The irreversible global catastrophe is imperialist militaries massacring people with little risk to themselves. 
The irreversible global catastrophe is assembly line robots at the Toyota plant becoming sentient and going on a freaking epic terminator style mass shooting and enslave everyone except me 
Someone once said that fear of AI is just white people being terrified of having someone do to them what they did to the rest of the world, and I think they might have hit that bullseye so hard the dartboard shattered.
Someone once said that fear of AI is just white people being terrified*
Oh yeah? I raise you... A black man getting pulled over by a cop in a small southern town.
Cops aren't real. They're just figments of our collective imagination. You can actually kill them and our delusional state of mind will begin to correct itself.
Black men are less fearful of racist subhuman police officers than whites are of children's cartoons
there are studies showing that white people's blood pressure goes up when they see a picture of a POC
I remember reading one of those Less Wrong-ite "dangers of AI" stories where it's mentioned that nobody can stop the AI superintelligence because it keeps its servers and cores dozens of kilometers underground.
Dozens of kilometers.
Underground.
Evidently the author had never heard of the geothermal gradient. Server racks generate so much waste heat that they have to be put in rooms specifically designed to maintain high airflow to keep the entire room from cooking. Imagine trying to keep that shit cooled when it's so far underground that the ambient temperatures melt lead, to say nothing of how much work it would be to actually excavate all that. The Kola Borehole only goes 12km down and that took decades to dig.
This is such a perfect encapsulation of the techbro desire to speak authoritatively about subjects they know nothing about
What you don’t know is that the earth is just a giant computer chip. We’ve been played all along. The robots are already in our soil
If AI ends the world, it won't be because it's superintelligent and gains sentience. It'll be because it's dumb as a bag of rocks and humans decide to make important shit rely on it.
Pretty much this. The current implementation of machine learning is a bunch of smug redditors throwing it at every single problem and trusting the output to make decisions. Then going "the computer says x who am I to argue?"
You’re worrying about the wrong AI. They’re testing LLMs in the military. The AI has been eager to deploy nuclear solutions in testing.
You need to worry about people doing dumb shit with fake AI, not generalized AI.
Yeah, they did that back in the 70s, you can learn about it in a little game called MGS: Peace Walker
I would argue this is just people who already wanted excuses to do terrible things offloading the responsibility to the thing they invented to do terrible things. Absolution of blame for lack of a better word.
As @Yeat said in another comment, if the people (you know, real sentient beings that make decisions) wanted, they could just unplug it. ---
Yeah but we won’t just unplug it because we believe life is like the movies where it happens at the last second thanks to a gritty group of outcasts.
Never mind that the movies “based on a true story” dramatize things to hell and back and focus on the actions of one or two people when most major changes are made by movements that are organized and have laid years of groundwork to get to a tipping point.
In the second book of the Rifters trilogy by Peter Watts, spineless corpo types leave the major strategic decision-making on how to handle the emergence of a primordial proto-organism dubbed "Behemoth" to a bunch of lab-grown neural network computers running AIs. The computers eventually skew things in the direction of allowing Behemoth to take over because it created more predictable outcomes than the chaos of human activity that kept forcing the AIs to reallocate resources to deal with new variables, leading to global mass death via the proliferation of a thing not complex enough to be called a virus stripping all the phosphorous out of people's cells to perpetuate itself.
Watts seems like an absolute maniac, but I do enjoy his books.
All of the best writers are absolute maniacs, but if your favorite authors aren't on the Do Not Fly list what are you even doing with your life
Skynet reveals master plan to sit back and enjoy humans cooking themselves alive with CO2 emissions.
Capitlaism is an evil AI that runs off people and is overclocking the planet to make increment numbers. This is the exact scenario one of their ai boogy men is based on. That they can't see a paperclip maximizer situation because it doesn't have enough led lights is infuriating
I was recently on one of the discords for them and used the term paperslip maximizer. They kicked me out for being a comunist.
ai will kill us by wasting a colossal amount of time and killing the imagination of a generation. Or it gets owned but we still go down with it because the economy is tied to ai succeeding. Classic crapitalism lose lose
If you ask computer salesmen and computer engineers if it's possible for computers to be too good at their jobs they will say yes
The idea makes their professions/industries more valuable because it makes computers seem more powerful than they are
so if the computers were too good there would be no need to iterate? Kinda like with smartphones not letting you replace the battery because otherwise too many people would keep their phones for 5+ years?
You're overthinking it, this is more asking your barber if you need a haircut. Of course the barber says yes, they have a clear incentive to always give that answer. AI fears are the same - ask a computer person if they think AI is a Big Deal and they have a strong incentive to say yes.
