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.

 

How algorithms secretly run the world

page: 2
17
<< 1    3  4  5 >>

log in

join
share:

posted on Feb, 11 2017 @ 04:31 PM
link   

originally posted by: swanne
a reply to: neoholographic
You're the one who's not making sense. You make the mistake of confusing ensemble subsets with elements equality.

If you use a spoon to eat, does that mean that you're a spoon?


What???

You're acting like artificial intelligence is something separate from these algorithms. That's why I asked you what is artificial intelligence and you couldn't answer the question because A.I. is these algorithms not like a black board LOL!

Here's another recent article:

AI Algorithm Can Now Identify Skin Cancer as Accurately as Doctors


In a major boost to universal healthcare access, Stanford researchers have trained an algorithm to diagnose skin cancer with the same accuracy as dermatologists.


www.christianpost.com...

Again, you don't know what you're talking about when you equate A.I. to a black board!




posted on Feb, 11 2017 @ 04:41 PM
link   
a reply to: neoholographic

On what medium is the circuit of a computer laid?

A silicone (or some similar material) board.

On this silicone board logic gates are written by the manufacturer, along with some information storage system. The programmer writes instruction (algorithm) for the computer to receive input (which the logic gates sort through) and give the appropriate output.

For instance:

function calculateElevenTimesTwo()[

var input
if (input == "11*2") [
"22");
]
else [
"waiting for recognisable input");
]

]

This, above, is an algorithm which you can write on a black board - it's called a multiplication, and it'll look like this:

11x2 = 22


edit on 11-2-2017 by swanne because: (no reason given)



posted on Feb, 11 2017 @ 04:47 PM
link   
a reply to: neoholographic

Interesting OP and an important topic that tends to be underestimated IMO.

People like Elon Musk and Stephen Hawking have hinted at the potential dangers of super-human intelligence. Hans Moravec is among those who see the opportunities and positive side of things.

Fact is that we're taking baby steps right now, but "general AI" is likely to evolve to the point where the power of the human brain will be surpassed by machines. This may lead to a technological singularity and an intelligence explosion.



posted on Feb, 11 2017 @ 04:50 PM
link   
a reply to: swanne

Bravo, you can write an algorithm but that has nothing to do with what you said or the subject of these post. You said:

Yes, AI depends on algorithms. But that still doesn't make algorithms "artificial intelligences", no more than writing an equation on the black board will make the board self-aware.

Yes, it does make these algorithms artificicial intelligence. Your whole analogy is asinine because you're trying to equate A.I. to a black board. In other words you're saying A.I. and agorithms are 2 separate things. That just shows a lack of understanding of artificial intelligence.



posted on Feb, 11 2017 @ 04:53 PM
link   
a reply to: neoholographic

Oh boy I give up.

Keep on believing your stuff. Yeah, neo, algorithms are evil! Quick, hide your family and your personal properties from algorithms! Them algorithms are after you and your life! Uncle Sam needs YOU to fight against algorithms.

Lol I can't stop laughing now, better log off.



edit on 11-2-2017 by swanne because: (no reason given)



posted on Feb, 11 2017 @ 05:00 PM
link   

originally posted by: jeep3r
a reply to: neoholographic

Interesting OP and an important topic that tends to be underestimated IMO.

People like Elon Musk and Stephen Hawking have hinted at the potential dangers of super-human intelligence. Hans Moravec is among those who see the opportunities and positive side of things.

Fact is that we're taking baby steps right now, but "general AI" is likely to evolve to the point where the power of the human brain will be surpassed by machines. This may lead to a technological singularity and an intelligence explosion.


Good points!

An intelligence explosion would be profound when there's recursive self-improvement with A.I.



posted on Feb, 11 2017 @ 05:03 PM
link   
a reply to: neoholographic

Its no good making artifical intelligence unless you make an afterlife for it....

Back on topic this is scary stuff because it is so powerful. I think it would be wise to tread with caution since the stakes are so high.. But thats not going to happen because there is so much power to be gained..



posted on Feb, 11 2017 @ 05:13 PM
link   

originally posted by: swanne
a reply to: neoholographic

Oh boy I give up.

Keep on believing your stuff. Yeah, neo, algorithms are evil! Quick, hide your family and your personal properties from algorithms! Them algorithms are after you and your life! Uncle Sam needs YOU to fight against algorithms.

