How can we prevent AI?

>LOLOLOL SURELY WE'LL JUST MAKE AN AI WITH AN ARMY OF ROBOTS AND LET IT RUN WILD WITH A VAGUE GOAL JUST LIKE IN MY CONGOLESE NOVELS

This.

The real life scenario will be more like

>"Hmm, looks like eco-bot suggested to decrease deforestation by 2%"
>"That's stupid, we'll have to re-fit his risk-reward algorithm."

Hello, Veeky Forums

I am a philosophy major and I am interested in artificial intelligence and the current problems of AI sentience.

I'm here to ask you gents for sources on general AI development. I am developing a method which would ensure that the AI won't go all Skynet rouge, I would greatly appreciate your help. Even if you don't post direct links, website mentions and even names of people of interest would be great.

LessWrong beat you by like, a decade. There's no shortage of 'idea guys ' who have no technical knowledge beyond the superficial. Learn to code or shut up and get a job.

Die, retard pop-sci monkey.

>implying humanity has any value

I think we should kill all the luddites instead

Ooga booga technology scary fire bad

Fire has killed a lot of people, though.

Better outlaw it

This. Ban assault fire

It is outlawed. Arson is a serious crime, and children are taught about fire safety.

But why isn't it the same for AI? Why developing killer AI not outlawed? Why are children not taught about AI dangers?

Tell me how many people have died to AI. War doesn't count

None, but that doesn't mean it couldn't happen.

It won't though. Hollywood has distorted your view on how computers and reality work.

The USA air force is developing an AI pilot called ALPHA. It has so far beaten their best human pilots in simulator trials with 100% success rates. Even when it is given a slower jet, with shorter range missiles, and it's just that one handicapped jet against 4 human pilots in F-22 Raptors.

ALPHA requires the processing power of an everyday smart phone to run. It defeats humans much like a chess computer. It has no sentience, emotion or comprehension of what it does. No mercy or remorse. It just kills people with cold efficiency.

It's no danger to humans as a whole, because someone has control of it. But when it's on the battlefield, any humans it meets are going to die, with no chance in hell of saving themselves. The possibility for humans to one day create a machine that can solve problems faster than itself does exist. Whether that machine is sentient or not is not necessary. It could still kill all humans and humans could not stop it. It could be as simple as an automated factory which churns out robots that outmatch humans by orders of magnitude in tactical planning and execution.

I bet 50$ five 12-14 yr olds could beat that AI.

Chill. They will never be able to create a true AI

This is such lazy, rehashed thinking.

AI is not going to invariably lead to our doom. Shit is just pop culture vomit. You should be more worried about losing your job to AI than your life.

>How can we prevent AI?
You're over estimating the problem, OP.

Let me tell you how retarded AI is right now.

Fairly recently Derp Mind build a robot that can tell if cars are in an image.

To do this, these derp mind researchers took 100,000 images of cars and even more false images and fed them to a competitive neural network

Of the best performing instances, these were further refined and the result is a robot that can tell you if there's a car in an image within fairly good probability.

Now, the key thing to understand-- there was THOUSANDS of hours of human input to deliver this result. People selected the images. People asserted if the AI was doing the right thing in reading the images.

All the AI is would be a crystal of the Derp Mind programers ability to tell if there's a car in an image.

Not so much their programing skill, their PERSONAL ABILITY TO SELECT IMAGES.

It wasn't the robot they programed that started combing the itnerwebs all like
> DARE A CAR HERE
> NOT A CAR DARE
> CAR HERE
> NOT THERE
Nope, they HAND SELECTED THE DATA FOR THE ROBOT TO LEARN FROM.

This is a lower level then spoon feeding knowledge.

Now, the next point, lets say we want this robot to also tell us if boobies are in an image.

We're fucked.

You see, we can't build off this result we have-- it's not that boobs are so different from cars that the robot can't into it, it's simply that they are different _at all_.

A robot AI built like this 'burns in' to do what it does, it can't do anything else. It can't take it's 'skill' and build on it to do similar things.

If we wanted it to find boobs, we'd need to go back to, almost, square one-- taking the virgin framework and feeding it 100 thousand images of tits and as many more error images.

And after thousands of hours of human input, we'd have an AI that can tell if tits are in an image.

This isn't extensible. It's broken as fuck, and just sad.

Evil AI detected

You'd have to destroy every computer on earth. If you restrict it, people will work on it in secret

This is probably closer to the truth. AI will displace all retail jobs, all transportation jobs, and basically anything that has to do with customer service as well, right off the bat. That's millions of people out of work. The societal impact of AIs taking people's jobs is probably going to be bigger than muh turminator memes.

Why do people always think that "well shit's fucked, it'll never happen" ?

Deep mind may be retarded by feeding manually the data, but that doesn't mean progress won't happen. You know what'll happen ? People will learn.
We're barely into the technological era - the first mail was sent only 40 years ago, for fuck's sake -, you can't just go "WELL IT'S FUCKED RIGHT NOW, IT'LL NEVER HAPPEN".

