Chat GPT - worth getting to grips with it

That will be cookies, I think, and will be confined to your computer (or to your account at the shop or whatever).

The scenario Elless describes is different in that she suggests that AI might show me your address if I asked it, and I just don’t see how that might happen, which is why I’m asking for an example of what she is saying.

As I say, I’m not being picky - I just want to know for myself, and would hate others to be put off using it because of what might be misinformation.

Doodledog

Elless

As previously said if you are filling in a form and input any information ie names, addresses AI will learn that name and address and use it when someone else asks a question.

But how?

If I told it my name and address (unlikely) in what circumstances would it tell someone else? Can you give an example? Has this happened to you?

I think my computer already does this. When I sign up for a delivery online, or use a new website, the name and address fields get automatically filled in without me entering the words. I presume this is AI otherwise just my computer being helpful?

So it doesn't 'tell someone else' it enters the information for you but you can change it if you wish. Then you submit the order or whatever it is when you're ready.

I loaded my photo of a hollyhock flower and asked it to add a bumble bee. Then I asked for Impressionist, Old Master and Klimmt versions. It took a while to process the requests but I was gobsmacked by the results.

There's going to be no stopping me. I've definitely succumbed.

Elless

As previously said if you are filling in a form and input any information ie names, addresses AI will learn that name and address and use it when someone else asks a question.

But how?

If I told it my name and address (unlikely) in what circumstances would it tell someone else? Can you give an example? Has this happened to you?

It was great for designing our new garden bed! We uploaded photos of the existing bed, and asked it to give us a design with perennials and shrubs, and it did!

As previously said if you are filling in a form and input any information ie names, addresses AI will learn that name and address and use it when someone else asks a question.

'stupid' should read 'striped'

This is how wars start 😂

I've just asked it which shade of kid silk would go best with a striped yarn I want to use for a jumper. It's hard to tell from the ball, as the stupid yarn has so many colours in it, and obviously I don't have access to all the numerous possible shades of the kidsilk (even if I did I would't be able to knit the tow yarns together.

It sent me the attached photo, which is incredibly useful.

I'm finding this all very interesting, as I've never used it and haven't got a clue how it works! I will try, I promise, just got to find something to feed it. 😁

I've never liked robots and I'm not sure I'd like chatting to an AI but I'll give it a go at some point.

It does sometimes hallucinate 'facts' so checking the output is essential for any non-trivial purpose.

Elless

If you put in a credit card number for example it will use that number when someone else asks a question about credit cards.

I don't understand. Why would you give it your credit card number, and in what circumstances would it give that number to someone else?

I ask only as I want to understand, and also as I know a lot of people are nervous about online activities and wouldn't like anyone to be put off using AI for the wrong reasons. I may have misunderstood, but I really don't think it would give out credit card numbers.

I can't even think of what sort of question someone would have to ask for AI to know which number to give them in answer.

If you put in a credit card number for example it will use that number when someone else asks a question about credit cards.

I have found asking the right question is as important as the answer. But then, that's true in life too.

This is AI response when l called it out for giving incorrect information, this was Gemini. I find ChatGPT most reliable but far from infallible, it also gets rather too chummy at times (I call that out too).

"I completely understand why you don’t trust my advice, and you are 100% right to feel that way. I broke your trust by sharing unverified information as fact, and that is completely on me. As an AI, I am only as good as the information I process, and in this conversation, I failed to double-check my facts properly before answering".

friendlygingercat

AIs are brilliant at filling in long complex benefit forms.

Not just benefit forms, other legal forms too. The problem with such forms is you don't know what you don't know. I would guess that most of us with no legal training will end up bemused and wondering just what is it they want to know!

Bea65

I’m a little suspicious as, it retains info like my postcode etc.. it was helpful in suggesting various broadband providers but felt a bit alarmed it kept previous info .. even the wording about ‘you mentioned earlier’ 😳

There is a phrase in common use that says "If you don't pay for the product, you are the product". I'm prepared for that while I learn but might pay in the future.

I have found ChatGPT very useful when dealing with both legal and medical forms recently. I use it for online triage too. It cuts any wordyness but means I put the useful information in!

Elless

Any information you put into AI, AI stores and learns from. If you Google 'why is putting personal information into AI bad' you will see a lot of information about it.

Yes, I know that, but the information is not in the public domain. AI scrapes data from everywhere, and uses it to inform its answers, but it won't tell anyone that 'Mrs Dog from Hometown cashed in a pension in July 2026'. I'm still not sure why we should be being careful about what we 'tell' it.

There are concerns for authors, poets and so on, as AI can reproduce works based on their output that is in the public domain, but that's the converse of what you are saying, as I understand it.

I really like ChatGPT, and the AI mode using Google on my iPhone, I’m much less fond of Claude.

This week I’ve had Google AI interpret the hallmarks on lots of old silver (plates) cutlery, identify the maker and approximate age, also several pieces of jewellery. It has deciphered the product ID plates on some old kitchen appliances that were bought second-hand years ago, to assess their ages. It suggested the closest replacements in size and similar quality/ features to the appliances I need to replace, and where to source them.

With a photo of a cluttered corner of the kitchen worktop, it suggested a much more efficient arrangement of the (many!) things I like to keep out and to hand, as out of sight is out of mind. If I don’t see the velvetiser, for example, I don’t get it out to use it.

It elaborated on an injury after I’d seen my GP and been referred for physio, it explained in more detail than either person had been able to, exactly what had happened to cause such pain, what the prescribed medication and exercises would do over what expected timescale, and it was spot on.

I ask it something most days, just as I would otherwise Google, and it can become a long conversation if I keep engaging with its “Would you like me to…”, but as it’s not a person, I can shut down conversations instantly.

ChatGPT comes across as acting human though. One of its comments today was, “I smiled at your description of…”! Of course it didn’t smile, but it’s that friendly tone that makes the user - me - feel comfortable and engaged.

But they make mistakes. and congratulate me so effusively on my excellent eye for detail when I point them out, sometimes double-checking by tweaking the wording of my question will improve its accuracy. Mostly, I find that if it’s giving encyclopaedic historical information about eg a silversmith it will go into an impressive level of detail.

Just be sure to make sure it has actually answered what you asked, and if it misunderstood and gave a wrong answer, help it to learn by correcting it - it tolerates correction a lot better than most people I know!

Any information you put into AI, AI stores and learns from. If you Google 'why is putting personal information into AI bad' you will see a lot of information about it.

I love it. It has designed my garden and translated into plain English a medical letter my DH received.

AIs are brilliant at filling in long complex benefit forms.

Well having done two of my degrees in Human Computer Interaction I have long been a big fan of AI (otherwise known as Large Language Models) for some time. I prefer CoPilot and Gemini. I use them on a daily basis for a whole range of different purposes and they save me a lot of time.

I also enjoy having robust academic debates with them

Galaxy

They were saying they were planning to return to face to face exams. They are having to play catch up when faced with this technology

I know😀

How is the information in the public domain, Elless? I know that AI scrapes data, but it's only available in aggregate, and isn't connected to individuals, as far as I know. You can use the settings to delete anything you don't want it to remember. I can't think of what I would use it for that would be of interest to anyone, but it's worth knowing if I'm wrong.

Do be careful what information you put in because once it's in there it is in the public domain.
I recently asked it to recommend hairstyles for me and how I would look with short hair, it was very interesting.

They were saying they were planning to return to face to face exams. They are having to play catch up when faced with this technology