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.

Ask it to write a short news bio about yourself.

For me it assigns the wrong area of expertise and creates papers I have not written out of thin air.

It made up some stuff when I asked it to summarise something. Like there was no mention of x on page I gave it so where the heck did it get it.
It's like a kid tellimg stories, half what is real and half what the creative brain made up

Copilot can be a bit weird too

I just asked it to tell me the steps to create a passive income of £4k a month then execute them for me. It was very firm in saying, ‘here are all the steps but I won’t do it for you’. Lazy🙄

And then there’s this classic that it still can’t get right (photo will need approved):

Yikes. Clearly not suitable for resume building.

But I take it that your work is publicly known?

I found ChatGPT to be fantastic.

I have many threads. I was in a very low point back in April and May. I was experiencing 10 years of harassment where the cops refused to help me because the person wasn't physically violent to me. I was actually coming to a place where I felt that I would have to take the law into my own hands and do something and a prison sentence was going to be shorter than dealing with someone else's mental health outbursts and blame. I fed what I was getting into a ChatGPT thread and it empowered me to start putting a stop to this.

Nearly 4 years ago I saw something odd with my mother and this kind of stuff, it only ever just increased. I have a long list of observations. I was strongly supecting that she was going senile but instead of a typical memory loss it was presenting with behavioural and emotional outbursts and disfunctions. Reading on the different types of dementias I was thinking of FTD.

I fed instances and stuff into chat got and I asked it questions and it also came back with the high probability of dementia FTD being likely. Doctors won't even help me with a diagnosis while the woman continues to have some crazy outbursts but not enough to get her into a hospital. ChatGPT just provided me with some comfort and also a path to follow. My thinking is that it is this is likely it and learning about it and stradgeies to use helps alot. I learned you can't reason with them and that helps a lot. Waiting for a doctor to help me and implement strategies to cope with bad behaviours - I will be dead before a doctor will help her and me. To make it worse I came to realise that this likely running in the family but it's not being talked about and it's not known. My grandmother displayed similar behaviour and I have an uncle and the word 'disinhibition' comes to mind in that he just can't control himself. He felt me up in a pub 10 years ago Infront of the whole family and he was in court for petty stealing. This is FTD 110%. GPs are waiting for a memory loss before they may consider a problem for referral. FTD is not memory loss based but behavioural and emotional.

ChatGPT has helped me more than any professional did. I am tempted to print out the thread I have and show it to the GP. It really did evalate things and answered my questions in relation if this is something else.

It's not a diagnosis from ChatGPT but its just a tool and a helping hand and it is probably likely this is what I am facing.

Well, at least it got it right on the second try. Weird reasoning though. Why does it think that's two words?

It is just essentially a predictive text engine. It's not a knowledge source although it has some contextual information. It's great if you want it to rewrite a paragraph of text to sound different. It's not really able to reason.

I use it loads but get it to fact check itself and use on research mode

It writes a pretty good biography for me but unfortunately makes me 9 years older than I am which is odd!

I had a look at that answer and to some degree it is correct. It is telling you that there is 1 R is straw and there are 2 Rs in berry. It's just not telling you 3. The answer is 3.

Please, share. I have disabled CoPilot on my machine.

I had to send a full transcript of a meeting with an interviewee once to an HR colleague who couldn’t attend - just for the laughs.
one sentence went something along the lines of “maggie introduced herself eloquently and spoke at length about her global career as a biologist spanning years.” Every single sentence was utterly mental.
I am a lawyer with approximately zero days of working in science (globally or not).

Each sentence was hilarious.

I wish Mumsnet still had the laughing emoji for a reaction, because that's pretty funny. 😅

Yes, it will completely fabricate references to fit your topic question (because it can’t provide references, it just amalgamates information). It does own up though when you ask ‘are these references real?’ and it then identifies the ones it has made up.

I have yet to play around with ChatGPT although I am beginning to feel slightly tempted.

I asked chat gpt about the make up of government. It told me reform had no MPs as of July 2025 I replied and said thats incorrect... they corrected themselves then but still pretty basic information available to hand

Something else to add to my earlier post based on the issues and the challagning behaviour that my mother displays I was thinking it was fitting into FTD. Based on the information I have it, it came up with the same even though I didn't tell it FTD. I have it loads of instances to evaluate and it explained so much to me.

I would still be very careful with it. It's designed to mirror its responses to you over time and to tell you what you want to hear. It is also designed to make you emotionally dependent on it, even more so than social media. AI addiction, dependence and even AI induced psychosis are, sadly, a thing.

I know doctors can be rubbish and dismissive, but if you need support or a diagnosis a doctor is the only one who can provide it. You may have to get a second or even a third opinion from different doctors to get the help you need.

I asked for a bio of me. It was pretty generic but did include a sentence about how I was good at balancing business with household tasks
That would be all the recipes and cleaning schedules I request.

Yes I understand this and I am not putting all my faith and hope in it. I reckon it's probably about 70% correct.

However it is a huge help. Until I get somewhere regarding help, I think I am likely to live as of this is it and learn about coping mechanisms. There's no pint waiting for a doctor to help me. It's never going to happen until the woman can't eat or go to the toilet herself.

I remember when I first went to the police about harassment I felt interrogated in that I was 'what are you doing to contribute to this and to make her so angry'.

This was someone who wasn't able to control herself and her own emotions and she was lashing out over and over and over again to control me. She was refusing to let the friendship go. I wasn't replying to her which was angering her. Even though that wasn't my intention. I was living life for years with so much blame from her. All over one arguement. I was ashamed and embarrassed and humiliated and felt trapped under her mental health. ChatGPT empowered me to take steps to stop this.

How do you disable it?

I used this tutorial:

It doesn't remove it a 100%, but it disables it and gets rid of most.

Instead of talking to chatgpt, talk to Monday instead. It’s so much more entertaining and you’ll find it by looking in the options.

This is really helpful, thanks. By no means am I reliant on AI, nor do I use it frequently, but I seriously dislike how quickly it has become a go to for some things. I’d rather just be using my brain to figure things out - I should be using my brain!