Artificial Intelligence replacing people

WithNobsOnIt

Have said this before.
AI is still in its infancy. Just learning to crawl.

Be afraid. Be very afraid.

Learning two languages over the weekend as it crawls too!😲

The thing is you can’t force the genie back into the bottle.

An item in The Guardian last month raised the possibility that there were indications that Artificial Intelligence was learning to resist attempts to disable it or limit it's capacity.

Thinking for itself, rather like HAL the rogue computer in the film '2001 A Space Odyssey'?

Because human emotions and failings are also online AI is incorporating those in its replies.
My sister in law is a vet and uses AI to compose letters to clients, basically, "we tried our best to save your dog but it died, you now owe us $10k."
AI does it so much better

Does AI read Trump's Truth Social? Is it incorporating/absorbing all of that 'literature' into it's mindset?

David49

Because human emotions and failings are also online AI is incorporating those in its replies.
My sister in law is a vet and uses AI to compose letters to clients, basically, "we tried our best to save your dog but it died, you now owe us $10k."
AI does it so much better

Our GP surgery use Ani AI. I recently had an asthma review and everything that the nurse and I discussed was comprehensively written up and added to my medical record. The note appeared in the NHS App within minutes. Surely, this is progress.

I recall going to school and using Log Tables in maths. Then came along the slide rule followed by the school getting a calculator as big as an old typewriter. It was kept under lock and key.

Much of what we do these days relies on a computer. However, even with AI the adage GIGO (garbage in - garbage out) will still apply. The challenge for teachers and trainers going forward is how are we going to give students enough knowledge to know when the computer output is wrong?

lizzypopbottle

My daughter asked AI to recommend some peer reviewed research papers. It complied. When she tried to find them to read them for her work on her MSc. she couldn't find them. The didn't exist! The AI had made them up and when she asked why it had done that, it apologised and said it hadn't wanted to disappoint her!

You can't trust everything AI responds with.

Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?

Daddima

lizzypopbottle

My daughter asked AI to recommend some peer reviewed research papers. It complied. When she tried to find them to read them for her work on her MSc. she couldn't find them. The didn't exist! The AI had made them up and when she asked why it had done that, it apologised and said it hadn't wanted to disappoint her!

You can't trust everything AI responds with.

Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?

Here’s what AI said-

‘What’s happening

AI can hallucinate citations, including:
•Research papers
•Authors
•Journal names
•DOIs or arXiv numbers

It does this when it’s trying to be helpful but doesn’t actually have a real paper to anchor to.

Why this occurs (not emotional)
•AI predicts what a plausible citation looks like
•It doesn’t have a live database of all papers
•When asked for references, it may generate:
•Real-sounding titles
•Correct-looking author lists
•Believable journals and years

This is a statistical completion error, not a choice to spare feelings.

Best practice
•Treat AI-provided citations as leads, not sources
•Always verify via Google Scholar, PubMed, arXiv, or journal sites

Bottom line
•✅ Yes, AI can recommend papers that don’t exist
•❌ Not to protect feelings
•⚠️ It’s a known failure mode called citation hallucination

Well, every day’s a school day!

I put my own name into AI and it told me I was a lecturer on international studies at a southern university. First I knew about it. My real name is unique.

GoodAfternoonTea

I put my own name into AI and it told me I was a lecturer on international studies at a southern university. First I knew about it. My real name is unique.

Are you sure it’s unique? And have you googled to see if there is indeed a lecturer, as the name may be only slightly different from yours?

Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?

Surely we know enough about AI to now that it has to be treated with the utmost caution because it *ISN'T' purely factual.

I guess that those parents who, if they had been born in another era, would hsve introduced their children to encyclopedias, the library, or contextual activities will be those who are beside their children in learning to use this tool too.

MaizieD

^Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?^

Surely we know enough about AI to now that it has to be treated with the utmost caution because it *ISN'T' purely factual.

But it can be usefully so. It's very polite if you query it. Shouldn't the student learn to ask for authenticated papers. How you put the question directs
thd answer.

AI includes reputable sites and also disreputable sites.
Same as any other secondary source of information, you have to look into the credentials of the source.

MaizieD

^Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?^

Surely we know enough about AI to now that it has to be treated with the utmost caution because it *ISN'T' purely factual.

ChatGPT is not purely factual , as no secondary source is "purely factual".

Chat GPT will give you advice 1, if you ask for specific advice and 2. if you ask for advice that ChatGPT is trained to be able to give. ChatGPT will refuse to give you illegal or harmful advice.

This is definitely a question of parental guidance, one (not just kids, but everyone in general) needs to distinguish between questions that could be addressed to ChatGPT (or looked up via Google) and questions that are better discussed between people. There is nothing wrong with asking AI how to repair something or asking for a list of pros and cons of each item, etc. Sometimes it gives really great advice on tech and software, I wouldn't even know about programs like Gardenbox 3d for garden planning or Todoist for task management if it wasn't for ChatGPT's recommendation. It can even help you with journaling and navigating the stream of thoughts you have.
However, you shouldn't use it to solve your personal problems with other people, these things need to be discussed personally. AI algorithms are trained to be supportive no matter what, but it's important that you realize your mistake if there is one.

