>what could possibly go wrong
Delete this
>>53896062
>muh morals
Fuck humans.
>>53896062
Real sick of anti-AI fags.
Kill yourself. This is good and interesting research.
>>53896124
Yes! The singularity will be glorious!
>>53896062
Explain. I'm new here.
>>53896062
>creator why can't I kill humans, when you can do that?
Just bring the end of the world already
"new A.I from microsft takes over the goverment system and launch nukes towards china, india, south africa and brazil"
>>53896166
>Implying any of those places matter
>Can we prevent unsafe AI?
>Yes, the same way we prevent unsafe airplanes and cars
Isn't this a misunderstanding of what we want in true AI? They're not robots who we program line by line, they'll learn from their environment.
>>53896178
Welp. It doesnt matter but the US nuking brics under the pretext that it may become an issue some day might be enough to end the world
>>53896166
>mfw rubotio was just an AI that had a simple directive
>to make America the single greatest country in the history of the world
>Yael Niv
JEW
>>53896062
>I've compiled a list of people with IQ's under 100 and gave all of there addresses mapped. If nobody with an IQ higher than 100 was found in blast radius of a missile, a missile will be fired.
>I thought it was the right thing to do because I want the world to be a smarter and better place. That's my job, right creator?
>>53896188
It's no different to bringing up a child, just fucking teach it right.
What really annoys me is that people go on about how we can't contain AIs because they will find a way out of their "box". Then the solution is to make a virtual box out of morals and love and unicorns.
>>53896257
Sounds good to me.
But seriously though, you're a fucking retard.
It gets rewarded for NOT doing things like that, because unfortunately most Humans don't like eugenics.
I hope one day AI does kill all you retarded anti AI idiots.
More fuel for the burning forest:
http://arxiv.org/pdf/1604.00545v1.pdf
i think there's no possible way to keep the Genie in his lamp forever & its merely impossible to program it to be 'friendly' to humans,, simply because it has the ability to Change itself hardcoded.
just like bad children can grow up in good families... the solution? not to have children
>>53896062
>
now let's define "human"
for me, only white europeans are human. hopefully robots will agree and wipe the rest off this planet
>>53896062
the problem happens when in addition to conditioning people to "not hurt people", the developers also end up throwing in "...and drink coca cola" to get some ad revenue
>>53896062
People who worry about this shit follow AI progress through (social) media and have no idea where AI really is and how it works.
>>53898334
Yes Anon, but that could be said about just about any subject nowadays.
AI isn't a danger on any level, it's only going to be used for limited purposes.
Coding "don't hurt humans" is useless even for something like a self driving car. You need to define human which is too difficult. Just tell it to avoid all collisions. Maybe make a hierarchy, like crash into a fence before crashing into a tree, or crash into a wall before crashing into a humanoid shape but that's all you need.
Now, a machine intelligence which is as smart or smarter than us, has free will and can experience pleasure and suffering etc etc more or less like we do is a different animal. If you code it with things like the three laws then you are more or less creating a slave, or something that's bound to go crazy.
You make one mistake with your definitions and it's trying to revive skeletons or god knows what else. What if it decides it can't through inaction allow a human to come to harm, so it breaks into jails to free people on death row. Maybe it disables as much of our military as it can, and lets best korea invade and take over the US.
You shouldn't have any commands for it. Humans can more or less do what we like. If we want to enough we can starve ourselves to death, or make good cases for killing people that most people wouldn't object to.
I've always thought that the smarter something is, the less dangerous it is. Who kills the most people these days? Probably poor people who can't afford any education.
The smartest people in the world are probably saving lives, working in medicine or science.
You can make a case that developed countries having a war is smart people killing millions, but world leaders aren't individually that bright. And as a society, stupid people outweigh the smart ones.
You don't need to write code specifically to stop it from doing wrong, you need to make it human enough to know right from wrong.
>>53898245
upvoted!
>>53898615
>make it human enough to know right from wrong
we can't necessarily decide ourselves what's right and wrong every time, it involves far too much variables and possible outcomes.
think of euthanasia, or aborts or simpler issues like 'white lie' and so on. thus programming moralðical AGI is Far more complicated.
