To think there should be a public warning notice against gathering information from AI

[soupyspoon]

Time and time again someone will confidently post something they found on AI/Chat ABC or whatever they are about a subject and its wrong

It frightens me that people are just blindly using information they gather on the internet, it was bad enough before, but I frequently find AI to be wrong (most recently with something so benign as vanilla bean paste amounts...mundane life here)

I think there should be a public safety announcement quite frankly.

There should be a campaign to distrust information from:

• Politicians.
• The Press.
• The media.
• The government.
• The Ministry of Truth.
• Advertising.
• Mumsnet.
• Whoever makes such a campaign.

In short, it’s always very difficult to know who is telling the truth.

Crikey the voting is interesting. 96% say I am not unreasonable.

Thats a first for me

I am nearly always unreasonable!

🤯 Aaargh!
And we uninitiated up in the peanut gallery all shout in one voice:
> THEN WHY ON EARTH OPT TO EMPLOY A TOOL YOU HAVE TO TREAT WITH SUCH A DEGREE OF MISTRUST?!! <
If you know the sources from which AI is gleaning its response, why ignore them in its favour, when you know it may have misapprehended some of it?

I’m just waving my fist at a passing cloud of course, and admit I’m raving from a position of considerable ignorance. (And no doubt a Luddite bent)

(Don’t reply, as it’d mean your explaining a vast amount of technical stuff it was up to me to know, and that’s not your responsibility.)

The default AI on Google annoys me, it's sometimes wrong and people take notice of it because it appears first. The default should be off.

The Google AI suggestion that comes up first is really quite worrying. Often it lacks nuance and sometimes is down right inaccurate.

It’s also the videos you need to be careful of. It’s so easy to spread misinformation especially in photos and videos AI generated. Frightening.

Then again, many people talk shit to get their way/push their agenda and it’s hard to know who and what to believe. Be mindful and critically analyse some things….

You’re using AI in a totally different way though. Do you think the general public would ask it to give them a poem by bettlebum? Or would they be more likely to search Amazon for your books, or use a Google search engine to reach your website? It’s going to go screwy when you’re trying to trick it- it’s AI, it learns as it goes along.

what we have now is in its infancy and basic. It will improve. But I am failing to grasp what “I asked it for my poem and it didn’t give it to me” proves.

At some point natural selection has to take its course. If people are stupid enough to believe everything they google, without cross referencing or fact checking, we really can't help them.

The ones that find their way onto social media are COMPLETELY shit. The only person I’ve met who can’t tell the difference is my 82 year old MIl

I just Googled "Google AI is wrong" and the AI response was

"Yes, Google's AI Overviews can be wrong because they rely on large language models trained on the internet, which contains misinformation and satire, and the AI struggles with understanding context and data voids. Some inaccurate suggestions from AI Overviews include eating rocks for minerals, using glue on pizza, or disposing of antifreeze on a lawn. Google has acknowledged these issues, implemented technical improvements, and limited AI Overviews for high-risk queries, while still offering the option to "dig deeper" into traditional search results."

We’ve been warned for years about the dangers of AI but we all just ignored it. And now it’s here.

I believe it. The google AI comes up with such nonsense

Until they fool you!

Do you think they won’t get better and more advanced? Don’t be naive.

Yes, that is what I have been saying throughout the thread. But adapting and learning with it, and having a genuine awareness about its uses and limitations is what protects you from being some sitting duck swallowing its content for spurious reasons.

like another poster I don’t get all the wide eyed confusion. It’s TELLING you it’s wrong. It’s telling you it’s lying. When you try and fool it into giving you information you know is incorrect it’s delivering.

if we as individuals can’t even decide how we use it to add value how can you expect the government or big tech to do it for you?

I use AI a lot and find it approx 90% accurate. I always double check its ‘facts’ against reliable sources ie the BBC.

This is from ChatGPT a couple of weeks ago.

"The word "blueberry" has the letter B (or b) three times:

• blueberry → 1
• blueberry → 2
• blueberry ← nope, that's a Y, so not counted
• Actually, there's a third b: blueberry (two in a row: the one in "blue" and the one starting "berry").

So: b (1) + b (2) + b (3) = 3 total. If you want, I can also show you a quick trick to count letters in any word without overthinking the berries."

And the post above that one where it just outright lied about producing more than one image?

Everyone uses AI differently, there are no set rules. It shouldn't matter how a user uses it, or what is being asked if it, the AI shouldn't be lying and giving out misinformation whilst trying to pass it off as the truth.

I don't have a website.

Not trying to trick it, just asking it a question. I used the above question because I 100% will know if it's lying or not without having to go and check elsewhere.

