The Ultimate A.I. .. Echelon?

I was reading a different thread, and a poster made this statement:

Sometime in the year 2010 the amount of useful knowledge in the world will be doubling every hour.

If Artificial Intelligence ever comes into being. The rate at which signals travel in a human brain is 100ft/sec, but in an electronic brain with a human cognitive level those same signals would travel at the speed of light. This would mean that when they turn on the first A.I. it would do 9,000,000 years worth of thinking and discovery in the 1st year of operation. The human race would jump that far ahead for every year an A.I. is working.

Quote slightly edited

This quote really got me thinking. Pretty much all our knowledge is more or less being posted on the internet. Everything from cars, to guns, to science and math.

If you had in your control, a system such as the famed Echelon, that scours the internet, phones, etc, and collects every bit of 'data'.. how easy would it be to right code that cross references every bit of knowledge, every bit of belief and every bit of physical law and in return, have a super computer that in itself, learns from everything it records?

Imagine for and instance, this AI records a thread on ATS regarding the JFK assassination.. It takes what it recorded, adds it to a database and cross referances it with everything in the 'JFK Assassination' database to see if it holds any truth. If not, it stores it regardless and uses this data as a later cross reference.

I honestly beleive the capability of the Echelon system is way downplayed. We believe the only thing it can do is look for "key words" and record the source. Do you honestly think the world's largest and smartest computer runs on such trivial code? I myself could probably write code that can do just that.. hell, YaHoo! and Google have it already. Its called a search engine.

Now imagine this "search engine" aka Echelon records every bit of data and cross references it with its data base and learns from it. The code to do this would not be horribly hard to achieve. It runs a search for let's say.. "Apples" and records all 30 million Google hits it recieves. It then scours every hit and finds similiarities between all the hits and discards the rest. It now know's what Apples are and has a several other million hits of knowledge it can then research.. much like the human brain..

One thing that I always found odd about our government, is that they only disclose sensitive information when it has either a) become to hard to hide or b) the mass population beleives heavily that it's not what it is.

So the government learns that everyone thinks Echelon is a mere search engine, and they then release the "misinformation" that it indeed is a mere search engine.. but in truth, it may be something way way way more then that..

What are your thoughts?

[edit on 10/29/2004 by QuietSoul]

I think the cod for actual learning would be quite large. Have you ever heard of COG? It is an MIT project on AI. I believe it took 12 computers just to track movement. I hate to think of the thousands of pages of code written just for that. Not to mention all the debugging

[edit on 10/29/04 by Kidfinger]

Well, I understand the code would be quite long, thousands of pages long.. but at what cost?

I mean, if you break it down, it could be as simple as an if/and/else statement.. it checks to see if X (google search) isin Db (Database) and performs a comparison of data if it is.. If its not, it records X to Db

But if you had in your hands, that worlds largest database of .. well.. everything.. and pretty much all the money you could desire to hire programmers.. would you not try to compile this data?

The data in itself could make it the absolute smartest "person" on the planet.

Ask it a question, it cross references that question with all its knowledge and tells you the answer.. how cool would that be!

Edit add:

I absolutly refuse to beleive that Echelon is a mere search engine. Anyone can type in "Bomb gun president kill" into Google and get several hits. If you had access to all the IP ports, you could infact 'google' email media's as well. Too simple... there has to be more to it.

[edit on 10/29/2004 by QuietSoul]

Thats fine for checking the data base, but to have knowledge, you must know how to use it. What kind of code would you write that would learn from its mistakes? How would you write the code to teach the AI to differintiate between the true stuff on the net and the bunk stuff? Im not saying it cant be done, just that if it were, it would take an eternity to debug it all. I think it would be more like hundreds of thousands of pages of code.

Just read your edit. I agree there is more to Echelon than we are being told. It definitly is NOT a search engine on steroids, although that may be one aspect of the system. If any body had the money and time to Debug coding of this nature and put it to use, it would be the Gov.

Originally posted by Kidfinger
Thats fine for checking the data base, but to have knowledge, you must know how to use it. What kind of code would you write that would learn from its mistakes? How would you write the code to teach the AI to differintiate between the true stuff on the net and the bunk stuff? Im not saying it cant be done, just that if it were, it would take an eternity to debug it all. I think it would be more like hundreds of thousands of pages of code.

Well, I understand its complicity. But I dont underestimate the ability of coders and writers. I mean, the very system we use now (windows, linux, ect) is a mass of code..

But in my little untrained eyes, I dont see how it could be to hard..

-First you teach it what every word means. (ie, upload a dictionary)
-Teach it how to read (which would be hard, I know)

You then give it a starting point... I'll use my old example..

Apples.

