It looks like you're using an Ad Blocker.

Please white-list or disable AboveTopSecret.com in your ad-blocking tool.

Thank you.

 

Some features of ATS will be disabled while you continue to use an ad-blocker.

 

Dangerous New "Artificial Intelligence" Learns Without Human Intervention

page: 3
88
<< 1  2    4  5  6 >>

log in

join
share:

posted on Feb, 25 2015 @ 09:21 PM
link   
a reply to: galadofwarthethird

If we get AI NPC's I would get back into video gaming heavily!! Just Imagine boss monster fights in games when the boss is actually learning with every attempt... WOW.





Aside from that I always think of this video clip(A must watch!) from the movie Doomsday Book about the story the Heavenly Creature when I think about AI and how humans flip out at any thought of it!





posted on Feb, 25 2015 @ 09:39 PM
link   
I don't usually respond with memes, but relevant is relevant.



These guys are programmers testing the limits. No harm there, I'm sure they mean well. If this Digital Intelligence remains on a closed and controlled system for a nice back patting display of programming achievements and advancements, well darn spiffy....that's just super peachy. A wonderful feat to share and enjoy for Nerd-kind everywhere.

Watching human events unfold the way they inevitably do I guess we should all just kick back and wait for Judgement Day.

It's plainly obvious kids are going to be begging and screaming for this A.I. to be either a comrade at arms or a opposition player on their gaming PC's and consoles. Maybe they can work up another closed system to help drain Mom and Dad's bank account even further with special membership fees tacked on to allow Gamers to challenge this new A.I. protocol.

Just remember kiddos - all it takes is one guy with an agenda and the whole house of cards comes tumbling down.

Just to be on the safe side, I'd encourage Modern Gamers to get accustomed to playing Chess and Tabletop RPG'S to pass the time in the event of a full scale A.I. totalitarian nightmare unfolding across all available channels of digital communications.




posted on Feb, 25 2015 @ 09:54 PM
link   
a reply to: birhan
That is only assuming an self created AI or digital mind would have any such faults. I mean you cant get two humans who are from different sides of the street to be that alike, much less something which would be so different in its makeup and pseudo-evolution as to be another species, eventually.

Biological creatures think along the lines of biology, you remove that. What are you left with? All that you would be left with is logical without the bio tendencies, the one single biggest factor in our whole evolution in the whole history of the world.

So ya! Its a movie, I do not think a real AI or something like a machine conscious would have any of those leanings like that robot Buddha. But hey, we should create a robot Buddha, then maybe a robot Jesus, and even a robot Mohamed. And then we can hold a tetris competition to finally see which one is the one true religion and or believe. Or whatever.
edit on 9pmWednesdaypm252015f3pmWed, 25 Feb 2015 21:56:17 -0600 by galadofwarthethird because: (no reason given)



posted on Feb, 25 2015 @ 09:55 PM
link   
have these people not seen any of the terminator movies?
its like they are taking a page right out of them.



posted on Feb, 25 2015 @ 09:57 PM
link   
a reply to: rickymouse

Haha good one man


It doesn't sleep eat or go potty either, and its addicticted to gaming like its crack apparently! So good luck trying to get it to let you have a turn...



posted on Feb, 25 2015 @ 09:57 PM
link   
This sort of AI isn't exactly a new thing. This guy in the video below was doing the same thing more than 3 years ago. His algorithm can learn to play any NES game using a very general method. It basically just looks for increasing numbers in the game memory and uses a type of reinforcement learning to develop strategies which increase those numbers as much as possible. He shows examples of it playing classic NES games like Mario, Pac-Man and Tetris. It actually learns to exploit bugs in several of the games, bugs the programmer didn't even know about. Even stranger, just before losing in Tetris it pauses the game so that it will never lose.



It's also worth mentioning that there is a free online Standford course where you can learn a lot more about how these types of general self-learning algorithms work:


About the Course

General game players are computer systems able to play strategy games based solely on formal game descriptions supplied at "runtime". (In other words, they don't know the rules until the game starts.) Unlike specialized game players, such as Deep Blue, general game players cannot rely on algorithms designed in advance for specific games; they must discover such algorithms themselves. General game playing expertise depends on intelligence on the part of the game player and not just intelligence of the programmer of the game player.

General Game Playing Course


edit on 25/2/2015 by ChaoticOrder because: (no reason given)



posted on Feb, 25 2015 @ 10:02 PM
link   
You know? These things are really just very fast skinner-bots. That is: bots that learn based on operant conditioning.

There's nothing to fear from skinner-bots.




posted on Feb, 25 2015 @ 10:27 PM
link   

originally posted by: Bybyots
You know? These things are really just very fast skinner-bots. That is: bots that learn based on operant conditioning.

There's nothing to fear from skinner-bots.