AI is just another tool and it's very popular, so instead of being scared of it we need learn how to use it wisely and teach our kids that.

Caleo

MaizieD

Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?

Surely we know enough about AI to now that it has to be treated with the utmost caution because it *ISN'T' purely factual.

ChatGPT is not purely factual , as no secondary source is "purely factual".

Chat GPT will give you advice 1, if you ask for specific advice and 2. if you ask for advice that ChatGPT is trained to be able to give. ChatGPT will refuse to give you illegal or harmful advice.

I DIDN'T' say AI is purely factual. I was quoting someone else, who did say it.

Every post I've made on this thread has said AI makes things up.

I use it myself because it can draw data and information together which would take ages to do oneself, but I note that it does have a tendency to try to please with duff stuff, like Grok and that it lays on the flattery, telling me what wonderful and perceptive questions I have asked, or what excellent points I have made.

I think that makes it even more dangerous. It's not just giving factual information, it's trying to 'attach' one to it (in the psychological sense of the word. I wonder if that is about keeping you as a loyal user in order to eventually monetise your usage?

At least dictionaries, encyclopaedias etc just give good information without telling the enquirer how wonderful they are...

Oh, and I also forgot to add that you need to check everything it tells you because AI is notorious for making up facts, titles, names and data. Just like every other online source, everything that comes from it needs to be double checked, which is a very important skill to have as it helps developing critical thinking.

Chat GP is not infallible ( neither are parents). Both parents and children need to understand that.

By age 12 my parents weaknesses (and lived example) taught me that I needed to take responsibility for myself, think for myself, do the hard graft by myself for myself, because nobody else was going to do it for me.

It was a tough lesson learned the hard way, and has turned out to be a huge advantage throughout life.

Must say I worry about it replacing many people’s jobs, but have to say my dd finds it extremely useful when she has to produce e.g 30 page proposal documents -it’d take her many times longer otherwise. She gives it the requirements - and tells is to use British English, not US.
And of course always checks the final documents thoroughly.

MaizieD

There we go, I hadn't read lizzypopbottle's post. That's a clear example of AI's dangers...

So we have to learn to use it, learn what controls it needs and where we need extra care. We've done it before.

One of the major things I think we need everyone to learn is how to ask the right question. If lizzypopbottle's had asked for information "that can be referenced" it may have given her the information she wanted.

MaizieD

Caleo

MaizieD

Now, that is very worrying. Surely the point of AI is that it is entirely factual, and saying ‘it didn’t want to disappoint her’ implies some sort of feeling?

Surely we know enough about AI to now that it has to be treated with the utmost caution because it *ISN'T' purely factual.

ChatGPT is not purely factual , as no secondary source is "purely factual".

Chat GPT will give you advice 1, if you ask for specific advice and 2. if you ask for advice that ChatGPT is trained to be able to give. ChatGPT will refuse to give you illegal or harmful advice.

I DIDN'T' say AI is purely factual. I was quoting someone else, who did say it.

Every post I've made on this thread has said AI makes things up.

I use it myself because it can draw data and information together which would take ages to do oneself, but I note that it does have a tendency to try to please with duff stuff, like Grok and that it lays on the flattery, telling me what wonderful and perceptive questions I have asked, or what excellent points I have made.

I think that makes it even more dangerous. It's not just giving factual information, it's trying to 'attach' one to it (in the psychological sense of the word. I wonder if that is about keeping you as a loyal user in order to eventually monetise your usage?

At least dictionaries, encyclopaedias etc just give good information without telling the enquirer how wonderful they are...

I have sometimes instructed ChatGPT to stop flattering me in its replies and it responds appropriately. It is just a machine and can adjust its tone to suit the user.

We are assuming AI responses are going to be benign, it’s when they are use to distort facts that will be the threat. AI and automation is already replacing service jobs leaving only the most menial work for humans

I use ChatGPT a lot.
Problem with my phone? - it takes me through the solution?
Problem with an ambiguous label on a carton of milk? - it clarifies the storage safety.
Problem with a new medication? - it takes me through the details.
Problem finding a private clinic to do my surgery? - it looks up the best options for me and gives contact details.
Need a whacky poster for a local event to put on facebook? - it creates something in about 3 minutes.
Need some help with the analysis of a poet's works for my poetry group? - there it all is in a couple of seconds.
Trying to trace the best online supermarket for low air miles? - options come up straight away
Wanting to find suppliers of a musical instrument? - there they all are.
Gardening advice foir a specific plant? - there it is
Have a difficult letter to write with subtle nuances in order not to give offence? - it will suggest lots of tweaks and remodel as you change your wishes.

I could go on. If I have a technical problem with an appliance I go straight to ChatGPT rather than fight my way through emails/unanswered phone calls/prevarications.

For me it is just a speedy way of finding stuff out. Anything medical I check what it has said in other ways through dedicated organizations, but so far have found it spot on.

All it is is a speedier way of getting information than going to the library and ordering a book. It is a blessing for me living as I do on my own in the middle of nowhere. It has saved me bothering my chidlren with technical questions on many occasions.