It does a nifty Google Search for apples. It reads everything it can about apples. It cross references the data. If it has 1000 hits that apples are good and 20 hits that apples are bad, it records that apples are good.

I know, I know, it seems overwhelming the think such a system MAY exist.. but honestly, and I'm repeating myself, if you had the power, and the money, would you try? Or would you setup a super-duper search engine that flashes a red light everytime it finds a simple trivial "keyword"

Originally posted by Kidfinger
I agree there is more to Echelon than we are being told. It definitly is NOT a search engine on steroids, although that may be one aspect of the system. If any body had the money and time to Debug coding of this nature and put it to use, it would be the Gov.

I believe that it was Mr. Nyquist who suggested that the real use of Echelon was actually to have the ability to shut the internet down at any time.

When you realize just how powerful a medium the internet is in transferral of information - compliaments of Google and Yahoo etc - than you can bet your bottom dollar there are 'corporations' out there that want to make sure that they have the ability to 'control' the most powerful tool ever invented by mankind...

Originally posted by QuietSoul

Originally posted by Kidfinger
Thats fine for checking the data base, but to have knowledge, you must know how to use it. What kind of code would you write that would learn from its mistakes?

I dont underestimate the ability of coders and writers. I mean, the very system we use now (windows, linux, ect) is a mass of code..

But in my little untrained eyes, I dont see how it could be to hard..

-First you teach it what every word means. (ie, upload a dictionary)
-Teach it how to read (which would be hard, I know)

You then give it a starting point... I'll use my old example..

Apples.

It does a nifty Google Search for apples. It reads everything it can about apples. It cross references the data. If it has 1000 hits that apples are good and 20 hits that apples are bad, it records that apples are good.

The problem isn't amassing the data -- google itself caches most web pages. Nor is the problem databasing and categorizing the data - again, google does this for searches and advertisers.

The ability to analyse data based on rule sets is the foundation of electronic computing today - if/and/then/else statements.

Teaching a computer to think for itself is the problem here. You can't just "upload a dictionary" to a computer and the computer then "knows" words. It doesn't "learn" anything about the words except the rules that you have already given it to deal with the data.

The only conclusions a computer can draw about information are those conclusions that we tell it to reach -- note that I say "tell," not "teach." We tell it rules, we don't teach it rules.

The difference? An electronic computer program cannot break the rules we give it. It cannot want to break the rules. This functionality does not exist for the computer. If it did, then the computer would have the ability to learn.

Most AI research these days is being directed away from thinking and more towards recognition - for instance, recognition of things in images. The computer is still not thinking and deciding, it is just recommending based on pattern searches, which are based on rules.

AI has a long way to go and in my opinion will not be possible without some type of organic component, be it DNA computing or whatever.

Zip

Interesting point there ZIP, I dont know, I can see what you mean about the whole not being able to learn since it abids by rules, but what if we were to rewrite these rules ? I dont know how, but it could possibly be done.

As for the ability to cross-reference previous knowledge, its something that could work. I guess the brain works in the same way, and dont we have rules that we have to abid to when learning ? If we can make the same rules for a computer we will be started somewhere.

Sorry i do not know much about the capabilities of this prog or any computer for that matter. It seems to me however that a prog like the one you are tallking about would be in a constent state of cross refrencing. With all the info you are talking about even with a super comp. like the ones we will have in 5 years or so, it would take way to long to search its database for every thing that it has stored. Even when it is cross refrencing it would be still taking in info and needing to cross refrence that. I highly doubt this is a very legit thing.

[edit on 29-10-2004 by Juiceifer]

Originally posted by Juiceifer
Sorry i do not know much about the capabilities of this prog or any computer for that matter. It seems to me however that a prog like the one you are tallking about would be in a constent state of cross refrencing. With all the info you are talking about even with a super comp. like the ones we will have in 5 years or so, it would take way to long to search its database for every thing that it has stored. Even when it is cross refrencing it would be still taking in info and needing to cross refrence that. I highly doubt this is a very legit thing.

[edit on 29-10-2004 by Juiceifer]

I look at this as being very unimportant. The slow speed of cross referencing database information in large databases comes from being unable to fit the entire database into a quickly accessible, solid state storage medium (think RAM). A plethora of new data storage technology is currently emerging and barring that, with enough RAM or solid state memory storage, this issue is nil.

In my opinion, a decent speed of computation for AI has already been reached and breached, to the point of not being even a minor issue. The problem with AI is in theory - how the hell do we make something artificially intelligent?

Zip

Exactly...when we say that something is AI we mean in the sense that is can think, but still it aint really thinking....AI is something that is gonna take a long time....Since anything can make a suggestion on a basis of information but producing something, unique is something different...