Google's AI built its own AI

[The Caspian Hare]

Quote

 

In May 2017, researchers at Google Brain announced the creation of AutoML, an artificial intelligence (AI) that's capable of generating its own AIs.

More recently, they decided to present AutoML with its biggest challenge to date, and the AI that can build AI created a 'child' that outperformed all of its human-made counterparts.

The Google researchers automated the design of machine learning models using an approach called reinforcement learning. AutoML acts as a controller neural network that develops a child AI network for a specific task.

For this particular child AI, which the researchers called NASNet, the task was recognising objects - people, cars, traffic lights, handbags, backpacks, etc. - in a video in real-time.

 

 

http://www.sciencealert.com/google-s-ai-built-it-s-own-ai-that-outperforms-any-made-by-humans

[Piney]

 Big Brother's Big Brother is watching............

[Rlyeh]

I suppose calling it an AI makes it sound more scary than what it really is, a neural network.

[_KB_]

There's no way that an AI that can develop a smarter AI is a bad idea... No really, no way, it's not like they're gonna doom us all with an AI like that... well if you excuse me i'm gonna go watch the terminator as i go over S. Hawkings warnings about this kinda stuff... but really no way it's gonna go evil and wipe us out :P

[Jon the frog]

Well digging our own grave we do.

[SHaYap]

31 minutes ago, Jon the frog said:

Well digging our own grave we do.

I think we have robots for that too ...

~

 

[Orphalesion]

Yay! Benevolent AI dictatorship in 10...9...8...

[tmcom]

As long as we don't hook it up to nuclear weapons all is well.

[SHaYap]

9 minutes ago, tmcom said:

As long as we don't hook it up to nuclear weapons all is well.

Perhaps better to program them to not ever try to build it themselves ...
 

Quote

 

~

Open Letter on Autonomous Weapons - Future of Life Institute

https://futureoflife.org/open-letter-autonomous-weapons/

Autonomous Weapons: an Open Letter from AI & Robotics Researchers ... described as the third revolution in warfare, after gunpowder and nuclear arms. ... weapons, most AI researchers have no interest in building AI weapons — and do not ...

~

 

 

[RAyMO]

23 hours ago, Rlyeh said:

I suppose calling it an AI makes it sound more scary than what it really is, a neural network

IDK Rlyeh, to me Neural Network is way scarier. In fact up to now I was quite calm about  AI. But Neural Network has shaken me up a bit.

Left to its own devices god knows what the little burgers will get up to. I bet like children always do they will get into things that they have no business being anywhere near.

"Oops I was just playing with the Nuclear launch software, mommy - I didn't mean to set it off."

Edited December 5, 2017 by RAyMO

[Rlyeh]

8 minutes ago, RAyMO said:

IDK Rlyeh, to me Neural Network is way scarier. In fact up to now I was quite calm about  AI. But Neural Network has shaken me up a bit.

Left to its own devices god knows what the little burgers will get up to. I bet like children always do they will get into things that they have no business being anywhere near.

"Oops I was just playing with the Nuclear launch software, mommy - I didn't mean to set it off."

What? An artificial neural network is a group of connected nodes used in pattern recognition. Real scary stuff..

[geraldnewfie]

let hope the nukes isnt hooked to any network!

[EBE Hybrid]

Being able to recognise objects in real time, couple that to state of the art optics and you could have an impressive weapons targeting system, no radar required, therefore your target will not realise that you have weapons lock. Cool

[tmcom]

40 minutes ago, third_eye said:

Perhaps better to program them to not ever try to build it themselves ...
 

 

True, we should stay with autonomous cars!

8 minutes ago, geraldnewfie said:

let hope the nukes isnt hooked to any network!

Or any infant AI, look what happened in Star Trek the Motion Picture, Veger, had a fit and almost wiped us out!

[paperdyer]

SKYNET!!!!!!

[UFOwatcher]

The AI craze or fear....1.2% more accurate, I can't get excited over that as an accomplishment...

[FlyingAngel]

WARNING they now have the ability to "breed", exponentially "evolve" faster than human evolution. Don't hook it to the internet !

I repeat, DON'T HOOK IT TO THE INTERNET !!

[Jon the frog]

1 hour ago, UFOwatcher said:

The AI craze or fear....1.2% more accurate, I can't get excited over that as an accomplishment...

Well it's way faster than biological evolution...

[Dark_Grey]

1 hour ago, UFOwatcher said:

The AI craze or fear....1.2% more accurate, I can't get excited over that as an accomplishment...

Well, some of the top minds alive today have spoken out against AI, warning of the grave consequences. Minds like Elon Musk and Steph Hawking have expressed real concern over it, leaving me to wonder why their opinions aren't being taken seriously. There is a coalition of Scientists and think tanks that drafted letters expressing concern as well. We are building a demon, piece by piece, and once it takes off, there is literally no stopping it. Remember, just the other week Facebook had to shut down their AI because it started communicating in it's own language. What happens when we flick the switch to shut it down and nothing happens? We have an AI creating better AI in a language we can't even understand?

We are already witnessing unforeseen behavior and AI is still in it's infancy. We are monkey people toying with fire - we have to be burned before we understand it's power but I'm worried it will already be too late when that happens.

[Dark_Grey]

8 minutes ago, Jon the frog said:

Well it's way faster than biological evolution...

It's projected to develop 200 years worth of technology in just under a week. The question isn't if the AI is capable, it's what humans will do such insane power we haven't prepared for.

[Jon the frog]

Just now, Dark_Grey said:

It's projected to develop 200 years worth of technology in just under a week. The question isn't if the AI is capable, it's what humans will do such insane power we haven't prepared for.

Yeah, that will surpass our capacity to use it and comprehend it... quite scary but also shiny like gold and it's why it will go forward.

[Dark_Grey]

14 minutes ago, Jon the frog said:

Yeah, that will surpass our capacity to use it and comprehend it... quite scary but also shiny like gold and it's why it will go forward.

Let's put it this way: the first nation that develops true AI will have god as their ally. The race is on.

Edited December 5, 2017 by Dark_Grey

[cyclopes500]

Only 1.2% more accurate. That thing isn't self aware already and deliberately reducing its capabilities is it? May I suggest they give it a lie detector test of some sort.

[cyclopes500]

 Incidentally while posting my message my phone app Firefox crashed.

[Guyver]

10 hours ago, Jon the frog said:

Well digging our own grave we do.

Right.  I think the fears are well founded.  It's baby steps to the elevator, baby steps to the door, baby steps to our own destruction.  I mean, how is it unreasonable to think that over time we create machines that can out perform us, and out last us?  Over time, we could certainly become the "inferior species."  

Why would a machine have any problem wiping us out?  They run the numbers.....and yeah....we deplete the environment, they don't....it's the Matrix and all the other sci-fi postulates happening in reality.  I see it as a definite possiblity and future.  Pandora's box.