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

True. We’ve had some v near misses too.

@LimeShade
wstch the documentary before calling it a conspiracy theory. I’m not saying that all all. It’s just something that sat with me as strange for a child in a third world country in the early 80s. It’s not passed on rumour. The interviews are on film for anyone to find.

I am utterly terrified by this. Admittedly I am pregnant so hormones don't help, but I'm so afraid for my children and so sad that I can do nothing to protect them from it. They deserve to be able to grow up and live a life. They are innocent and had no say in anything.

The future is so volatile and in the hands of the mega wealthy. I only hope they also have children and it gives them pause.

I don't know why more people aren't completely terrified or why governments seem to be so complacent. I flip between they know something I don't and it'll be ok, or it's a 'don't look up' scenario. I can only imagine the people at the top don't actually want human extinction. Everyone talks about the AI arms race with China, but surely they don't want human extinction too??

@Scoffingbiscuits and @HolidayHideaway ,

I agree re nukes.

If you assign any reasonably probability of them being used in a year and think it has to be binomially distributed, over half a century or a century it becomes a significant probability.

And we have had some very near misses. (Details below courtesy of AI, it does have its uses). From memory, the 1983 one was literally one man’s instantaneous decision away from Armageddon.

The Cuban Missile Crisis (1962): During a blockade, US warships dropped depth charges on a Soviet submarine. Unaware they were practice charges, the sub's commander wanted to launch a nuclear torpedo. Second-in-command Vasili Arkhipov refused to give his required consent, saving the world from nuclear retaliation. 1]
The Soviet Early Warning False Alarm (1983): Lieutenant Colonel Stanislav Petrov was on duty when Soviet satellites reported a US intercontinental ballistic missile strike. Suspecting a computer error rather than a real attack, he disobeyed protocol and reported it as a false alarm, which it turned out to be (triggered by sunlight glinting off clouds). 1]
The 1980 Faulty Computer Chip: A defective computer chip in a US NORAD communication system falsely signaled that hundreds, and then thousands, of Soviet missiles were incoming. Strategic bombers scrambled and missile crews prepped for war before the system was identified as a glitch. 1]

If the warnings are true, then AI agents are already capable of doing potentially destructive/harmful things independently. The main fear of them is that they will end up "out of control". If that is a legitimate possibility then "China will beat the US" or "the US will beat China" is a moot point. It's the same problem as war elephants, great when they are under control attacking the other side as technology, algorithms, hacking etc. But deliberately "developing" an out of control elephant that's going to charge about all over the place isn't in anyone's best interest. It's not like China winning that race is going to hurt us more than America winning that race.

Really I think that the "AI will develop consciousness and decide to kill all humans" is a scare story spread to make AI companies look better (!?) But deliberately trying to develop agents that act independently including in malicious ways in order to "beat China" isnt a great plan.

AI kind of is humans killing all humans, because if AI kills us its because humans developed it and didn't stop it

No that’s a terrible plan in that last line but idk what they’re developing specifically. Just that there seems to be competition between the two.

Although China seems to be doing it cheaper/more efficiently. And America who have about 10 times the database capability (in terms of area covered) are not surprisingly measuring their success by this metric. But I think it's set up a mindset that the more database, the better. When in reality if a LLM can solve a maths puzzle a SLM trained on more tailored datasets could solve the same maths problem far more efficiently. And be less likely to randomly hack into a different companies website.

At the moment it feels like that joke about rich Russians after the fall of communism. One asks the other how much his watch cost and he tells him 10,000 dollars. You fool says the other, I bought the same watch for 15,000 dollars.

If you think the world is safely in the hands of the grown-ups, it's very clear that you're wrong.
I read something from a woman who used to work for Trump (in his first term). Apparently he could not get his head round the idea that launching nuclear missiles was a bad idea. He kept on coming back to the nuclear missiles, and why people thought he shouldn't use them. He couldn't remember or understand the reason, and seemed to have a great interest in pushing the red button. It's not surprising really - for someone who lives to be powerful, that must be the greatest high.

It's very hard to get our heads round the idea not only that humanity (or much of humanity) will end (eg due to nuclear war, biological or chemical attacks, global warming, etc), but that it is likely to end very soon and that quite probably we, and very probably our children, will experience it.

I suspect we're nearing the stage when everything changes very quickly and drastically, and we accept that there's no point reading science fiction or dystopian fiction or even end of the world fiction, because we're living it.

I think we’ve pretty much always thought we’re on brink of Armageddon…NB: 60s/80s.

Has anyone read Annie Jacobsen - Nuclear War? & now, Biological War?

Unfortunately a lot of top AI bods have stated they think it’s not a bad thing if AI succeeds humans. I think the figure was about 15%.

Don’t underestimate the craziness of some of these super intelligent people

Isn’t it more likely eventually we merge with AI rather than they wipe us out?

Now I know why David Lynch was so obsessed with electricity and faceless forces in his movies and TV shows.

Wake up.

YANBU

I also don’t understand how we are just looking no at AI taking all our jobs. I am due to lose my fairly high level job this year to AI. I am 55 and the government tells me I have to work for decades more. So what’s the actual plan then?

It's terrifying, and has been for some time. We are already run by robots. You can't pick up the phone and talk to anyone now, but have to go through a serries of ridiculous Q&A sessions with a bot. Millions are out of work because AI has taken over. The world can grind to a total halt, and has done so many times over the past couple of years (think worldwide aviation stopped, inability to actually go shopping and pay for you goods!) due to automation.

The IQ of the young in this country has dipped significantly. Kids are being taught through AI and IT. They don't need to research, read, write, and can use AI to pen letters.

As for the Arts - I'm hoping that there is a bit of a come back on this one, where composers, artists, photographers, etc are fighting back. The whole 'compose/create art in the style of ....' is beginning to be held up for the bollox it is. However, it is inconsceivable that anyone can have to audacity to think it possible to recreate the development and imagination of art.

Robots will never have creative imagination.

I’m a similar age and there’s going to be a huge employment crisis before long imo, given that ageism’s still so rife in the workplace. Personally I think some kind of universal basic income conversation will have to be had sooner or later, I can’t see how else we will get around the effects of AI.

Can someone explain how this income conversion will work?

Most people want to work, be useful, contribute and socialise with other imams through work.

Well, let’s think. Humans create an alien intelligence far greater than our own. AI will regard us as an inferior species and either kill us or subjugate us. How dumb can we get?

Dumb or greedy. AI is making a few people
very rich.

AI was officially named in 1956 not human like we are today but the concept of AI was there and being developed computer programs skill solving .. its absolutely terrifying sooner or later they will outperform humans.

It's like these bloody AI programmers never saw Terminator.

I think it’s more a case of the stupid humans will program the inanimate AI to do stupid things.
It’s the humans we need to watch out for.

Not sure if it’s already been shared or if it will achieve anything but I’ve signed this petition

I’m not specifically pro a UBI, just can’t see how it can be avoided if AI eventually removes many people’s sources of income and they can’t move into something AI-proof. I agree work is important for the reasons you state here, but there are other ways to find purpose in life. I guess a lot of our current beliefs would have to be re-examined.