Active discussions

Meet the Other Phone. Flexible and made to last.

Meet the Other Phone.
Flexible and made to last.

Buy now

Please or to access all these features

Talk
Chat

Join the discussion and chat with other Mumsnetters about everyday life, relationships and parenting.

Original poster

The threat to humanity from AI

111 replies

Youreeavinalaff · 01/04/2023 14:03

Just read a terrifying article about the threat AI poses to humanity.
https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/
ChatGPT (AI chatbot) is amazing - my kids have been talking with it and setting it tasks and challenges. It's incredible. Apparently it's making such sudden and huge leaps in its abilities that Elon Musk and others are petitioning to pause it along with AI labs and programmes for 6 months whilst safety mechanisms and limitations are put in place. Reading this article however makes me feel that these won't be sufficient. We're about the unleash something we can't control by the look of it - or something we can't persuade all nations to agree to. I am worried.

The Open Letter on AI Doesn't Go Far Enough

One of the earliest researchers to analyze the prospect of powerful Artificial Intelligence warns of a bleak scenario

https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough

OP posts:
stargirl1701 · 01/04/2023 21:38

You need some dystopian sci-fi, OP.

The 100 is a show where an AI takes power.

Person of Interest has 2 battling AIs.

And, the classic: The Terminator. 3 billion human lives ended in on August 29th, 1997...

BertieBotts · 01/04/2023 21:51

Well MyCat put it much better and more succinctly than my ramble 😂

Mycathatesmecuddling · 01/04/2023 21:54

BertieBotts · 01/04/2023 21:51

Well MyCat put it much better and more succinctly than my ramble 😂

To be fair I loved your - how do we reduce the need for paperclips - kill all humans example. Shows how easily it could all go wrong if someone is having an absentminded day 😂

"Sure ask the AI about paperclips, it's not like it's nuclear war, what could possibly go wrong...."

Switchwitch · 01/04/2023 21:57

But we can easily stop AI by saying "I know you are but what am I" on a loop.

Mycathatesmecuddling · 01/04/2023 22:03

Switchwitch · 01/04/2023 21:57

But we can easily stop AI by saying "I know you are but what am I" on a loop.

So basically AI needs the equivalent of a child proof lid 😂

In fairness what we have now is not really the true AI which people mean when they talk about the technological singularity

Although I do wonder if anyone ever considered how successful even that kind of AI would be if you put it in a room full of 6 year olds 😂

AI didn't destroy the world it had a nervous breakdown and gave school teachers a 200% raise instead...

shadypines · 01/04/2023 23:17

I know it's sad @UnDruidlyWords I saw a mother walking her 2/3 yr old along main road to nursery, on her phone, seemingly oblivious to the fact he kept stepping into the road and then she crossed the side road without looking. Sweet Jesus where is the parenting?
I think these are the people AI will take advantage of, those whose common sense and care has already gone to batshit.

Hawkins003 · 01/04/2023 23:23

Youreeavinalaff · 01/04/2023 14:03

Just read a terrifying article about the threat AI poses to humanity.
https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/
ChatGPT (AI chatbot) is amazing - my kids have been talking with it and setting it tasks and challenges. It's incredible. Apparently it's making such sudden and huge leaps in its abilities that Elon Musk and others are petitioning to pause it along with AI labs and programmes for 6 months whilst safety mechanisms and limitations are put in place. Reading this article however makes me feel that these won't be sufficient. We're about the unleash something we can't control by the look of it - or something we can't persuade all nations to agree to. I am worried.

More likely so they can monetise it,
We also have mid journey for art and it's holy spirit. @Youreeavinalaff

Hawkins003 · 01/04/2023 23:24

The designers, developers ect knew ai was advancing eons ago,

Original poster
Youreeavinalaff · 01/04/2023 23:53

Thanks stargirl1701 for the Sci fi recommendations, I haven't watched any of them so will look them up - I generally love a dystopian drama, although post zombie apocalyptic worlds are usually my thing. I just re-watched Bladerunner with my daughter. It questions what happens when AI does develop feelings, emotions and desires. Everyone seems to assume that it will remain robotic and inhuman and will never understand the nuances of being human, and take everything literally, but I don't know why that would be the case if it had access to all human knowledge and experience.

