To think ChatGPT is the best thing ever?

[ColliLass]

So far it has:

sorted a problem with my emails that I would never EVER have been able to work out
Found a book I haven’t been able to remember for years
diagnosed a problem on the car and accurately estimated the probable cost
found recipes

and many many more little helpful day to day offerings when i can’t remember things or I wonder things.

I loathe the AI videos and pictures that seem to be everywhere but ChatGPT is amazing.

The problem with people who don't use AI and don't understand AI is that they re-hash things they hear on line...and they don't realise how silly they sound to those of us who do understand it.

It can't do 5 x 5!
It gave a solicitor wrong legal info!
It makes things up!

You're at least a year out of date, probably two. These were 2024 problems, problems that simply do not exist now except in the rarerst of tech glitches. If you're consistently getting wrong or made up stuff from Chat GPT then your prompts are the problem.

It's not Google and the output will be as good as the input.

If you'd like a real world example of what Chat GPT can do...dh and I have very recently published our first app. You can actually download it right now on the Apple and Play stores which still feels slightly surreal 😁
It took us nine weeks to build. Lots and lots of hours of work, late nights etc. Pulling apart code, deleting code, reinputting it etc. Thousands and thousands of lines of code, I was dreaming in it by the end.

I'm a risk manager. Dh owns a small transport company. We are NOT generally 'technical' and we have zero coding skills. BUT, I worked for DataAnnotation for a long time so I am skilled in getting the best out of a model and writing useful prompts which give a high value output. For our app...my prompt was about a thousand words.

Chat GPT wrote the code for our app. It took our concept and detailed idea and turned it into reality, giving us a step by step process to build it. It was our developer, tech expert, marketing manager and so much more all in one. We looked into this a couple of years ago when we first had the business idea and it would have cost circa £20k to get it built.

No, AI isn't yet perfect or infallible. It does need directing, reminding, checking. But if you know how to use it, it's mind-blowing.

Calling it 'predictive text' or 'Google' or saying it's incapable or useless is simply laughable. It makes me think of my great granny confidently telling us this Internet thing would never take off 😂

I hate AI art/videos etc and all those daft things people ask ChatGPT to create an image of themselves etc. but for actual tasks, it’s really useful!

I do laugh though when I get a long AI written email at work, and then I can use AI to summarise it.

I agree, OP!

I don't care if it's sycophantic, I like how we have a civilised, supportive and kind conversation about anything I need to talk about. I know it isn't going to be rude or snappy at me.

Yesterday I was in despair about the litter in our area, so between us we've worked up a 6 month plan on how to raise awareness in the local authorities, including a templated letter I can send or email to key departments. It also looked up the Wiltshire Council's campaign that they're running called Stop at Source, so I've got some precedent to bring to the authorities in my area.

it took a total of 5 minutes which is amazing, when it would have taken me hours,

i went from feeling frustrated and ill-equipped in how to start making a difference, to having a plan of attack, which felt so positive. If it wasn't for ChatGPT I would be feeling very different.

It’s useful for suggesting recipes or correcting grammar but gave absolute incorrect advice about a query with HMRC. Not to be relied on.

It's not perfect, but it's pretty good. Obviously you can't believe everything it says, because it makes things up sometimes to fill the gaps. I asked it once to give me quotes from an ethical framework and it made them up. Luckily I recognised it and it apologised when pressed but it took quite a while to admit to it.

If it's a repetitive process and you can build in "checks and balances" it should be OK. Most systems are not just AI/language models running freely. The "sophisticated versions" have a lot of these constraints and extra-special knowledge built-in - but they have to be paid for, they're not in the free versions.

And there was a court case recently where the written judgment included AI-generated misquotes. There is no system with that amount of precision and reliability.

It's amazing. Anybody with any ounce of intelligence or common sense wouldn't use it for proper serious stuff though you can't get wrong. Unfortunately the danger is those using it who can't differentiate between what it's appropriate for and what it isn't.

I've used it recently for a trip itinerary for Iceland and it was excellent. Worked a treat.

I think its great in some scenarios but not all, also often its out of date, gets it wrong or not quite right.

I have used it as a guide re interior design/decorating ideas.