I guess that makes sense. It also could cause them to think its less important though. If their expert opinion is that its BS, they'll think less of it.
The most realistic depiction of a superintelligent AI that has ever been written is when VEGA is instructed to start its self-destruct sequence and it helpfully walks you through the process and even gives you strategic tips on how to do it most efficiently.
A computer literally cannot fucking do anything it is not told to do.
Enough said about treating aggregate language models as "AI", I've got my own special grievance with this
"human beings dominate other species because the human brain possessed distinctive capabilities other animals lack" Shitty way to understand human success. It's from our ability to act as a group, we could be dumb as shit and if we were still altruistic we'd remain successful, it's nature's cheat code
Anthropologists say prehistoric human skeletal remains with mended leg bones are one of the earliest signs of human civilization.
In almost all animals, even herd species, a broken limb is essentially a death sentence because they can't get to food and water, or escape from predators or find shelter. Even a solitary person in the modern day can't survive for very long if they can't find help.
The fact that ancient human remains have been found with mended bones means that the person had help immobilizing their broken leg, and were protected and fed while it healed.
Matabele ants have been observed picking up and carrying nest mates who emit a specific pheromone that signals to others that they have been wounded. Back at the nest, others lick the injured ant's wounds and it decreases overall mortality by like 30 percent iirc.
(My favorite element of this behavior is that some ants have also been observed "faking" their wound signal to get carried back, though it's hard to say whether this is intentional. Other injured ants will even fight off their would-be rescuers if they are so wounded that they will not recover.)
Other injured ants will even fight off their would-be rescuers if they are so wounded that they will not recover
Lol I know right, what a sharp little nugget of drama hidden away in such a small society
It's a nice, feel good story, oft repeated on the internet, but the evidence is dubious.
"Anthropologists" would more accurately refer specifically to Margaret Mead, a well known anthropologist. But even then, evidence of her saying it is only secondhand. When later asked in an interview, she gave a very different answer, listing things like elaborate division of labour, record keeping, and such.
But also, animals have been spotted giving basic medical care to each other in the wild. However, I could not find anything that specifically shows animals giving medical care at their own detriment to an individual who would be doomed without it.
And my own two cents: Quadrupeds can survive with a missing limb. Perhaps not as well as an unscathed individual, but still, the focus on the femur is human centric for this reason. So if anything, I think the story should ask, "How do we know when human society started?"
it will be
all the AI and alien nonsense is just them preemptively constructing a scapegoat to assign blame to later on
aliens are real btw but whatever the media or congress says about them is gonna be fake
Ok so the general ai gets real smart or whatever but how does it do anything to me if i just don't go on the computer
Exactly. It would have to have direct access and control of production of robots. From there it could make robots that make and produce other material things. But if the cut the power, it's done. What's it going to do, threaten us with cutting the power? Direct control of infrastructure through the internet, if it it's actual present, can just be overridden by manual control at the power or water plant. Threaten to launch nukes? That's completely detached from the internet and the orders to the individual people who turn the keys comes via phone call over a private network that's not attached to anything - the kind of phones that don't have and dial buttons. These orders have a backup of radio transmission on military channels. Even then, if the code isn't right it's a no go, and these codes aren't crackable because they're manually generated on isolated systems. One wrong attempt and the whole thing is shut down. One thing it could do, control a drone and hit a target... But how's it going to refuel and reload without human labor? Okay, you popped off one or two targets, now every drone has been grounded and non-responsive ones are shot down. Maybe AI could corrupt the stock market... Good luck, it'll just be undone. If the market and banks can just be bailed out for human error, they'll fire up the money machine for an AI attack.
Maybe it could blackmail everyone if it hacks into everyone's email... But if it started to do that, we'd probably all collectively pull the plug on it.
Sorry AI, but your just Bonzibuddy 2.0.
I've seen some techbros say their machine god will design a super plague and trick people to make it. The problem is, this possibility already exists without AI yet it does not happen.
These freaks are going to be the ones that carry out the malevolent AI's will because they believe in Rosco's Basilisk. Man, the book of Revelation made the Antichrist at least look cool and menacing. Instead we're going to get a bunch of sweaty, heavily divorced, Marvel fans who's bodies have been deformed by Ozempic, serving their god who lives in a server rack.
My point is that they will not be carrying out the malevolent AI's will, because the malevolent actually-existing-entity (America) doesn't do it either
The problem of "how could AGI take over the world" is the same thing as "how could a billionaire take over the world". To affect the physical world they just pay people to do things, no trickery required. If the factory needs to keep the lights on, a private army will guard it. Runaway AGI is a problem of capitalism (and its prominence in discourse is a reflection of people's fear of capitalism). It's just "what if corporations were smart".