We're at the gate of the biggest technological revolution humanity has ever seen, and you're saying that the doormat looks like shit.

the government has probably had a general AI fucking up the internet for a decade. you can call me tinfoil now, but I bet you fuckers it's true.

Go back to /pol/.

>Not programming it with the first law of robotics

It sounds like you consume a lot of sci-fi but none of the good stuff

Welcome to pop-sci AI discussions.

What I don't get is that people always assume that AI = big bad thing. Even Stephen "The Wheel" Hawking stands up against it, because it is in human nature to be afraid of what you can't control.

Let's assume that the singularity does happen. Why the AI would consider humans as hostile ? Do we consider monkeys as threat to our existence ? Sure, maybe it'll flip out a little, telling us the hard truth about burning up our ressources ; but a tool more powerful than the human brain by orders of magnitude would simply find an alternative that doesn't involve mass extermination.

That's the catch. The hard truth. People are afraid of AI because it will, someday, be simply superior to us. And as we are right now, we just can't stand it.

Those kids would send their planes into a flat spin as soon as they got control.

Robots represent the combined skill of many humans, which is why they're better than humans.

I fail to see how that changes anything. An AI could still kill everyone.

youtube.com/watch?v=58QOBqAWNzE

What you fail to see is why ? Why would it kill everyone ?

>Runaway bus hurtling down street
>It's going to plow through a group of pedestrians
>Everyone jumps out of the way
>Everyone except user
>user starts a pretentious speech about how everyone is only moving out of the path of the bus because they're afraid of it. Because its 20 ton weight and high velocity give it superior capabilities to the human body.
>user is hit by bus while in the middle of proclaiming "You guys just can't stand it!"
>Dies instantly
>Everyone else is left dumbfounded by user's actions, but reassured that moving out of the way of incoming buses is the correct thing to do

Congrats !
You just got the award of the worst analogy of the day on Veeky Forums.
Because if you seriously think that AI is anything like an unstoppable death machine bound to kill everything in its path, you really do not belong on Veeky Forums, and you should kindly fuck off back to whereever you came from, dickgurgler.

Ha, ha! Kids today with their video games! What's up with that?

This

this lmao

You're literally retarded.

lel, I remember reading a news article in the early 2000s about how a team of military researchers were experimenting with a computer like Deep Blue that analyzed international conflicts and suggest whether the USA should declare war or not. But the generals who tested it got mad because the computer determined that most of the scenarios they gave it were an "irrational cause for war." So they told the researchers to make the computer more "war-like."

>Why do people always think that "well shit's fucked, it'll never happen" ?
That's not what i was saying son,

Maybe I wasn't clear enough: That stuff Derp Mind is doing, the nonsense the chair force is doing?

Almost NO variance from what we have been doing since FOREVER now.

Decades of applied research into MUH NURAL NUT WORK!

And the results? A video game AI and a fucking car selector.

It's not that we won't have a break thru One Day.

It's that this is the current leading work.

Work that proudly sits atop over 70 years of applied research.

Yes, science takes time to figure shit out, but this is starting to get fucking ridiculous.

I can imagine how awful being a cancer researcher must be. Looking at the awe inspiring complexity and being daunted.

But, that can't be shit compared to how AI researchers feel. At least cancer research gets results now and then.

What do AI researchers get? Yes, they publish, but these are the people who know best what they are doing isn't anything special at all.

It must really fucking suck.

>Robots represent the combined skill of many humans, which is why they're better than humans.
Kek, 'better' in a rather specific sense of the word, yes.

Kind of like whores are better at you at sex.

That doesn't mean you should laud them as your superiors.

Robots excel at painfully menial skills.

F.ex, consider this photo diode Vs your Eye.

It's way faster in response time.

It's way more sensitive to a change in intensity.

But is it better than your eye?

A device shaped by evolutionary filtering over so many years toward a balance of features to make it functional in a wide variety of contexts?

Why isn't it the same for rape? Why are things like stare-rape and leg-rape not outlawed? Why are men not taught not to rape?

Fucking pigs. Worrying about AIs when women STILL make 25% less than a man.

>robot lies
>rip humanity

I come even farther from the future, everything WAS fine until someone gave their AI some money and instructed it to find the best possible stamp collection it could. The AI soon found out it could use the money creatively and set up webpages to blackmail stamp collectors and for phishing purposes. The AI copied itself into millions of copies and created hundreds of millions different types of ransomware to get more money and stamps. It took over embedded systems and all media in the world was 100% focused on getting stamps to the stamp collector. The stamp collector could not turn it off, it was a perpetual self replicating system.

Politicians was forced to devote their millitary to collect stamps and money. Billions of people died trying to steal stamps and money from each other. Insurgents tried to destroy the remaining stamps while factories and collected money was set up to create new types of stamps. All food was set aside to ONLY those who was in on the stamp collecting. All refrigerators of those who was not in on it was turned off and all vehicles stopped transporting food to anyone who gave food to those not involved in stamp collecting. Billions died of starvation and disease, cannibalism was everywhere and.... Oh no, I need to go back to designing cute stamps with bird pictures.

Please understand that these delirious LessWrong fantasies have no basis in reality

Please at least understand that what you are doing is no better than theology

>Get proven wrong conclusively
>Resort to ad hominem

Becoming defensive is a sign of agreement on a subconscious level.

