To think ChatGPT is the best thing ever?

I think there already are. The SEO industry has rapidly shifted to optimising for AI rather than google rankings.

It hallucinates and it keeps your data and uses it for training the LLM. Environmentally it’s not great either.

I think AI makes me cleverer. It gives me the confidence to give almost anything a try. As I mentioned in an earlier post I do a lot of DIY property maintenance and car repairs etc and AI is very helpful. It’s so fast and, generally, accurate. I take a photo of something and then it tells me how to do something. I often then explore the issue further. AI presents the information so clearly and quickly. It helps me learn. I’m quite a techy person and perfectly capable of googling things but AI is so much better.
Im not stupid and I’m not going to become stupid by using AI. I know to double check things and I know the limitations of AI.
The pp who mentioned AI told her to go to A&E because she missed some periods presumably missed the clear AI medical disclaimer which reads something like “I am an AI, not a doctor. Do not use this information to make medical decisions” It’s the same as googling some thing or asking on Mumsnet, you have to blindly follow a suggestion like that while, presumably, ignoring the rest of the advice ai would have given alongside.

Also, do people who claim they will never use AI actually understand how daft that sounds.

You do realise that Google is powered by AI. It summarises what it finds into natural language outputs rather than what it used to do, just signpost to relevant websites. The narrative is Google Gemini, its AI model, summarising what it has found, trawling the internet for relevant info to satisfy the search

there is nowhere nowadays that isn't integrated with AI in one form or another.

If you ask it not to make things up in settings, then also ask it to provide the link to the info it gives you or ringfence where it should look (I.e. only in the documents you have uploaded and to tell you where in those documents it’s found it), you can check the output quite easily. I’ve got the paid version, maybe that’s different.

It is to me. But I guess I do focus on the bits it tells me that seems wrong or I disagree with.
And it pissed me off because it came up and annoyed me. So I remembered it.

Whenever I've tried it it gets things wrong. I assume this is because its fed by the most popular search results rather than having any discretion.
I find just googling gives faster and better results because I can weed out the crap myself, which ChatGBT cannot do.

I liken it to a calculator, do they make people lazy for not working out the maths themselves, no, they make them more efficient.

You must have some coding skills though? To be able to check that the code it wrote for you was correct? Did it also act as a tester to test that the app works as intended?

No it’s shite and will result in humans being a species incapable of independent thought.

I’ve heard of situations where employees get into grievance hearings, are asked about points in their letter, and know absolutely nothing about them

This is what I don't understand. For the people who think that ChatGPT is a good idea for this sort of thing, wouldn't they at least sense check what it had written and whether it was a reasonable representation of what they wanted to say?

See also academic work. No-one is fooled that you used AI, it's blindly obvious to an expert when the content is nonsense so wouldn't you at least only submit what you actually understood?

I think it might have had some effect, as I asked it just now and it it's given me almost exact quotes but has chopped them up a little bit. When challenged it said it assumed I was looking for usable excerpts for discussion so it gave shortened quotations. Said it should have defaulted to verbatim quotations. So it assumes things, unless you repeatedly emphasise what you need.

Huge fan of Chatgbt use it for all sorts of things, find it so helpful.

Hm, this makes me wonder, what is the effect when AI admits to a mistake? Does that affect what answers it gives later to simlar questions, to you or to others? Does that decrease its confidence in its response? Is it using reinforcement learning? So many questions....

Snap.

It’s tiring advising on 25 page grievances which sound impressive but say absolutely nothing of any substance, and contain assertions that have no legal basis whatsoever.

How is in the best interests of anyone (employee included) to have to wade through reems of nonsense to try to work out what the actual complaint is.

I’ve heard of situations where employees get into grievance hearings, are asked about points in their letter, and know absolutely nothing about them.

I always eye roll when people are so ignorant towards the environmental impact. Instead of jumping on the trend like minions using valuable resources to pick your wallpaper.
I have the very basic in education skills yet it floors me that apparently intelligent educated people are so greedy and stupid.
The technology companies will be happily taking the profits from the minions.

LLMs are constructed from the raw material of online human made content.
Because they are useful and popular LLMs are increasingly replacing human-made content.
You can see where this is going ... the LLMs are eating themselves and will disappear up their own arseholes.

It's making people stupid and lazy, not to mention the environmental cost.

DH uses it to write code. He was computer programming in the 70s and started work writing programmes in binary. So he's very good at writing good prompts. Also the answer it gives in this kind of use can be verified and tested.

The problem with wider societal arguments against AI, whilst enormously important, is that very few people care enough to change their own behaviour in order to be part of a societal shift. Like climate change generally.

People seemingly don't give a shit about their own data, or the very real potential of mass surveillance, which astonishes me. The growing use of facial recognition tech, plus AI should give pause, but usually doesn't. How many use use a VPN, Tor browser, or temporary email addresses to sign up to things?

The only arguments that people seem to hear are 'it will give you wrong answers' (true, but not always and perhaps just infrequently enough that we start believing it) and 'it does you personally damage'. Which not everyone seems to get. There is already evidence that having Google available (phone in your pocket, or a computer) makes people overestimate what they actually know or remember.

A huge danger of LLMs (and this is already happening, due to the internet generally) is growth of radical cynicism. When you can't trust previously-trusted forms of news or data or information, and we are completely bathed in endless streams of information, it becomes more likely that you believe and trust nothing. That way lies conspiracy mindedness and political radicalisation. Its not a new trend, but AI risks speeding this exponentially, as it does everything else.

The argument of 'if you don't use it, you'll be left behind' is insane. It is literally designed for dummies. You don't have to get 'good' at using it. You have to get good at critically analysing the output - and the skills for this aren't learnt by using AI.

I also tried this but just a heads up - it isn’t always accurate as people have said. It told
me to take the kids to Istanbul Legoland which has literally been shut down for years and I had to correct it.

@daisychain01 Seems a bit ironic to use it for help with environmental issues 😩😂

And people say don't use ai it gets things wrong while confidently giving an incorrect answer 😂😂😂

Well if it told you to go to hospital under some circumstance as one point in a wall of well reasoned common sense advice that’s not really the same as “I told it I had my period and it told me to go to hospital” is it?

It guzzles resources, lies to you, takes your inputs and reuses them, and has just put thousands of people out of jobs where I live. Just use Google.