I think it would be naive to assume that we can contain AI as an advanced amalgamation of Alexa, Google etc which will simply continue to serve us. Maybe in the short term, but looking at some of those graphs and data provided by previous posters, the exponential evolution of AI suggests that it will turn into something else quite quickly.
Yes, I've seen some of the art, it's impressive. We were also getting chatgpt to write poetry for us on a variety of topics in the style of different poets last weekend and it was astonishingly good - some of them could easily pass for rediscovered lost works. Like a previous poster said, it's the arts that are probably going to be made redundant quickest.

OP posts:
Hawkins003 · 02/04/2023 00:02

@Youreeavinalaff an excellent show with the capabilities for ai is called person of interest, with the two artificial super intelligences, called the machine and Samaritan

MobilityCat · 02/04/2023 00:43

shadypines · 01/04/2023 23:17

I know it's sad @UnDruidlyWords I saw a mother walking her 2/3 yr old along main road to nursery, on her phone, seemingly oblivious to the fact he kept stepping into the road and then she crossed the side road without looking. Sweet Jesus where is the parenting?
I think these are the people AI will take advantage of, those whose common sense and care has already gone to batshit.

Driverless cars use AI already to protect "mother and child"

JamSandle · 02/04/2023 00:45

MobilityCat · 01/04/2023 19:03

I've had Alexa for years and all it really does apart from controlling WiFi connected devices, it just seems to search the Web for answers to any questions.

Alexa is a much more 'early doors' form of AI.

PussBilledDuckyPlait · 02/04/2023 00:50

Switchwitch · 01/04/2023 21:57

But we can easily stop AI by saying "I know you are but what am I" on a loop.

Surely we can even more easily stop it by pulling out the plug.

JamSandle · 02/04/2023 00:51

I think it's an awful and terrifying next step in our story.

FlosCampi · 02/04/2023 01:02

I'm slightly disappointed you're sensible OP, I was hoping for the seamonsters /royal lizards/ alien motorways conspiracy theorist who sometimes posts late at night!

Dracuuule · 02/04/2023 09:33

I think the most dangerous thing about AI is that it gathers and utilises data that is already accessible out there by using specific algorithms. It can be easily manipulated. You can ask certain sensitive questions about race or trans ideology and it will give certain answers that are from a particular viewpoint.
I'm not saying it's right or wrong but it only follows the biases that algorithm writers have or the information it can access.
If a lot of information out there believes humanity is the cause of climate change, then yes, it may conclude that humanity should be wiped out.
If all the information on the cause is out there as specifically being air travel, it may ground planes as a solution if it was capable of doing so.
This is what makes it potentially unsafe.

furtheringeducation · 02/04/2023 09:38

sorrynotathome · 01/04/2023 15:18

Climate change is FAR more of an immediate and obvious threat to humanity. I think we should ask AI to work on that.

In that case AI would definitely get rid of humans as we are the cause of climate change

Dracuuule · 02/04/2023 09:48

"In that case AI would definitely get rid of humans as we are the cause of climate change"

But we can also be seen as part of the solution, if the information about it is out there. AI cannot think for itself and just gathers information from the internet. That's what it's learning from.

Mycathatesmecuddling · 02/04/2023 11:23

PussBilledDuckyPlait · 02/04/2023 00:50

Surely we can even more easily stop it by pulling out the plug.

Show quote history

If AI is in the cloud you might not be near the plug

You could, for example, try to shut down the cloud instance that is running the AI, until the AI takes control of that

But these things are not necessarily running on one computer with one plug

Although I personally think we are some ways of an AI that takes over if it ever happens

PussBilledDuckyPlait · 02/04/2023 11:27

Mycathatesmecuddling · 02/04/2023 11:23

If AI is in the cloud you might not be near the plug

You could, for example, try to shut down the cloud instance that is running the AI, until the AI takes control of that

But these things are not necessarily running on one computer with one plug

Although I personally think we are some ways of an AI that takes over if it ever happens

Show quote history

In that case, take a hammer to the computer.

