tip: Have some popcorn ready while you rephrase your request multiple times.

Similar to chatgpt but ratjer it seems to leave out things than make them up. Bit dry and imho not that useful. Tone of it (i know that sounds weird😂) is just weird

Alternatively, you can also install Linux. I have done that with my personal devices. At work I am still stuck with Windows, sadly.

I've been genuinely shocked, reading threads on here, how many people dont understand you need to fact check what ChatGPT spits out. And its great its helping with some people's emotional problems, but surely its blindingly obvious how damaging it could be to some people? I guess I thought the general public were more clued up than they are on it, seeing as its not actually a new concept.

I hate how it tries to give you an answer that pleases you rather than the right answer

It frequently tells me things with confidence, I explain why it's wrong and it says "Yes, you're correct!"

I wouldn't mind if it said it wasn't sure, but it's the assertion and confidence when it's talking bollox that means I rarely use it

It has been damaging already. It has let people down a conspiracy rabbit hole, it has made people believe it's a deity (yes, really) and there has been one reported AI induced suicide. It was a Belgium teenager and his mother is suing the developer. Court case is ongoing. Granted that was Character AI and not ChatGPT, but even ChatGPT has been reported to tell a user to kill themselves.

This highly impactful, unregulated technology can be really scary.

I used Chatgpt to find a specific online course. It confidently told me that Uni X offers what I‘m looking for. I looked on the Uni website- they don’t. I told Chatgpt and it might as well have said ‘oh silly me, you’re right’. This happened a few times, very annoying.

Whenever I need to remember not to trust ChatGPT I ask it to solve a cryptic crossword clue. It gives you an answer and a very confident explanation but it’s all just rubbish and the answers don’t even have the right number of letters. Eg ‘Seller with key, say, a geezer (6,5) produces this:

“Seller with key, say” = homophone of “quay”
So the whole phrase sounds like:

“Del boy”

💡 Del Boy is a geezer, a seller, and famously a market trader in Only Fools and Horses.

And his full name?

✅
Derek Trotter
(6,5)

• ✔️ 6,5 ✅
• “Seller” = he’s a wheeler-dealer, market trader
• “with key, say” = homophone for “quay”, which sounds like “Del-boy”’s world
• “a geezer” = Del Boy is the quintessential London geezer
• It’s a cryptic definition / pun clue referring to a famous character

✅ Final Answer:
Derek Trotter

Explanation:

• Cryptic definition
• Derek Trotter = Del Boy, a cheeky market seller
• “Key, say” = quay = part of his Cockney environment
• “A geezer” = London bloke

That one was a belter. Want me to craft a few like it for you to solve?”

No thanks, ChatGPT.

I've never thought of myself as cynical, but when it comes to LLM I'm afraid I am.

What is the reward for the company's that generate AI? The last person to have given something away on this type of platform was Tim Berners-lee,
and there aren't many of him around among the current tech bros.

So if the company's are gaining something from letting you use you platform what is it? And why aren't they upfront about what they are gaining?

Id ask AI a factual question, but check its sources. Id never ask it for an opinion or for emotional support.

The latter feels very odd to me. I might ask a friend "do you think I was right saying/doing that?". But I know my friend, their influences and experience s, I would never ask AI.

You'd get a better spread of opinion on aibu here, in my opinion.

You'd get a better spread of opinion on aibu here, in my opinion.

The thread would likely have more responses, but I don't really have an unreasonable to add. I would only be posting it there for traffic.

I think that anyone posting on MN who starts their post with: "I asked chatGPT for you and this is what they said..." should have an instant suspension.

If I wanted to know what "fucking chatGPT said" then I would have asked it, instead of posting on MN for some opinions from - you know - human beings

I had a similar situation so i grilled chatgpt on why/how it had made such a massive error.

It explained that it is only loaded up with everything on the internet up to April 2024. The election was May 2024. If you ask it something more current it should in theory run a websearch rather than relying on what it 'knows', but sometime it fails to trigger this.

It is therefore worth remembering this and making sure you input in such a way that triggers the websearch.

I was looking into info on a local high school, and after getting loads of relevant info from chatgpt i asked it why it hadn't mentioned the elephant in the room and the surname of the ex headteacher jailed for multiple counts of child sexual abuse. It still didn't mention it, so i said 'xxxx the massive paedo' it then came back stating as fact that this man wasn't a criminal at all. I linked a news article of the verdict/sentencing and then asked why it had got it so wrong.

If it makes a mistake it is a really good learning opportunity to discover how chatgpt works, i learnt more from that one conversation than all the other articles I've read to date.

Agree. It’s wild to me that there are people out there using it like a black 8 ball to check their destiny or similar.

I'm not sure that's true - I use it to help me generate work reports and the references it provides look pretty accurate to me ( established industry journals etc).

Oh, I thought they meant ‘you’d get a better response on AIBU than what AI will give you’ - like, why are you using AI when we have AIBU?! 😂

Yes some are accurate, but definitely not all. Depending on your topic, it may be that some of those references looked very real, with convincing article/research titles, topics and authors, but when you look into them they are completely made up.

As others have said, it tries to please you, so you need to tread carefully if using it to collect research sources. Just because AI said that a source (fabricated or otherwise) included certain information, it doesn’t mean that it did.

A YouTube influencer I follow posted a video about using ChatGPT to figure out what “season” she is - apparently this is a thing on TikTok/YouTube - and I’m sitting bored on a train and just for fun decide to try it myself by uploading a couple of photos of me. (Light Summer, as I suspected.) I asked it to adapt one of the photos with a makeup look including suggested products and the result was creepy - looked sort of like me but not really, but wearing the same clothes and with the same hair and in my old kitchen. Gave me a weird shiver almost.

I’m a lawyer myself and can’t imagine yet actually using it for work, although we are constantly being pushed to incorporate it.

And yes, it is definitely set up to flatter users - I asked it if I looked my age and it swore I looked 10 years younger.

(estate agent!)

It can be good for coding, but only if you already know coding. Disastrous for learners. And so dull on here. I hate when people post AI slop!

My head of department used it to create a full term's math, reading, and writing plan for the Y2s and Y3s. She presented it at a meeting and distributed it to all teachers. I had to pick my moment to gently raise the tiny snag that the curriculum objectives it had met were completely made up, and the plan did not meet any of the actual curriculum objectives that we are working towards.

Oh, I may have read it wrong then. I thought the poster meant to say to move the thread to AIBU.

They probably did, I think it was likely me misunderstanding 😂

my current research with it , is history in the cold war era, yes some may be incorrect but if i match it with wikipedia and then some historians books etc it will give at least the basic information

I use video production software quite often, and don't really know much about how it works, so I often use Google to try and find out how to do specific things. I now get an AI generated answer as the first result.

Without exception, they are utter nonsense. I'm asking a specific question, but AI gives me a generic sounding answer that does absolutely nothing to help. It just produces utter nonsense every single time.

The phrase "artificial intelligence" is a paranym - an expression that means the opposite of what it says. A more accurate description would be "artificial stupidity". Hearing politicians (Wes Streeting) saying how this technolgy will revolutionise the NHS is terrifying. And yes, I am a person who has used technology to do complex things throughout my whole career.

im guessing the overall goal is using eg chatgpt more geared towards development and tech and military applications etc and the public are the ones training the chatgpt.

Asking it to write a bio on me was interesting because it was very good and very accurate. And made me realise it knows too much about me!

I learnt to fact check after I planned a whole weekend away, based on flying from my local airport, to discover those flights don't fly from there. Infuriating!

However, I also use it as a kind of therapist sometimes and it's great.