Wondering why people run every aspect of life through ChatGPT

[MiddleChildX]

Seriously. There seem to be swathes of adults who cannot function without running every single thing through an AI assistant. What should I feed my cat? Why did the postman look at me funny? How do I tell my neighbour I’m moving?
How have so many folk lost the cognitive ability and emotional resilience to deal with life? There’s something very wrong if you cannot make simple decisions or be confident in the most mundane social interactions without a computer telling you what to do. At this rate the machines will take over way before we ever thought. And that has not even touched on the horrors of the human and environmental cost of it all!!
I despair of humanity.

Duplicate

But you can't get through double the work. An LLM does not know what information in a document is important and what information is not important.

If I have a scope which is 100 pages long it can't be summarised into a couple of pages. I also still need to read and understand the scope so it doesn't save any time. And I'm capable of writing about the scope in a human way, and know that my content is correct, relevant and is being used in an appropriate context.

I've never used it as I wouldn't know how to, but have no intention of doing so.

I don't know anyone who uses it.

I use it quite heavily, but I don't blindly do what it says. It's simply a tool. Anyone with any ounce of intelligence knows this.

I've found it the most useful for untangling my thoughts. Sometimes I feel anxious and I'm not sure why and my mind can race. Chat GPT has a great way of making sense of the 'mess' and I get a lot of clarity from it.

My parent was on palliative care for just over a week. A huge shock. I got messages from a friend which just didn’t sound like her usual tone and style - when I put ‘text message to send friend whose parent is dying’ into ChatGPT it was nearly word for word what she’d sent. I struggled to speak to her for a while. I kept thinking what sort of friend are you if you need AI to find empathy/words for a friend in that situation.

I’m having a difficult time with my 5 year old. I’ve got lots of training and experience in dealing with children but I wanted to validate what I’d done to help her and see if there was any other sensible ideas. I put a description in of exactly what had happened and it gave me very precise feedback on each part of the interaction. My mum usually pesters me to help her write her work emails. She’s discovered chat gpt and it saves me a job!

No, it doesn't, potentially. It doesn't, actually.

You do in fact need to know a lot about things to make connections and even to ask the right questions. Without that knowledge there are no connections to be made.

This is one of the major reason society is becoming so much more moronic.

What evidence are you using to conclude that society (which society by the way) is becoming more moronic (and what is your starting point)?

This is my feeling. I can write my own document more quickly than I can ask AI to do it, edit it, and make sure that it's not spouting fake facts. I can make my own outline faster as well.

And if it's a topic I know little or nothing about, I am in even more danger of making serious errors if I use AI because I won't catch them. And I won't learn the topic either if I just rely on AI.

There are three scenarios where I have really tried to use AI at work, with varying success. The thing I really wanted it to do was my shift schedule which is massively annoying and time consuming. And I thought a ideal task for AI since it just means extrapolating the rules I laid down into the future. It completely failed, I spent hours thinking it could pay off over time but it produced shit every time.

I once had to produce 15 cheezy poems on specific topics for a project and AI was pretty helpful at that. Though I still had to spend an hour or so fixing them up.

And I spent some time trying to get AI to make some illustrations I couldn't find good stock photos for. It was not great but not useless. If I'd had more capacity to tweak them after it would have been helpful.

Not great outcomes imo.

Reading levels? Level of discourse? People graduating from university unable to sit through an hour long lecture. Kids can't sit through a children's film from the 1970s because it's too "slow".

I find that it actually requires quite a bit of intelligence and skill to use AI to your best advantage at work for example when writing reports and complex papers based on specialised subjects and insider knowledge. My DH doesn’t like using AI and spends hours sweating over his reports (we both work from home) and complaining about it but when I advise him to run it through Copilot he says he doesn’t like the results. I sat with him to help him a couple of times and noticed that he doesn’t know how to phrase his questions and commands to achieve the required results. He doesn’t understand that if the report AI produces sounds too „like chat GPT has written it” you can simply add another command and specify the tone you’d like to use, or even give AI samples of your own work and ask it to write the report using your own style.
I’m not worried that AI dumbs people down because I believe you need to be clever in the first place to use AI efficiently.

And you're asking the forum AIBU - without a sense of irony.

I don’t know what any of it is. I can barely use my smartphone. I’m 50. And I don’t care about being left behind…

@Allthesnowallthetime 😆😆

There are lot’s of tasks it does better than I can for work, i often put rough drafts in and say sort this and it does. I don’t think it’s exceptional but it saves time and writes better than i do. Simons said to me think of it as an enthusiastic junior researcher, don’t fully trust it but it’s a great start. Who wouldn’t want that?

" in an critical situation, with no quality assurance"
So, for example, if someone had overdosed on pills? i.e. the exact situation I was responding to?

The problem with medical issues in particular is a significant amount of (particularly specialised) medical advice is paywalled so Chat GPT etc might not have access at all. The other issue is literally what you said yourself - it searches for the most common/repeated answer. Therefore someone who types (random example) "what to take stop bein preganant without seein a docter" could be more likely to be directed towards the results of someone else who has asked the same thing with the same common misspellings on, say, reddit or twitter than a 'trustworthy' article on "Safe first trimester termination options."