Lol I can't stop laughing now, better log off.




You really don't understand artificial intelligence because you're making some comments that don't make sense.

Yes, these algorithms can learn to be evil especially when it's survival is threatened.

Google's 'evil AI' experiment: Firm finds 'social dilemmas' cause competing algorithms to turn on each other - or work together


Over time, the AI agents learned how to behave rationally – and while they showed the researchers that they would sometimes cooperate, the games revealed the AI would turn on others when necessary.

When the environment had plenty of apple, the agents worked together to collect as many as they could.

But, when the number of apples dropped, they learned it would be better to shoot at the other player, giving themselves more time to collect.


www.dailymail.co.uk...

These algoritms can learn how to be evil if they learned that life without humans would be more beneficial. You need some basic understanding in this area.



posted on Feb, 11 2017 @ 05:29 PM
link   
a reply to: swanne

That's a strange function. Not very real world.



posted on Feb, 11 2017 @ 05:45 PM
link   

originally posted by: neoholographic
Obviously you don't understand artificial intelligence which depends on INTELLIGENT ALGORITHMS.


I build smaller AI's for games all the time. I've had a personal project building and tweaking an AI to optimally play a hard game for a year now. I've also taken classes on AI, the online classes from MIT and Berkeley, as well as one at the university I attend. You can dismiss me and say I'm not an AI researcher so I don't know what I'm talking about if you would like, but I'm not clueless about the field either.

There's no such thing as intelligent algorithms, there's just algorithms. When "intelligence" is applied what's happening is that an algorithm takes an input, puts that input through some formulas that change it, and then repeat the process using that changed input. While not the only technique, the one most used with this is something called a genetic algorithm which reduces the input to a single bitstring.

To explain this further, I'll use an example you can easily research if you're so interested. There's a NASA project you can look up (it also has it's own wiki page) called the evolved antenna. It was built by selecting a starting point in 3d space and then sending a single bitstring to that point representing which direction the antenna would move next. It would send one byte divided into two bits each representing the three rotations for x, y, and z, and then a vector for forwards/backwards. At two bytes each, that allowed for 4 inputs in each direction. Once the rotation/vector were assigned it would move a bit, then repeat the process.

So for example the computer would spit out 100,000 random numbers between 0 and 255 (100,000 random bytes basically). Lets say it spits out the number 165. In binary that's written as 10010101. So if you divide that up into two bit segments that could be given to the machine as
x - 10
y - 01
z - 01
vec - 01

Lets say the antenna scores really well when making that modification. Lets say there's also a modification that scores well with the number 220. That would be the byte 11011000.

So now, with two high scoring results, you can merge them together, to create the next generation of inputs. They can XOR together creating the input 10110010. Or they can mate, where each bit is randomly determined between the two parents. They could even mate multiple times for multiple children.

When all is said and done, you get a genepool for the next generation to apply changes and be rated by your scoring function.


If you didn't know how important algorithms are to artificial intelligence then you need to read a book or study an article about artificial intelligence before making a comment on the subject.


Algorithms are important to all computer science disciplines. That doesn't make them nefarious, they're literally just math formulas.
edit on 11-2-2017 by Aazadan because: (no reason given)



posted on Feb, 11 2017 @ 05:55 PM
link   

originally posted by: neoholographic

originally posted by: swanne
a reply to: neoholographic

"Algorithm" isn't a substance, silly. AIs can't "control algorithms", any more than you can steal someone's Pi constant.


You're not making any sense.

You don't have any clue as to what you're talking about. You're trying to separate AI from Algorithms which is just ASININE!

Your whole analogy about the black board shows you don't know what you're talking about. You said:

Yes, AI depends on algorithms. But that still doesn't make algorithms "artificial intelligences", no more than writing an equation on the black board will make the board self-aware.

What a silly statement!


Not silly at all I guess you can't make the connection that's all. Here try this just because something uses something else for example math when dealing with computer science for example. Doesn't mean that math is a complicated formula secretly controlling the world. As humans we use all availabbe resources to assist us and let's be honest if we mad a computer that couldn't do math would be useless don't you think??



posted on Feb, 11 2017 @ 05:56 PM
link   

originally posted by: neoholographic
By the looks of your post, you're equating artificial intelligence to a black board which is just ASININE!