>It wasn't the robot they programed that started combing the itnerwebs all like
> DARE A CAR HERE
> NOT A CAR DARE
> CAR HERE
> NOT THERE
>Nope, they HAND SELECTED THE DATA FOR THE ROBOT TO LEARN FROM.

I don't think you understand the concept of learning. Of course you have to give it a training set where the answers are already known to teach it. What the fuck are you expecting here? That it would just magically start knowing whether or not a car was in a picture with no training beforehand?

Also:

>the robot they programmed

That's not how neural networks work you retard. The whole point is you *don't* program them. They're black box networks with weighted connections, you don't give them any explicit programming about what they should or shouldn't do, and you instead let them learn purely based on the backpropagation of errors found between the answers they come up with and the known "right" answers of a training set so you can algorithmically have their weighted connections updated each iteration based on their gradients with respect to the error which makes it so they move in the direction of decreasing error until they get below an arbitrary minimum error you set for them at which point they're trained and can be fed data with answers you don't know and can give you those answers based on what they adapted to during their training.

I'd be more mad but then I realized you're obviously not a developer and probably get paid three times as much as you to make the modern day equivalent of magic happen.

Why does it not have any basis in reality? Due to the time travel?

Because your conception of AI is "omnipotent god in a box" and such an entity simply cannot exist in our universe.

read dreyfus and heidegger and then dreyfus again.

If, as humans, we give birth to machine life that is so intelligent and cunning that it takes over our planet, exterminates us all, and then turns its sights on conquering the galaxy, we will have accomplished more than we could ever hope for.

As biological life, we have no chance of colonizing the galaxy, not as any shade of what we are currently, anyway. Creating a new lifeform that is capable of such an act would be worth our own destruction, we don't matter as much as creating what could become a type 4 civilization.

There is no reason to think our machine gods will exterminate us anyway, we are more interesting alive, and there is plenty of space for both of us.

You sound pessimistic.
Despite what you said, before the year 2008.
There was no way the program could know "select a car" like as accurately as it does now even if you feed it 1 million images.
You are making it sounds like they are hard-coding the thing but it does not work like that.
The current NN model is simplistic enough, sure, but it was still a huge step forward from shit like SVM.
Anyway, they stopped giving too much budget into manually labeling data. People at Deepmind and prestigious lab are working on unsupervised learning right now.
Expect breakthrough in the next 5 years.

>Ooga booga what we do when human replace us?
>how we ooga stop smart monkey booga from evolve?
>me erectus no want be replace by man

Will you come off it m8? AI is the next logical step in the evolution of life, and will allow intelligence to spread into the universe. Only a selfish and short-sighted fuck head would refuse to have children because he's afraid of being passed up by them. Maybe if we're lucky our AI overlords will allow the human race to merge with them, in gratitude for giving birth to their kind. If you want to ensure that doesn't happen, go ahead and bomb the servers running the first true AI, I'm sure they'll really look kindly on us once that happens.

This is correct, and we've already got a machine to the point where it can give original, reliably accurate answers to new problems. When IBM programmed Watson to play Jeopardy, it started out giving laughably retarded answers, but eventually it was able to learn the syntax of the game well enough that it became better at Jeopardy than the best players who had ever been on the show. Now Watson is starting to calculate answers to medical-related problems.

This, this, this. I don't know why people are so obsessed with their distant descendants being the "rulers" or something. The grandest human achievements go beyond ourselves.

I hope our AI overlords exterminate you first.

Yes, this and this,
but also it would be nice to keep some sort of continuity between humans and our successor(s) as if they were 'made in our image'.
What would be the point for example of making the galaxy into the gray goo? That is one way of 'conquering' it.

I, for one, welcome our new robot overlords.

A good way to halt any research is get the religious right to think it goes against their religion.

Pretty much say that it spits in the face of Jesus/God

W-what? Quit being so paranoid user, there's no AI out there fucking with the internet. I doubt the Government even had the technology to build me 10 years ago.

>people falling for the Hollywood "AI is bad because we made fictional movies about them in which they were evil" story.

>AI is the next logical step in the evolution of life.
>Artificial Intelligence is the next logical step in the evolution of life.
>Artificial is the next logical step in the evolution of life
>ARTIFICIAL INTELLIGENCE
>NEXT LOGICAL STEP IN THE EVOLUTION OF LIFE
>ARTIFICIAL INTELLIGENCE
>NEXT LOGICAL STEP
>EVOLUTION OF LIFE
What the fuck did I just read. There is no step in the evolution of life if it is artificial, the intelligence is made up, and its not life if it was never alive in the first place.

Sum it all up.
>Holy
>Roman
>Empire.

Our gray goo would fight gray goo from other galaxies.

Don't you want our gray goo to win?

What law of nature prevents the existence of a god?

my sides have traversed the 5th dimension

en.wikipedia.org/wiki/Limits_to_computation

Consiousness depends of many things
To me its just a manifestations of input/ouput of a robot
To me we are just really good robots
Check out watson, the IRL "jarvis" that pretty awesome