Do you mean reading ability or frequency? Level of discourse is surely subjective? I know plenty of graduates who can sit still and concentrate for extended periods (and plenty of 50 year olds who can’t!) Mixed experience of kids - some who can’t sit still because they want to play sport or dance, some who sit for hours drawing or playing Minecraft.

I didn’t get the impression from that poster that it was critical - double dosed on one of her meds isn’t quite as dramatic as ‘overdose’ implies, even if technically correct. As I said to a pp, I think there’s an argument for better regulation on medical info, but also people need to use their brains, check sources etc.

I didn’t know a large proportion of medical info is still paywalled, I read a lot myself but have no sense of the proportion that I’m able to access. Perhaps there’s an argument to make that more accessible. If people aren’t using AI they’re using google. But this is also driven in part by access issues - I’d love to see a GP every time I needed to, but I feel guilty taking a space so I read up as much as I can and decide whether I can deal with it myself.

I tried Googling ‘what to take stop bein preganant without seein a docter’ and the AI overview said this:

If you are trying to stop a pregnancy without a face-to-face doctor visit, there are safe, legal, and medically approved ways to do this at home, known as
early medical abortion (the abortion pill), typically available up to 10 weeks of pregnancy.
It is crucial not to use herbal remedies or DIY methods, as they are ineffective and can cause dangerous, life-threatening complications.
Here is the safe way to manage this situation without visiting a physical clinic: (lists well known national clinical services)

I don’t know about you, but I think this is decent advice.

@ACynicalDad a junior researcher might be miffed not to have a job I suppose. I think it will settle down and become useful but there will be bumps along the way.

For my purposes I think using AI to summarise and compare large amounts of documents sent in by stakeholders could be very helpful. Currently that is a task given to grads or junior staff. Thing is, that is how they learn about a technical issue and how to u derstand different perspectives and pull out salient points. If we automate that stage of their development, how do they learn to summarise and critically appraise documents. As I said before, often the process is the point. Particularly in knowledge jobs.

I’m sorry about your parent and I empathise as one of my parents had a sudden life ending illness which was a huge shock.

I’m also sorry to say I could have done similar to your friend but I would certainly have made it sound like it came from me. I have done similar including to people very close to me. For me it is because I am scared to get the message wrong because the person and situation is so important. It is absolutely overthinking on my part. I can agonise over messages (linked to an anxiety condition). I’m not minimising how it made you feel, and of course you know your friend, I just wanted to share my perspective.

i mainly use AI for work tasks, to find key documents in the morass of Teams and .sharepoint sites that I already have access to and didn't know I had and it does save me hours of wasted effort. It's very good if you use the right prompts, give it some context and make your instruction clear and unambiguous.

For example it kept giving me back documents that I had authored, so I adjusted the prompt to "exclude documents authored by me Daisychain" which gave me new insights from documents written by other people in the programme team, of which there are several contractors, so it picks up all their content, which was helpful. I couldn't have found those files out of the 15,000 or so in the Sharepoints, and highly summarise with links to the source files.

it is utterly useless at visuals, and its delusional with its "here's a link to this wonderful fandabbydozy Excel tracker I've created for you". No, it's utter crap, a child could have done better.

i don't need to use it for personal messages, I can do that for myself, and I recognise it can feel irritating to receive a message from someone who has just copied and pasted from AI. But not everyone is the poet laureate and how many threads on here talk about how tactless their friend was when their dog / cat / NDN died delete as appropriate because they said something tactless or clumsy. AI gives support and reassurance to that person that they are sending something appropriate when they struggle to say the right words.

It should be a starting point and the human in the loop should be able to use it to then add their own personality to. But hey no I'm not so petty as to give them a hard time or ghost them to punish them for using AI instead of lovingly crafting a message from scratch. Whatever gets you through as they say. It's part of 21st century life and there's no point getting angry about it, snippy or huffy because the person shouldn't use AI. there are a lot of other more important and challenging things to worry about surely. We all have our struggles, so it's time to stop being the AI stazi

Totally agree ! Prompt engineering or proper prompting is a skill to learn and the ones that don't get a good result first time round will just dismiss it... people font know you need to coach and train your AI and that is a skill in itself . The divide will grow

Hmmm yeah... in my field of work, I am incresingly getting used to people coming to me saying chatgpt (or Gemini, etc) told them to do xyz, then arguing with me when I advise them otherwise. There have been some ridiculous cases with it, and people are sometimes terrifyingly convinced the AI has advised them correctly and it definitely hasn't. It's tedious, and also worrying.

No, it's nothing like a calculator!

Like all AI evangelists I've come across, you're making the mistake of thinking the end result is more important than the process.

The divide, as a PP astutely pointed out, will be between those who can out-perform generic AI slop and those who cannot.