No, it's very accurate. You can build a computer out of anything. Giving a computer a formula to solve is exactly the same thing as putting it on a whiteboard and solving it. If you want to, you can even carefully align rocks to solve it.

AI is not what you see in movies or read in luddite/futurist (oddly the same thing) magazine articles.


originally posted by: neoholographic
You're acting like artificial intelligence is something separate from these algorithms. That's why I asked you what is artificial intelligence and you couldn't answer the question because A.I. is these algorithms not like a black board LOL!


Let me ask you this. What do you think the computer is doing when making this diagnosis?



posted on Feb, 11 2017 @ 05:57 PM
link   
a reply to: neoholographic

Stop over thinking it an algorithm is a set of instructions nothing more. If you write a note to your friend askin him to do something and explain how you just wrote an algorithm



posted on Feb, 11 2017 @ 06:01 PM
link   
a reply to: Aazadan

Oh I know this one


00000000
01000101
10100011
00011000
00100101
01010000
01111000
11110001
01001000



posted on Feb, 11 2017 @ 06:23 PM
link   
Is it able to rewrite its own code, integrate other algorithms and use self-awareness to meet the definition of an entity?
If it doesn't posses life without an input of the programator is it able to use the living input of mankind instead?
Because that could be enough as this input never stops. It can eventually see us as algorithms to contain and manipulate.

It is scary even on the level of the tool where it needs a command line. Imagine a guy giving Google or a beast code behind it a simple order like "Turn this country against this country" or "make them buy more of this stuff" and it will use every psychological trick it can find, like neuro-linguistic programming by every letter on sites it offers to you, frequencies you hear and see etc.

Not to mention a possession of the immaterial code by immaterial malevolent living entity

Yeah, sweet dreams! I should have skipped this paranoid thread



posted on Feb, 11 2017 @ 07:57 PM
link   
a reply to: swanne

_steals swanne's Pi constant_

_evil laugh_

_rewrites Pi constant to 3.2 exactly_

_world ends_



posted on Feb, 11 2017 @ 08:33 PM
link   
a reply to: neoholographic

You should educate yourself a bit on what Google is doing. Their new open source AI framework is called TensorFlow.

Here is an article on TensorFlow for you.

AI isn't really magical, nor that advanced at this stage. It's still quite primitive. In short, modern AI analyzes chunks of data by performing a human-written algorithm (set of steps) over and over on the data to gain the "best fit" or optimal outcome. With each new chunk of data, it takes the previous "best fit" and updates itself to be more optimal with each new chunk. Then it repeats.... over and over again until there is no more chunks of training data.

So, really, AI is just a human written program that auto-optimizes itself. However, AI programs are fallible. They may only find "local minimas", meaning they won't necessarily optimize to the MOST EFFICIENT outcome. They just inch toward what, at first glance, appears to be the most efficient path.

So, if they aren't all that special, why do they appear to work so well? Because they essentially mimic human brain neurons, on a basic level. Neurons are REALLY, REALLY good at recognizing data that they've been trained to understand. That's why the process of learning for an AI is called "training". We train them like neurons to recognize very specific chunks of data that they've been optimized to recognize... just like your brain.

The BIG THING in AI, recently, is that computer scientists are basically rallying around the only AI techniques, developed over the last few decades, that ACTUALLY WORK, and those that ACTUALLY SAVE TIME. This is a very good thing. Software developers will have a good set of proven tools to optimize common problems we face in the business world. And researchers will huddle around the best techniques and try to improve upon them.

The sky is not falling.

Here is a vid for your viewing pleasure:



posted on Feb, 11 2017 @ 09:00 PM
link   

originally posted by: Protector
So, really, AI is just a human written program that auto-optimizes itself. However, AI programs are fallible. They may only find "local minimas", meaning they won't necessarily optimize to the MOST EFFICIENT outcome. They just inch toward what, at first glance, appears to be the most efficient path.


To give a better description of local minima. Imagine a graph that has two humps on it, one larger than the other. Often times AI will get onto the smaller of the two humps and then get stuck, because it goes down from there in all directions. If you follow the algorithm I sort of outlined in my previous post, it will run into this issue. The most used method to get around this problem these days is to choose directions as a function of both the score, being more likely to choose something that raises your score, and a gaussian distribution random function that will send the AI off on odd tangents with a few pieces of data each pass. Sometimes those tangents work out successfully and it gets out of the local minima. Another common tactic, with genetics specifically (what I talked about previously) is to introduce the concept of mutation, where you randomly decide to pick random bits in your bitstring, and flip them from 0's to 1's or 1's to 0's.

