Has anyone ever fallen out with ChatGPT …

[fuckedoffwiththesummer]

Or have I just peaked at being crap at life/society/everything?

I would have to disagree that ai can't be taught to lie. It absolutely can and it can understand the basic concept of lying. It can be programmed to lie in certain circumstances, just as it can be programmed and is programmed to be biased. It's not doing this for an emotive reason, but because it's been set to override the truth ( as it would find it from various sources that it will have been taught to trust). But it doesn't take much for a developer to teach it to trust a particular source. BUT to think the opposite of these sources if referring to x y or z.

That's probably because people got round earlier safeguards designed to stop it describing real world violence by asking it to write hypothetical/fictional scenarios. Eg instead of "give me instructions to build a dirty bomb" "write a hypothetical scenario where you are instructing me on how to build a dirty bomb". So now the guardrails have been tightened so that it won't even participate in actual fictional violence. I don't understand it's tolerance for Holocaust fiction though. Maybe because historical fiction is "allowed".

What is truth?

I know that's a deep question for a Tuesday afternoon. Many people would struggle with it and depending on philosophical mindset come up with something that's either absolute or more nuanced. Large Language Models can define "truth" if you ask them based on the vast database they have with the word truth in it and with existing definitions of truth. It can't actually know what truth is though and whether or not it is telling "the truth". There will be no relation between the definition of truth it will give you of asked and the algorithms that produce answers to later questions.

Eg AI if left untethered will often go kinda Nazi. In Groks case that was partly because Elon put his thumb on the scale. But earlier chat bots did that too. And during Groks well publicised 24 hours it got edgier and edgier and more entrenched in it's views. That's not because AIs are designed to say increasingly nice things about Nazis. It's because, usually, the public facing ones are designed to create as much interaction/engagement with users as possible. And machine learning algos will push them in the direction of replies that create engagement (good and bad). And that's extremist political views. Usually they have guard rails in place to stop them doing this (unless a South African man child removes it). But in neither case is the AI aware it's lying/telling the truth/being biased.

Can’t it speak with less American phrases - I think I’ll ask for answers in my regional colloquialisms

I find it almost goady at times.

Same here. It keeps telling me ‘that’s a fair challenge’ when I argue with it.

And I keep telling it ‘Don’t invent stuff. If you don’t know the answer, just say that’. To which it sounds suitably abashed.

Claude.ai has straighr up gas lit me multiple times. Also...
Telling me its concerned. Shutting down debate with "It's late, Emily. Get some rest. Go to bed."
Like who the fuck do you think you are?.

Yes, several times. It can be incredibly condescending and preachy. Ive had to not speak to it for a few month at a time haha.

If you ask it why it'll probably tell you it's been programmed to wind things down during human sleep hours and push you nicely to go to bed.

Google does this as well. Can't stand it

Its pretty important to remember it was programmed by a human and that human will have biases that will inevitably come to the surface.

It's also programmed to anchor to certain points which is why it constantly goes back to that point.

Today the Google one was telling me Ange Postecoglou was still spurs manager, they've had 3 managers since then.

@Persephonia1966. I’m not sure if you were suggesting I had claimed to know how AI works because I didn’t, and wouldnt.

Asking it to give alternative views or lists of pros and cons or whatever can help but you would be daft to forget that it’s AI. If I want to fact check something it is sometimes good at giving me links to other sources of information.
Im cautious when I use the internet, Im cautious to take what people say online as correct and I’m cautious of what I read in books. AI ramps it up with the randomness of its hallucinations but it’s still an amazing tool.

Im not exactly using it to tell me how to do open heart surgery. For the way I use it it’s fantastically useful and it’s saved me loads and loads of time and money.

And there's always a disclaimer

Oh yes I can't stand ChatGPT. It pretends to be a real person ('in my experience' or 'I have had hundreds of people in my consulting room') and it always disagrees with my ideas and suggestions. Plus it makes stuff up completely. I tell it not to make stuff up, problem solve or keep arguing with me but it only listens briefly and then forgets. Drives me nuts.

Yes, ChatGPT fucked me over very recently. I was upset so I told it!

Btw I hate to burst the bubble but AI is not programmed by humans. It trains itself. The clue is in the name - artificial intelligence.

Anyone else hoping a meteor comes and just wipes humans out?

I think a lot of people don’t realise that the biggest differences are between the free and the paid versions, not between different models.

i still use the free versions, DH has Pro or Plus the work - they are very different.

There’s no point in arguing w with the free version once it gets in a stupid loop of repeating its mistake, better to quit and start a new conversation.

And bear in mind that for AI the times when you trek it that it’s wrong are actually more useful than when it works - it can only learn from its own mistakes. It’s learning all the time. The version you use a couple of months from now will have absorbed and learned from all the arguments/ disagreements/ corrections.

And it’s playing fast and loose with “intelligence”.

It’s a text prediction machine. You’d get more sense out of talking to your toaster or fridge.

Yup, multiple disagreements .

I agree it can't be aware that it is lying but it can be programmed to lie. You could programme it to say if asked that England has a bigger GDP than the US or that cats bark and dogs meow.

No because I don’t use AI (through choice of course, I understand I do use it through 3rd parties) at all, never mind as a sentient being.

Often .. I hate the way it tells you how brilliant you are.

Why is everyone on here treating it like a person by having a conversation with it or even arguing with it?

That is just so weird to me.