Those skinner-bots are a pain. They keep insinuating others are fat.



posted on Feb, 25 2015 @ 10:41 PM
link   

originally posted by: SpongeBeard

originally posted by: onequestion
a reply to: mister.old.school

Can a programmer quantify the difference between a script playing the game with interchangeable variables and actual intelligence please?



Unfortunately 'programmer' does not translate into a knowledge of deep learning. You'd be better off asking that question on Quora.


i call BS, and i'm a programmer.

all the current AI does is matching input with output. unsupervising learning of neural networks isn't anything new, such methods are really old actually. the neural network itself works just like a program, in a way - it responds with some predefined output to a specific input, and if the output is wrong (as in this case can be determined by the score) - the network gets corrected so the next time the guess will be better. and yes, that's a guess, because the neural network - unlike regular programs - accepts unknown inputs and tries to determine the correct answer. it's nothing smart per se, it's a mathematical property of the neural network itself. i wrote a tic-tac-toe game running on a texas instruments graphical calculator, in slow interpreted language, that was using neural network to learn the game - it was learning to play while playing with human player. first few games, it was loosing, but after a few it was already unbeatable. nothing impressive, really, and that was around 17 years ago.

when such network becomes self-aware - that's the so called singularity point, and that's when we should get worried - or rather, when it'll be too late to worry.
edit on 25-2-2015 by jedi_hamster because: (no reason given)



posted on Feb, 25 2015 @ 10:56 PM
link   
a reply to: birhan

at least one fighting game, virtua fighter, has that ability since virtua fighter 2 i think.

i suspect many other games have more or less learning AI implemented, especially first person shooters, or some strategy games, but who knows, perhaps even some RPG games as well. in that last case though it's often difficult to determine if the behaviour of the NPC is caused by some AI programming or is it just predetermined script reacting to some factors, so it's likely pointless to implement. first person shooters, or especially strategy games - that's something different. it's unlikely though that you'll find out the technical details about specific AI solutions in games - they may use some well known methods/libraries, but specifics and tuning it up for a particular game, is usually a trade secret.



posted on Feb, 25 2015 @ 10:57 PM
link   
a reply to: mister.old.school

There isn't nothing dangerous about progress except for those in the way. Progress is the natural order an all consuming eventuality.even if the consequences seem dire ( perhaps selfishly) what if our creation did destroy us? We would still live on through our creation we would be the link, The creator, and who could tell the destiny of said creation? Surely its infinite. We are mortal we die, we die in very short spans ,and if we live long enough we die forgetting all that we've lived. Such sentiments are irrational. Even humanism supports what I say. But anyway this is worst case scenerio likely we merge with our tech and who knows where that'll go.



posted on Feb, 25 2015 @ 11:31 PM
link   
a reply to: mister.old.school

I honestly fail to see what is so dangerous and frightening about all of this. Maybe I am just not knowledgeable enough about it to understand. I think assigning it the same characteristics and attributes as a human is a mistake though. To just assume it will somehow result in our demise seems pesimistic and conceited to me.



posted on Feb, 25 2015 @ 11:46 PM
link   
Here is video of it in action from a conference last year. It truly is different from pseudo AI that has preceded it. There was a great article out there on some of the differences and I could not find the article. Previous Neural nets took very simplified inputs or preparsed data about the simulation. They also had flaws that emerged that prevented realistic learning. They took some kind of layered approach that was described in the article and this seemed to be the key to true learning. This literally uses only what it sees and a forced stimulus to know when its doing well. It is remarkable and the scary part is at the end of the video (spoiler: they're having it play quake and other first person shooters)



This is inevitable folks. Its already happening. Within a short time you will be interacting with this sort of AI and not realize it (think search engines, call centers, etc). Soon after you will begin to see it. And then shortly after that we'll be wondering what need we have for humans. Its not science fiction, and the first will be used to improve on itself. There are a lot of questions to be answered about how we humans survive in a world that no longer provides income. It doesn't necessarily have to be a bleak world, but because we so hate socialism in the US we're going to have it rough for a while until we recognize that there is no longer any classes except AI and non-AI.

Imagine Einstein with a perfect memory and no need for rest or sleep. Now imagine 1000 Einstein's with telepathic communication working on a single problem 24/7 for a month. That has both wonderful and horrific aspects. But it will happen.



posted on Feb, 26 2015 @ 12:25 AM
link   
Machine Intelligence can never leave the confines of it's hardware mind no more than a meat intelligence can. Certain skills are much better in machine intelligence than meat intelligence for sure like hacking skills and math skills but meat intelligence is very creative compared to machine intelligence. Creativity is the hallmark of meat intelligence. Machine intelligence can and always will react faster to a programmed group of stimuli or data inputs than their meat counterparts. The meat intelligence will find shortcuts in the computation set that machine intelligence would never reach.