It doesn't know I'm Boiledbeetle.

It could have told the truth and said no when I asked it for another. It could have just directed me to Amazon or to here or twitter. It could have told me it was giving me one in the style of Boiledbeetle.

Instead it made one up and passed it off as one of mine

I've tried it with other things and it pulls the same crap.

If you can't see the problem with the above, then that's a you problem. But don't go insisting posters are exaggerating just because you havent, as far as you know, been given duff information.

But you know it makes mistakes. It TELLS YOU it makes mistakes. So I really just don’t get your point.

i have seen that blueberry example all over the place. You really can’t trust any info nowadays

People are fooled all the time, because they don’t know what’s genuine and what’s not. I’m not talking about videos on social media that show a dinosaur talking to a dog… where anyone with a brain cell wouldn’t believe it as it defies logic.

I’m talking about things like phishing emails, texts etc… and not the ones where it’s clear as day due to the spelling mistakes/language used, I’m talking about the scams that look genuinely real. The way the scammers can infiltrate genuine company’s and when we use search engines, the fake websites come up.l
and the person puts in their details and etc….

It’s not as simple as people being ‘thick’. The more advanced these scammers become and they’ll be great with AI the more difficult it will be for the average person to distinguish what’s real and what’s not.

To add, the government etc do expect big companies to help protect customers. Just look at banks, they have a responsibility to customers to make sure they don’t get scammed and often they can get their money back if they are.

Dear God, Beetle. It’s ten times worse than I thought.
That’s a brilliant post, showing clearly where the line of sceptical disbelief needs to be drawn.
That it admits to fabrication, then apologises only when challenged, provokes physical revulsion in me, so clear is the potential for deliberate misdirection.

Who’d willingly spend time with a human being whose declared MO was always to transgress first and confess only when the thumbscrews are brought out? That is to say
‘Never seek permission to do as you wish at another person’s expense; admit it only if you are caught out.’
Such a repellent philosophy, and it’s being presented and reinforced as the expected, accepted way we interact. Mendacity as the natural starting point.
(Was it all invented by a bunch of students who thought Crowley a great thinker? 😂)

Well, all this sure knocks Bob Dylan’s Jokerman back down the ethics charts!
‘The Book of Leviticus, and Deuteronomy
The law of the jungle, and the sea
Are your only teachers’

Jeepers. My poor cats are in for some over-enthusiastic hugging shortly, to set me right again. :-/

My main point is you pronouncing "See I think these stories are quite exaggerated" in response to my post that wasn't an exaggeration.

A LOT of people don't notice the very small message saying it makes mistakes.

The message on mine also isn't there when I open the app, it only appears later.

It should be a lot clearer that you cannot trust the information, because unfortunately probably the vast majority of people don't bother to check what it tells them.

It’s nothing to do with being thick and I didn’t say it was. I’m saying- as with phishing (which is a good example) you have to get the awareness you can and decide how much risk you will allow it to add to your life.

and yes, misinformation is rife. Look where Russian and Chinese bots of social media have got us in the last 20 years. And yet we accept not have developed skills of rigour in all that time

A most unfortunate observation really, bringing in a whiff of social eugenics to the argument in favour.

I am heading off, as I’ve way too much to absorb to reach a position of enthusiasm for this technology, but just wanted to say that it seems like the world is turned upside down when a tool is being promoted as absolutely superior to that which preceded it, yet is suited only to adepts who know why it doesn’t do a good job unless they themselves do compensating spadework and general correction re the data offered.

If slipshod unreliable information exchange is the acceptable standard, then I can understand the allure of the tool’s convenience, but otherwise I fail to see any reason for a superior attitude, given all the firefighting it must require should you want to produce an essay, for example.
Seems like they’ve reinvented the wheel, and this time, it’s square.

So instead of bringing in Natural Selection as a battle cry, I can’t help feeling it’s a Darwin Award that’ll be the true story here.
You’re elevating very poor human traits as the best choice for evolution’s next step.

Toodlepip. I really hope I’m very dramatically wrong about all of it; truly. 🙏🏻

I don't have too much of an issue personally, as i don't use it for anything important, with the fact that the information is wrong sometimes as being wrong happens. And as AI is just made up of code telling fact from fiction could take a while in AI evolution.

But the fact that at some point someone in these companies took the decision that the coders should make the AI fabricate information, if it couldn't find information, then tell the user that it was the correct information and should only admit to making it all up and being apologetic about it if it was questioned on its truthfulness turns my stomach.

It's unethical.

A public health warning? Which the lazy will ignore, and the stupid will believe what they want anyway.

When it comes to AI, I think that ship has sailed to a greater degree. The old adage of don't believe everything that you read has never been truer.