Mycathatesmecuddling · 02/04/2023 11:37

PussBilledDuckyPlait · 02/04/2023 11:27

In that case, take a hammer to the computer.

Show quote history

If I am running an AI programme in Wales and the datacentre it's being hosted in is in Indiana, or even Ireland, how exactly is that going to happen

If you type something into Chatgpt online do you really think plugging your computer out makes it stop running?

Hawkins003 · 02/04/2023 11:46

PussBilledDuckyPlait · 02/04/2023 00:50

Surely we can even more easily stop it by pulling out the plug.

Show quote history

Think of ai as skynet, basically it would be in the majority of systems and computers, you would effectively have to use a virus to shut down the software, as their will be many data centres all connected, yes in the home pulling the plug shuts off your Wi-Fi but then the ai will still exist.

PussBilledDuckyPlait · 02/04/2023 11:47

Mycathatesmecuddling · 02/04/2023 11:37

If I am running an AI programme in Wales and the datacentre it's being hosted in is in Indiana, or even Ireland, how exactly is that going to happen

If you type something into Chatgpt online do you really think plugging your computer out makes it stop running?

Show quote history

It would stop it having any effect on me.

Hawkins003 · 02/04/2023 11:49

BertieBotts · 01/04/2023 19:26

There is no advantage to an AI to wipe out all humans so why would it choose it?

The difficulty here is in wording the goal. There is a really good explanation on the wait but why AI post linked above. But in short if the AI's goal is to find the most efficient way to create paperclips for example, it could potentially and eventually get so smart that it's just turning all matter on earth into paperclips. Let's say we thought about overproducing paperclips and set the goal instead to be for the machine to produce as many paperclips as humans need in real time, based on evaluating previous usage and predicting future usage. Well, the most efficient number to produce is 0. So it's actually in the machine's interest to ensure that nobody needs any more paperclips ever again. It could just create another machine to create paperclips for it, or invent something that means paperclips are obsolete. But the simplest way to ensure that humans don't need any more paperclips is probably just to kill all the humans.

We can't just tell it not to harm humans because what does that even mean? We don't really understand morality very well ourselves let alone being able to explain it to a machine that doesn't understand humans. You could tell it to prevent human suffering and actually the most efficient way for it to do that would be for it to kill everyone. So you can tell it explicitly not to kill anyone, but in order to eliminate suffering it just kind of unplugs everyone's brain. No more suffering, but no more anything either. We think it's obvious that no suffering meant that we still want to experience all the nice things about being human, but that's not obvious to a machine that only understands literally exactly what you tell it.

Language is full of assumption and shared understanding, most of which is very cultural and learned. Culture and learning that a machine would not have and we take for granted so would not be able to program it in.

People have strange ideas about what AI might be, they think of it as like an intelligent animal, like you can teach crows to do simple puzzles. Or they think it's like teaching a child, not really understanding that we only directly teach children a fraction of what they learn, most of their learning is primed in the brain and comes from experimentation having been honed through evolution (with lots of death along the way). Think about all the daft experiments that toddlers like to do that could get them killed or harmed or other people killed or harmed if they had more body strength, ability to manipulate computers, or adults weren't around to stop them.

Or they think AI is some malevolent force like a movie villain who just wants power. That's a very human interpretation, it's unlikely that AI would "want" this - but it's possible that AI might see power as a means towards some other goal in a way that we fail to predict.

It's a bit like the game mass effect, a civilisation built an ai construct to try to prevent a war with humans when humans developed artificial intelligence, and the Ai construct learned learned that it was better to cull humanity before they got too advanced to build the artificial intelligence.

Mycathatesmecuddling · 02/04/2023 11:58

PussBilledDuckyPlait · 02/04/2023 11:47

It would stop it having any effect on me.

Show quote history

Depends what you have asked it really

The kind of AI we are talking about won't be stopped by plugging out one individual computer