AI is highly likely to destroy humans, says Elon Musk

[Pakistanian]

http://www.independent.co.uk/life-s...nce-openai-neuralink-ai-warning-a8074821.html

“Maybe there's a five to 10 percent chance of success [of making AI safe],” he told Neuralink staff after showing them a documentary on AI, reports Rolling Stone.

He also told them that he invested in DeepMind in order to keep an eye on Google’s development of AI.

Mr Musk has called for the companies working on AI to slow down to ensure they don’t unintentionally build something unsafe.

Gadgets and tech news in pictures
+
show all
“Between Facebook, Google and Amazon – and arguably Apple, but they seem to care about privacy – they have more information about you than you can remember,” he told Rolling Stone.

“There's a lot of risk in concentration of power. So if AGI [artificial general intelligence] represents an extreme level of power, should that be controlled by a few people at Google with no oversight?”

Though he didn't expand on what sort of threat it could pose, he's previously said that AI is “a fundamental risk” to the existence of human civilisation.




READ MORE
Stephen Hawking warns AI 'may replace humans altogether'
He believes its development needs to be regulated “proactively”.

“I have exposure to the most cutting-edge AI and I think people should be really concerned about it,” he said in July.

“I keep sounding the alarm bell but until people see robots going down the street killing people, they don’t know how to react because it seems so ethereal."

 

[RickyG]

Imagine Dajjal is a robot all along.

 

[Nostalgic]

Dajjal Robotics Inc would be a great name for a startup.

 

[Pakistanian]

Imagine Dajjal is a robot all along.

Dajjal imo is more of a metaphor, it could be in the shape of system like technology or could allude the state of the world we live in.

 

[Suleiman]

Dajjal will be known as the Terminator in the west.

 

[pillionrider]

I hope we are all able to settle on which religion AI follows before we are destroyed. At least we will know which of us was right all along

 

[Syed1]

AI technology should have a kill-switch if things get too out of hand (which they will).


Example of AI technology learning and improving itself:


There is a stretch of highway in California on which human drivers don't really struggle since they can anticipate the turn and angle of highway and when to brake and accelerate, and at what angle to steer. When Teslas in autopilot mode first started driving on that stretch they struggled a great deal, once or twice went out of lane and alerted the driver to immediately take control. However, over-time when more and more Teslas drove on that stretch and they each shared their data with the other through machine learning the car learned how to navigate that stretch perfectly. All of this was achieved with zero human input. It is a little scary what machines can do if you teach them how to learn.

 

[BlackShadow]

Don't know about destroying life, but AI gives me hope of finding an algorithm defined partner. This has been a fantasy since Blade Runner.

 

[Syed1]

Imagine the superiority of the human brain, we can drive a car while day dreaming and thinking about the wonderful nihari that is waiting for us at home (or being on the phone - which is a garbage habit to have) and still drive safely. Meanwhile a car needs about 200 sensors and several cameras, radars, lidars etc streaming a tera-byte of data per second just to keep the car in between two lines. Machines have a long way to go before they can eclipse humans.

 

[Abdul]

AI technology should have a kill-switch if things get too out of hand (which they will).


Example of AI technology learning and improving itself:


There is a stretch of highway in California on which human drivers don't really struggle since they can anticipate the turn and angle of highway and when to brake and accelerate, and at what angle to steer. When Teslas in autopilot mode first started driving on that stretch they struggled a great deal, once or twice went out of lane and alerted the driver to immediately take control. However, over-time when more and more Teslas drove on that stretch and they each shared their data with the other through machine learning the car learned how to navigate that stretch perfectly. All of this was achieved with zero human input. It is a little scary what machines can do if you teach them how to learn.

Do you think it'll happen in this century?

 

[Syed1]

Do you think it'll happen in this century?

Obviously... have you seen how rapidly technology is advancing.

 

[BlackShadow]

Imagine the superiority of the human brain, we can drive a car while day dreaming and thinking about the wonderful nihari that is waiting for us at home (or being on the phone - which is a garbage habit to have) and still drive safely. Meanwhile a car needs about 200 sensors and several cameras, radars, lidars etc streaming a tera-byte of data per second just to keep the car in between two lines. Machines have a long way to go before they can eclipse humans.

Do biological cameras and radars our brain relies on don't count as sensors? The scary thing about machines is the rate at which they evolve. They can surpass millions of years of human evolution in a matter of few decades.

 

[good guy]

this thing will only will happened if we have cheap oil

which do not have so much as per hsbc report

 

[Cpt. Rishwat]

AI technology should have a kill-switch if things get too out of hand (which they will).


