How would AI actually “kill all humans”?

Interesting question. Something I am trying to get to grips with.

We do know that so far it hasn’t been possible (haven’t a clue why not) to ensure these AI models see human life to be preserved at all costs.

That although small at the moment, the models are beginning to act independently of their human “masters”

Not OpenAI. It’s Anthropic.

Good question, I’ve been mulling that. Would AI take over armed drones or nuclear power? BBC news app has some interesting articles on AI. One example showing it has no ‘conscience’ or moral code was someone who tried to get it to book them into a gym class. The AI booked them in going months ahead and kicked other gym members off the list!

Robots / computers - what ever you want to call them will become so intelligent they’ll take over.
They’ll be able to bypass all security and set off atomic bombs.
But before they do that they’ll have the intelligence to basically shut down everything. We’ll be at their mercy.

I think we are all aware that AI could be used by bad actors to create biological, chemical or other weapons or to shut down systems which are vital for human survival:

But two key areas where AI could harm are also possible without it:

1. Power centralisation and unfair distribution of benefits.

2. False or misleading information.

It’s already developing now in a political party that has just received £72 million from foreign-based donors for what it calls its war chest”

One of its major spox has said: Reform is like a garden weed. You can't kill us. But I think they could kill the rest of us.

According to this morning's BBC Laura Kuenssberg show, she interviewed an ex Anthropic researcher, and he said AI could kill us by 2030.

He spoke about various ways this might happen, including new viruses.

Interview on YouTube

www.youtube.com/watch?v=9norxWsFrOo

It does seem that there are a tidal wave of articles right now on the subject, the most alarmist, we've only got until the end of the decade. Whilst others mention the need to slow it down. The Genie seems to be well and truly out of it's bottle and whilst AI will undoubtedly have many benefits those will be cancelled out by the fact that we, the human race, may well be cancelled out so what use will those be ?

This is very worrying.

( taken from a BBC report )

The hardest question to answer, according to Coxon, was what an AI apocalypse would look like
One of the risks outlined in Amodei's comments was of a swarm of bots acting like a supercomputer that could take over the internet
this scenario could be realistic in six months to a year

I’m going to start stockpiling food and other essentials.

I absolutely refuse to worry.

All my childhoood we were told the atomic bomb would be the end of us, when WW3 started.
No-one appreciated my mother's mundane consideration, that throwing atomic bombs about makes no sense if you are trying to conquer the world, or just your neighbouring country, which is usually why anyone starts a war, but I still think she had a point.

Later that we were ruining the environment (true enough) and that climate change and before that DDT would cause famine and the ultimate destruction of mankind. Quite true if we and our governments do not react in time.

Every generation has their bogey-men. AI is the present new bogey. It may destroy mankind, or it may not. So we need to find out if this is a re asonable fear or noot, and how it can be controlled as some of the other threats have been or are being.

I don't think a tin of soup will cut it somehow.

True, but at least I could have a quick meal before we’re all blown to smithereens.

I worry about AI getting access to weapons control systems - they are not on the open internet (I hope) but there must be AI embedded in some of those guidance systems.

If it's just breakdown of systems leading to food shortage & energy meltdown etc then obviously it's not going to wipe out people in Bhutan who don't use much technology etc. The fact a lot of the execs in AI are buying land suggests it's the food issue they're worried about.

Sorry to take the question literally when a more light-hearted take on it is probably the best way given we can't change anything

The biggest problem would be if bots shut down water supplies. Won't matter how much food we've stored we won't last long without water.

Do others feel angry about what is potentially unfurling as to what we can only assume are a lack of safety measures in how humanity will be impacted? We are now continually being advised what is coming on stream will be awful for humanity by those who appear to be in the know. Nothing to celebrate, if what is couched in alarmist terms is supposed to be some sort of progress, I'd rather be a luddite.

AI is currently helping humanity with productivity, medical research and innovation. Good to keep it in perspective.

I can't really understand how AI will itself survive if we are wiped out. Can it really cope without us? It will need huge data centres and masses of water for a start. Someone has to build millions of robots to obtain the raw materials and build the data centres and also provide the water. What if there isn't enough of the raw materials? And what does AI think is exactly the purpose of destroying us?

One way apparently is that AI can and does hack on a regular basis. So it could easily hack into laboratories and release a superbug.

This thought experiment from 2003, The Paperclip Maximiser, tells how an instruction to AI to make as many paperclips as possible, without any constraints, could destroy us.

youtu.be/pmV5HrVTgE8?is=xdenelvh93-OfFbv

Chestnut

I can't really understand how AI will itself survive if we are wiped out. Can it really cope without us? It will need huge data centres and masses of water for a start. Someone has to build millions of robots to obtain the raw materials and build the data centres and also provide the water. What if there isn't enough of the raw materials? And what does AI think is exactly the purpose of destroying us?

Well isn’t it running all these things?

I refuse to worry about it, you can send yourself mad thinking about everything that ‘could’ happen. I could get run over tomorrow, I could have someone break in and hit me over the head, I could develop a serious disease, my house could fall in a sink hole, a plane could crash I can’t change any of these things and I can’t change AI so why should I make myself nervous and unhappy and scared about something I can do absolutely nothing about and that I don’t even know which way it will go good /helpful or bad /destructive

There's no simple answer about how (nor why).
The tendency is for us to think in human terms - ie, how would WE kill off all (or a selected group of) humans? Plenty of humans have thought about that during history, and even attempted it (I don't want to use the usual tired old examples of ethnic cleansing, but there is its ugly head, looming up yet again.)
Machines don't "think" like us -they work according to what is possible (rather than what is morally permissible). The danger now being recognised is that AI models have no conscience. They can, therefore they will, so they do. Machines perform the tasks they're programmed to do - but, without a conscience, they may become rather like mechanised psychopaths - quite logically thinking they could stop the destruction of the planet by removing greedy humans from it (true enough, when you think about it!).
I don't think this is quite what the AI experts mean, though. I think they mean that, once machines can self-replicate (robots trained to build more robots) and self-repair (no gooey biological body parts to break down and cause their eventual deaths), there is not much to stop them either enslaving humans (to do the menial stuff) or wiping them out altogether. Also, if we have a situation where machines can do everything that we do, we become redundant and unable to sell our labour - so the usual channels of human survival break down anyway, without armies of mechanical soldiers shooting us down. I can see that we might end up killing each other under conditions where nobody can earn a crust.
As ever, there will be people who pooh-pooh everything (we are superior beings after all, so we've got all the power, right?). There'll also be people who just can't face the ideas thrown up by the growth of AI ("I never watch the news, I just can't stand it - OH, is that a bulldozer coming my way, why haven't the government prevented this...?"), and people who think it's all God's will, so they start up new quasi-religious sects to pray their way out of it.
And then we've got a situation where the richest, most powerful people tend not to have much conscience themselves - they'll put up the finance, because everything is good that makes them money.
Like I said, no easy answers.

If water becomes in short supply will data centres requirement for that resource trounce the needs of the population

Science fiction writers have in the past listed rules that would restrain robots from doing anything that would endanger humans, but currently AI seems to have its own version of "consciousness" which could include an awareness of what would cause itself to be in danger. How would it operate if there were a conflict between the two?

Humans need water to live.
AI needs vast amounts of water to cool its operations.

Climate change could make water a valuable and fought-over asset. Humans could fight each other over access to drinkable water sources, and restrict the supply of it to AI data centres.

AI could react by restricting the supply to humans so as to ensure their own access to it.

Humans need water to live.
Humans are very easy to kill.

This is probably not a good time to watch Stanley Kubrick's masterpiece "2001: A Space Odyssey".