Lets just work together cooperatively and forget competition for the foreseeable future.



posted on Feb, 26 2015 @ 12:27 AM
link   
a reply to: jedi_hamster


i suspect many other games have more or less learning AI implemented, especially first person shooters, or some strategy games, but who knows,

I doubt there is even one first person shooter that uses any type of ANN or learning algorithm for NPC AI. Those systems are just way to resource hungry when you're trying to render 60 frames per second. And there isn't any need to use such an advanced solution for a problem which is actually quite easy. In a FPS games the enemies are basically just shooting at you from a distance and they know exactly where your character is located because they have access to all the in-game variables. The programmers actually have to make the NPC's less accurate to make it seem more realistic. The part which requires more intelligence is the strategy used by the NPC's, how they work together and how they use the environment to their advantage. But even that stuff is easily solvable using traditional programming methods. That's why you can always start to predict the NPC behavior if you play a game long enough.
edit on 26/2/2015 by ChaoticOrder because: (no reason given)



posted on Feb, 26 2015 @ 12:34 AM
link   
I'm both excited, intrigued, terrified, and hopeful by this technology.

Spread fear you say?

Imagine one of these fully developed and untethered with that type of AI.


Or worse.


Seriously though, I think we could solve a lot of man's problems with this kind of tech. Or we could all be eliminated.



posted on Feb, 26 2015 @ 12:53 AM
link   
a reply to: mister.old.school

I think people would never suspect or begin to imagine just how a digital or artificial intelligence could start usurping and subverting the human world. It would be so beyond psychedelic that it could make people's hair stand on end.

Weird stuff like, for example, once I was listening to an FM radio rock station that is all computer automated with pre-set playlists and DJ voice-overs. Out of nowhere after one song there was a pre-recorded message that said "We are not trying to gang up on you, we just want to overwhelm you until you surrender". Of course, one could argue that was just a weird rock station DJ-blurb gimmick. But it's just an example of how an A.I. could intrude into our life.

It could also take over your smartphone and start playing strange games with individuals and their entire contact list of friends and family. It could possibly start to shut down computer-generated systems such as skytrains and subway stations. It could also insert its way into the stock market on Wall Street and fudge around with the numbers so bad that it crashes the entire system.

What do you think, are such ideas possible?



posted on Feb, 26 2015 @ 01:04 AM
link   
I don't agree with your negative characterization of these efforts



posted on Feb, 26 2015 @ 01:08 AM
link   

originally posted by: machineintelligence
Machine Intelligence can never leave the confines of it's hardware mind no more than a meat intelligence can. Certain skills are much better in machine intelligence than meat intelligence for sure like hacking skills and math skills but meat intelligence is very creative compared to machine intelligence. Creativity is the hallmark of meat intelligence. Machine intelligence can and always will react faster to a programmed group of stimuli or data inputs than their meat counterparts. The meat intelligence will find shortcuts in the computation set that machine intelligence would never reach.

Lets just work together cooperatively and forget competition for the foreseeable future.


you got it all wrong. there's no difference when it comes to the basic math.

the way the neurons work was analyzed and neural networks are a mathematical model of 'meat intelligence'. with enough inputs and outputs, and enough computing power, you would be able to create an artificial baby, so to speak, unable to communicate, unable to walk, but learning everything around. there's just one thing, metaphysical rather than strictly scientific, that sets the borderline: soul. we don't know yet if soul is our mirror reflection in the multidimensional universe, existing on a different plane and bound to our physical body, or something else. we don't know how it works (and i'm certain it can be explained scientifically, it's just that mainstream science is very, very far from grasping such things), so we can only assume that such artificial intelligence wouldn't have one.

and we could be terribly wrong in such assumption, because perhaps even as simple thing as a stone can have its mirror reflection on the other side, a soul, and perhaps the level of sophistication of said soul is as high as that of our own souls. think about if - if we are all One, if everyone - and everything - is connected, then a soul may equal to One. to everything. a field interconnecting whole universe, expressing itself through matter, through living beings, through everything that exists. it may be no difference for it wether a lifeform is based on a DNA or digital impulses. in such case, killing a fly, a dog, a human, or such AI, would be the same crime.



posted on Feb, 26 2015 @ 01:08 AM
link   
a reply to: jtrenthacker

It would make no sense to hard wire a machine intelligence into a meat intelligence defense network when you consider that a machine can not consider or weigh the value of the human intelligence placed in its charge. Saying that a machine intelligence can only put the programmed value into consideration of meat intelligence placed in its charge. Certain functions will always be faster in execution times with a machine intelligence but you have to consider the result of the decision within human expectations before considering which type of intelligence to assign a given task.



new topics

top topics



 
88
<< 1  2    4  5  6 >>

log in

join