Example of AI technology learning and improving itself:


There is a stretch of highway in California on which human drivers don't really struggle since they can anticipate the turn and angle of highway and when to brake and accelerate, and at what angle to steer. When Teslas in autopilot mode first started driving on that stretch they struggled a great deal, once or twice went out of lane and alerted the driver to immediately take control. However, over-time when more and more Teslas drove on that stretch and they each shared their data with the other through machine learning the car learned how to navigate that stretch perfectly. All of this was achieved with zero human input. It is a little scary what machines can do if you teach them how to learn.

Wouldn't any half decent AI learn to override any kill switch?

 

[Madplayer]

Wouldn't any half decent AI learn to override any kill switch?

True. Kill switches wont work.
I think if Machines reach a dangerous level of intelligence, humans would have to adapt by losing part of their humanity. Only a human+machine hybrid would survive in competition against machines.

 

[troodon]

True. Kill switches wont work.
I think if Machines reach a dangerous level of intelligence, humans would have to adapt by losing part of their humanity. Only a human+machine hybrid would survive in competition against machines.

Humans will all be cyborgs in the coming centuries. This is done to override death.

 

[Madplayer]

Humans will all be cyborgs in the coming centuries. This is done to override death.

Override Natural death only isnt it?

And What could be the consequences? Travel to other planets (provided cyborgs wont need oxygen) and over population (provided reproduction isnt shunned completely). It would interesting to see how humanity adapts.

 

[PetroDollars]

Ofcourse there can be a kill switch but how good will it be when 'the AI network' will be a billiom times more intelligent then humans.

read 'Transhumanism - a dangerous idea' by david livingston - an islamic scholar

 

[s28]

Scary

<iframe width="560" height="315" src="https://www.youtube.com/embed/ecClODh4zYk" frameborder="0" allow="autoplay; encrypted-media" allowfullscreen></iframe>

 

[Yossarian]

Drones are already being designed to use AI technology to identify and eliminate targets on a hit list without the drones being controlled by human operators.

Similarly, tanks, planes, warships and subs that have no military personnel on board are currently on drawing boards being designed to fight wars using AI.

It's not far fetched to imagine, in the not so far away future, the Armed Forces of two Superpowers (say USA and China) being completely based around AI technologies. A mistake by one or the other, and suddenly they're fighting an all out war including attacking each other's infrastructure and population centres.

 

[Donal Cozzie]

To be fair I must've slaughtered thousands of AI's on Age of Empires alone let alone the hundreds of other games I've played so only fair they get a chance at striking back.

 

[DeadBall]

I think Trump will destroy humans much earlier but he doesn't seem to have any intelligence at all so I don't know if this qualifies.

 

[DeadBall]

To be fair I must've slaughtered thousands of AI's on Age of Empires alone let alone the hundreds of other games I've played so only fair they get a chance at striking back.

You are a nerdy (what I remember from the post your pics thread pic) yet fair man.

 

[KingKhanWC]

Drones are already being designed to use AI technology to identify and eliminate targets on a hit list without the drones being controlled by human operators.

Similarly, tanks, planes, warships and subs that have no military personnel on board are currently on drawing boards being designed to fight wars using AI.

It's not far fetched to imagine, in the not so far away future, the Armed Forces of two Superpowers (say USA and China) being completely based around AI technologies. A mistake by one or the other, and suddenly they're fighting an all out war including attacking each other's infrastructure and population centres.

What about robot soldiers?

The problem for superpowers such as the US is the deaths of people is what holds them back from putting more boots on the ground, the public dont have the stomach for it. Since Iraq in 2003, the US has only managed to bomb from the air.

 

[maverick85]

There is this theory of singularity, where humans are a link to the next phase of evolution. We are all but small flashes of intelligence limited by our decaying bodies.

 

[Yossarian]

What about robot soldiers?

The problem for superpowers such as the US is the deaths of people is what holds them back from putting more boots on the ground, the public dont have the stomach for it. Since Iraq in 2003, the US has only managed to bomb from the air.

"Robot soldiers" don't need to have 2 arms, 2 legs and look like human beings. They simply need to be all-terrain devices with numerous sensors/ 'tentacles/arms' and multiple types of weapons.

 

[CricketCartoons]

World needs a reboot. Whatever brings it faster. Whether AI or clash of civilizations.