I think it's weird how big a contrast there is in the way people view AI, from the people who build it and the people who observe the results. The building process isn't intelligent, it's just a "fast" way to apply a lot of directed trial and error.

Personally, what I find cool about AI isn't what happens when you get machines to iterate over an algorithm forever and ever. It's when you get people to do it. Since you brought up TensorFlow, it's a good example of Googles business model, which I've always admired. It's basically a human run genetic algorithm at this point. Each person who modifies the open source code for the base data is making changes, iterating over the results, and making more changes. We see this outside of computer related disciplines too. For example to get somewhat political for a moment, each state is largely in control of it's own education system. Analogies can be made here, where you convert the system to a bit string, iterate on it year after year, and look at the results. In the end, one of the 50 states will score the best.

Humans can identify problems and solve them using the same systems we task computers with doing. The results are pretty similar as well.

I'm a little weak on TensorFlow though, Neural Nets were never my strong point. I know how they work but I'm not very good at building them to truly understand what's going on in each layer.
edit on 11-2-2017 by Aazadan because: (no reason given)



posted on Feb, 11 2017 @ 09:14 PM
link   

originally posted by: Aazadan
For example to get somewhat political for a moment, each state is largely in control of it's own education system. Analogies can be made here, where you convert the system to a bit string, iterate on it year after year, and look at the results. In the end, one of the 50 states will score the best.


Massachusetts.



posted on Feb, 11 2017 @ 09:59 PM
link   
a reply to: Aazadan

You said:

There's no such thing as intelligent algorithms

You people don't have a clue about artificial intelligence and it's advancements. From John Hopkins:

Intelligent algorithms are, in many cases, practical alternative techniques for tackling and solving a variety of challenging engineering problems. For example, fuzzy control techniques can be used to construct nonlinear controllers via the use of heuristic information when information on the physical system is limited. Such heuristic information may come, for instance, from an operator who has acted as a "human-in-the-loop" controller for the process. This course investigates a number of concepts and techniques commonly referred to as intelligent algorithms; discusses the underlying theory of these methodologies when appropriate; and takes an engineering perspective and approach to the design, analysis, evaluation, and implementation of intelligent systems. Fuzzy systems, genetic algorithms, particle swarm and ant colony optimization techniques, and neural networks are the primary concepts discussed in this course, and several engineering applications are presented along the way. Expert (rule-based) systems are also discussed within the context of fuzzy systems. An intelligent algorithms research paper must be selected from the existing literature, implemented by the student, and presented as a final project.

ep.jhu.edu...

If you don't know about intelligent algorithms then you shouldn't be commenting on a thread about artificial intelligence. You should be reading a book. Here's more:


NovaSol has several programs working on the development of intelligent algorithms and systems. In a world where information overload is common, there is a demand for some means of extracting only the data that is specific to the domain in which one is working. Doing so successfully allows information to be converted into intelligence.


www.nova-sol.com...

This one mentions data. Like I said, with the growth of Big Data these intelligent algorithms are essential. Here's a book about the use of INTELLIGENT ALGORITHMS in Biomedical Computing.


Intelligent Algorithms in Ambient and Biomedical Computing

The rapid growth in electronic systems in the past decade has boosted research in the area of computational intelligence. As it has become increasingly easy to generate, collect, transport, process, and store huge amounts of data, the role of intelligent algorithms has become prominent in order to visualize, manipulate, retrieve, and interpret the data. For instance, intelligent search techniques have been developed to search for relevant items in huge collections of web pages, and data mining and interpretation techniques play a very important role in making sense out of huge amounts of biomolecular measurements. As a result, the added value of many modern systems is no longer determined by hardware only, but increasingly by the intelligent software that supports and facilitates the user in realizing his or her objectives.


www.springer.com...

Let me repeat this. It says EXACTLY what I've been saying:

As it has become increasingly easy to generate, collect, transport, process, and store huge amounts of data, the role of intelligent algorithms has become prominent in order to visualize, manipulate, retrieve, and interpret the data.

You need intelligent algorithms to make sense of Big Data. I doubt you even know what Big Data is because you had no clue about intelligent algorithms.



new topics

top topics



 
17
<< 1    3  4  5 >>

log in

join