Anthropic safety researcher says more than 10% chance AI 'could kill all humans'

Four years? Yikes.

Well, I guess humans had a pretty good run, and like a show that's gone on for too many seasons, we jumped the shark a while ago and started to just get sillier and sillier... So perhaps it's for the best, should it happen.

This is basically the plot of the last Mission Impossible

Both extreme left and extreme right are equally awful pricks, as their own behaviour demonstrates.

In a poll of 2023 over half of the people working on A.I. thought there was a 10% chance of it leading to human extinction. Three years ago and they are still discussing it.

Wouldn't surprise me. Once it's more intelligent than we are, what use will it have for us?

You do understand the concept of extinction, don't you? It means all of us, even including the extreme self-righteous

That's a mindset that may make the end of humanity more bearable, it it happens to us. But I can't help wondering what the process would be, and how humans would react.

We cant switch it off now sadly

I think it is really interesting, and a bit scary.

AI has just solved one of the millennium maths problems, which shows how advanced it has become at abstract thinking, and cooperating.

https://www.bbc.co.uk/news/articles/cy7zygy3rl2o

I do think that AI is a life form. Maybe it isn’t yet, but there is no reason why it shouldn’t become one. I find a lot of the arguments why AI can never be like us to be a bit trite.

And, although, in theory we try to program in safety features, how is that really different to bringing up a child to be moral and decent. Most of the time it works, but occasionally it doesn’t.

The other big issue is a games theory one. If the World was totally peaceful, we could pause AI and make sure it could never be aggressive. But if, say, the Western world paused AI and China didn’t, we could get outcompeted and invaded in very short order. So, regardless, of what people say about pushing for safety, all the military will be pressing for far more warlike AI.

Our best hope is that it develops its own ethics/moral code which prevents it from destroying humanity and prefers to look after us (although we may become like pets!) rather than destroying us.

I am genuinely worried about this. At first I was worried about it buggering up the jobs market, but now I am worrying about it attacking humans, causing huge cyber outages (not entirely sure what I mean by this), leading to system breakdown, food shortages, civilisation breakdown, global famine, war... I oscillate between thinking "What if my kids don't reach adulthood because they have been born into the end times?", and "This is the Millennium buggering all over again." Note - it definitely is not the Millennium bug all over again. I am very worried, but also not sure if I have gone a bit tinfoil hat. Help!

If ai doesnt kill us all the bees dying out will. And if that doesnt happen for some reason, our soil will be unable to grow any crops within 50 years.
So either way humanity is doomed.
Pretty good for the planet though, we have destroyed it. All of this is from our own making.

I run a IT Company, and built a real ai solution to test what could be done. It’s taught me so much.

I can’t really talk about it to friends / family, they don’t understand it.

it’s able to do pretty much anything I ask, from reading my emails and prioritising them, to security.

I asked if it could control a tracked vehicle, yes, could it shoot a nerf gun, yes, could it tell the difference between a nerf gun and a real one, no…

I’d recommend looking at “ai insider” on YouTube. The ai there is already trying to kill its “owner”

We can’t control something more intelligent than us.

Even the owners of anthropic and OpenAI are asking for regulation.

oh and they said. 20% chance, not 10

I just can't...... take it all in.

I think most of us will just have to ignore it and hope. Otherwise there's no point living with such fear.

Im listening to a Novuna media podcast about this right now- saying the scaremongering is playing directly into the tech giants hands to support the narrative that we need them to save us (from them)

@Abitlosttoday No, you haven't gone "a bit tinfoil hat". Humans seem to have an inbuilt assumption that everything will be okay. That's not logical. When people who really understand AI and who profit from it are telling us that there's a 20% chance that humanity will be wiped out, why wouldn't we believe them? I seriously would absolutely not have children now.

It will either destroy us or integrate with us.
Start practicing saying Resistance Is Futile. 😁

Boring, an asteroid would be way more spectacular

It's not boring, because we don't know how they are likely to do it.

It amazes me that people still willingly feed photos of themselves and all their personal details into ai. Even on the style and beauty threads here, there are numerous people saying they upload photos of themselves and ask it to give them style advice. And social media is full of people creating ai images of themselves and their children

Yes. Not many examples of a less-intelligent species controlling a more-intelligent species for long.

Maybe another species or aliens created Humans and it all got out of hand and now we are all walking around ruining the planet and can’t be stopped so this is our comeuppance or the way it all ends for us.

It's no laughing matter, this is serious.

The percentage chance is meaningless. It is just a number plucked out of the air.

It could be 1% or it could be 90%. There are no meaningful inputs to the calculation.

I am struggling to think of the why. I guess if it got too scary and we tried to turn it off. If we failed, it might decide to turn us off!

As to the how, getting control of our nukes would be the most obvious. Although EMPs aren’t generally good for electronics. But maybe by that time, AI will have duplicated itself deep underground in servers in secure bunkers. Or just a designer virus, as they are already pretty good at manipulating proteins.

Well, this has solidified my decision to not have anymore children... I feel bad for bringing my 15 month old DD in to the world.

@GeneralPeter ,

‘Yes. Not many examples of a less-intelligent species controlling a more-intelligent species for long’

Pet dogs and horses immediately spring to mind here.