 

[KingKhanWC]

"Robot soldiers" don't need to have 2 arms, 2 legs and look like human beings. They simply need to be all-terrain devices with numerous sensors/ 'tentacles/arms' and multiple types of weapons.

Do you have any example of how this all terrain device would look like? Let's use the most difficult nation to occupy as an example, Afghanistan. The terrain is very tough , steep mountains, obstacles such as trees, rocks etc and this is just the north. The speed of such devices would be very slow and easy targets for anyone fighting against them.

 

[Robert]

Do you have any example of how this all terrain device would look like? Let's use the most difficult nation to occupy as an example, Afghanistan. The terrain is very tough , steep mountains, obstacles such as trees, rocks etc and this is just the north. The speed of such devices would be very slow and easy targets for anyone fighting against them.

I bet those Boston robots could manage that ok. They could call up robot drones as air support.

 

[Robert]

What about robot soldiers?

The problem for superpowers such as the US is the deaths of people is what holds them back from putting more boots on the ground, the public dont have the stomach for it. Since Iraq in 2003, the US has only managed to bomb from the air.

There are US soldiers in many countries. Djibouti for instance.

 

[DeadBall]

I think Trump may well cause that before them but he was talking about intelligence after all, artificial or not.

 

[KingKhanWC]

I bet those Boston robots could manage that ok. They could call up robot drones as air support.

Air support becomes very limited as the terrain is full of mountains, trees and rocks. They cannot see targets from an ariel position, the only way is to physically climb to the location of the enemy. Those robots will be far too slow and will be shot to pieces before they even take a few steps.

There are US soldiers in many countries. Djibouti for instance.

In terms of occupation they cannot openly send in hundreds of thousands of soldiers any longer, their deaths are too much for the public to handle esp when it's known their lives were wasted as they were in Afghanistan and Iraq. Also the US forces are vastly over rated in terms of great warriors and the same goes for the British. Both have struggled to defeat a village resistance force in Afghanistan and then restored to bombing the village from air.

 

[Yossarian]

Do you have any example of how this all terrain device would look like? Let's use the most difficult nation to occupy as an example, Afghanistan. The terrain is very tough , steep mountains, obstacles such as trees, rocks etc and this is just the north. The speed of such devices would be very slow and easy targets for anyone fighting against them.

Since you're the one who brought up the possibility of robot soldiers, and you appear to disagree that they just need to be all terrain devices that don't need to have 2 arms, 2 legs and look like human beings, does that mean you think they have to have 2 arms, 2 legs, look, walk, interact with the locals and fight like human beings?

 

[KingKhanWC]

Since you're the one who brought up the possibility of robot soldiers, and you appear to disagree that they just need to be all terrain devices that don't need to have 2 arms, 2 legs and look like human beings, does that mean you think they have to have 2 arms, 2 legs, look, walk, interact with the locals and fight like human beings?

No not at all, I never mentioned any type of design involving legs or arms. I just wondered if you had any idea. I just don't feel vehicles will not be good enough for the various terrains they would face in places such as Afghanistan. Maybe in time they will develop robots which are flexible and able to maneuver themselves climbing mountains, through thick forrests, swimming through rivers, getting themselves out of a ditch etc etc

 

[Yossarian]

No not at all, I never mentioned any type of design involving legs or arms. I just wondered if you had any idea. I just don't feel vehicles will not be good enough for the various terrains they would face in places such as Afghanistan. Maybe in time they will develop robots which are flexible and able to maneuver themselves climbing mountains, through thick forrests, swimming through rivers, getting themselves out of a ditch etc etc

Being robots, and thus not feeling pain or needing to breathe, as well as being much more robust than human skin and bones, I doubt if the physical constraints you mention will be a major issue considering the rate at which all these technologies, from AI to miniature drones, are developing.

 

[Robert]

Miniature drones, aye.

We will start to see swarms of killer bots the size of wasps or smaller, unstoppable by gunfire, descending on people and stinging or even eviscerating them.

 

[s28]

Scary

<iframe width="560" height="315" src="https://www.youtube.com/embed/ecClODh4zYk" frameborder="0" allow="autoplay; encrypted-media" allowfullscreen></iframe>

Gee I wonder where Robert got that astoundingly original idea from

 

[maverick85]

 

[hussain123]

AI is inevitable.

 

[Kianig89]

Exactly , these tools are helpful but they are very addictive , let's see what happens

 

[topspin]

Will the dajjal use AI?

 

[Kianig89]

Dajjal might be AI himself,

 

[Ball Blazer]

launching AI themselves and then making these predictions...