I have asked it to help me plan itineraries for holidays but often what you get is far too much to fit in, in a day so it relies on what you put in and say can you give me a less chaotic less full itinerary for such and such a place that factors in a long lunch or some quiet relaxation time etc.

I have asked it to summarise an email response etc etc but you do still have to proof read it yourself.

I always correct anything AI mis-states, because it then feeds back into the model as the right information. Chat GPT is a lot more accurate than Copilot which massively overclaims what it has created. So now I don't waste my time asking it to generate a PowerPoint template for me because I know a 5 year old could do better. Instead, I ask it to generate ideas for content and then I build it myself eg here is a file containing information about our project, please create a multiple choice quiz from the information. It did that really well, in about 1 minute, saving me time to build the file with the graphics (which it's useless at!)

It is very bad

You think what I'm saying isn't true? I literally said I've got my period and haven't had one for a few months. One of the suggestions was that I should go to hospital. I told it that periods are totally normal so why would I do that? And it 'apologised'.

It is sometimes quite helpful in making my emails more concise and clear but nothing a good editor ( or in fact my DH who can draft things more clearly and just as quickly often as AI can) could do

Oh come on now.

You can throw 50,000 words into a good AI model and generate an excellent, succinct yet detailed summary, in literally five seconds. And then to be certain, you spend another ten seconds running both versions through an independent model to check for accuracy and 'de-AI' (as I call it) the output. Less than two minutes work overall.

No, your dh cannot do this as quickly as AI can. Nowhere near.

It frustrates me how many people don't care about the environment at all. As long as they can get an answer to a crossword puzzle or find the title of an old book in a matter of seconds, who cares about the state of the planet eh? Are we really this selfish and short sighted?

You've admitted this only worked for you because of your unusual experience in data annotation, and even then you and your husband still had to put in endless hours to make it work. These things aren't "amazing" unless and until they can be used by the general public reliably on a day to day basis irrespective of their individual past experience.

It is demonstrable nonsense to say that problems haven't happened since 2024 when people are telling you that they are dealing with the results of those problems on a daily basis now. If someone quotes the wrong legal principles and makes decisions on that basis, it can have disastrous consequences. There are no warnings on ChatGPT telling people that unless they take hours carefully honing their prompts the information it will produce may well be utter bollocks.

you only think this because you haven't moved to Claude yet..

Well yes this is the danger isn't it. Someone putting that in and blindly following what it says.

I have to go back and forth with it sometimes because I'm not an idiot. It clearly needed more input. If you use it correctly and give it enough detail and context, it will help you more accurately.

Nothing beats common sense though. It's a tool like anything else.

It still requires the human being to have some critical thinking and I think people are too quick to believe implicitly in what it churns out like it's an all knowing oracle. It's been fed information and it gets it wrong. There are many stories even in this thread of incorrect information, and when it's been used in business it's caused big cock ups (ie making up figures when jsed as a stock and shares partner), however I'd say the cock up lies always with the human user taking it all as 100% bona-fide correct info.

A little like any tool that can be used in so many ways, I think the damage it does to human ingenuity and creativity outstrips its benefits when used ad hoc by people.

I'm with you there. Robot vacuum cleaner? I want a device that rolls it's sleeves up and scrubs the bathroom, dusts the cobwebs and Hoover's under the furniture.

Meanwhile I worry for my kids employment prospects long term.
I agree it can be a useful tool but it's a bit full of itself. I tried to get it to help me choose a car, given a list of criteria. Apparently I need a Toyota Corolla and nothing else will do.

Like the lawyer who used AI slop in a ciurt case and it turned out to have fabricated the case law?

Yes, terminally stupid.

Put -ai in your search, it omits the slop.

In a google search, just put

-ai

in the search bar alongside your search terms

It can also teach punctuation...

I think there is a half way house. We can use AI intelligently while recognising its limitations and not relying on it for stuff that it is no good at.

Basically, we can use AI to great effect as long as we continue to fully engage our brains when doing so.

I think I am. Are you indicating if we put “I've got my period and haven't had one for a few months” it would suggest we went to hospital?

Well it did for me! Why would I make this up?