[Snowman]

I suppose that's why you would only use an ANN function when you absolutely need adaptability. Although, I will say this. The way processors are multiplying in speed, efficiency, and how many processors are virtualizing multicore functionality, many things that use to take minutes now only takes seconds to do.

Yesterday, I was using a program that let me play around with ANN. It had an example code and it told me how long it should take to finish teaching it. Its said, "just under a minute". I built my pc last year and bought the new stuff. I didn't buy the most expensive, but I didn't buy the cheapest either. Anyway, it took literally 2 seconds to completely finish teaching the ANN. This is to be expected considering Moore's law.

There is also coming a memristor computer revolution. It will be full of ultra compact processors with multi-terabytes of memory on-board. It will have insane speed, and will make sticks of RAM obsolete, hard-drives will be ultra tiny SSDs, processors will have little need for cooling. Video cards would go the way of the Dodo, since cpu's could also do this more efficiently. Basically, Moore's law should still be valid for a long time to come. (I guess now I'm a tech prophet, Art  )

It might one day be the norm to use ANNs with everything since computer power will no longer be an obstacle. Personally, I'm looking forward to all of this too, as long as its not abused. 

You know, for so many years programmers have been working hard to make there programs fast and efficient. They did this because they only had so few resources to work with. I suspect that programming efficiency will gradually disappear over time, seeing that constraints will no longer apply. However, coding will probably lean towards readability more than anything. Lots of times programmers have sacrificed good looking code over fast code, again this would no longer be necessary. Even Visual Studio has the ability to write code using only pseudo code.. I don't recall what this type of programming is actually called. Spydaz should remember though.

I don't mean to say I know any of this for sure, its just a guess. Other opinions are welcome. I'll try not to fall off my soapbox as I descend.

As for this recent cloud-based movement, there is a place for it in our society. Lots of people are non-technical, and cloud-based apps lean towards less headaches for users. But, as for serious PC users and developers, we will always need self-contained, non-web based units. Of course, these 'units' might be the size of today's cell phones.

I sure broke off on a tangent there  .

[Art]

Ah yes...the Great and Powerful Oz, uhh Snowman!! Please, pay no attention to the little man behind the curtain!
He's busy politely pontificating instead of , well, you know...other stuff!

Me? I've been busy shoveling snow from the walk, deck, driveways and two cars! This is getting taxing, tedious and too much!

[Snowman]

Hey!! I'm doing something Athena related... promise... today I've been going over some more English lessons and have been applying those rules to an NLP structure.. I have to be thorough, and so it becomes tedious, slow... boring  . It's much easier just to pontificate (nice word by the way). 

You know you could just call the NSA and say you found a terrorist cell operating from snow-forts in your neighborhood. It won't be long and they'll probably swoop in and fire-bomb the entire area.. no more shoveling snow. Problem Solved!!! Hmmm... I almost sense other potential problems with that solution..

    

[8pla.net]

Snowman said, "I don't mean to say I know any of this for sure, its just a guess. Other opinions are welcome."   Well, there is a version of .NET which I learned about in expensive trade school which requires no programmers at all.

[Art]

Hmm...Hey Snowy...I like that idea!! Let THEM do the "removal". Wow...that might work! Let me think on it before I commit (or get committed)!

[Don Patrick]

I like the video Ranch posted. I rarely (if ever) see evolving AI turn out to do what they were meant to do. I've seen robots do the same but they didn't get beyond the wobble & crawl stage, presumably because real life doesn't allow for 30000 tries in anything less than a year. Interesting to note is that the AI did not learn to walk but to "fall with style", which goes to illustrate that out of all possibilities, the chance that an evolving AI will end up with human-like characteristics is not a given.

Neural nets are still magic to me, I only get that they figure out statistical patterns from loads and loads of data/attempts. Now I wonder what happens if we placed one rock on that otherwise 99% flat terrain that the AI is trains on. Will it eventually change its global walking pattern to constantly brace for rocks, or would it develop a sub-pattern to overcome rocks? My guess is neither, as the rock is statistically insignificant, yet that means it would always fumble at the rock.

You can apply this scenario to NLP: Suppose the flat surface represents conventional word sequences, and the rock represents an idiom, or any other kind of inconsistent exception which we know language is full of. As a human child you are allowed to misinterpret and ask questions like "Why was the elephant wearing your pajamas?" and you'd probably get an explanation that it wasn't meant literally. That is a kind of learning that only requires one single instance of training, and allows for exceptions as well as conventions.
I do not agree with ignoring the pajama-wearing elephant example entirely though. That should only be the case if we assume that the AI is dumb to a fault, whereas the example served to illustrate a point which was not to be ignored. There is no rule or impossibility against programming AI to understand analogies, examples and the non-literal meanings of figurative speech. The trick is to recognise them when they do occur.

As for the grammar programming: If your mental stamina isn't up for prolonged tedious, you might want to look around for existing grammar parsers that meet your demands just for tagging adjectives and verbs and nouns and such. When I began there weren't none around, otherwise I'd most probably have used one.

[spydaz]

 Cloud based apps are good for artificial intelligence as a single mind could communicate to all it's clients and learning would be centralised ... At this moment it's all about price .... Window azure platform "looks good" yet it does cost ...

I'm not sure which code can be generated by the "codedom" engine in visual studio. But code generation ... I was thinking along the lines of building templates which could be adjusted to fit the circumstance ..

NLP: I have found that when using propositional logic .. (Inference engine ) that there are many more sentence types to distinguish ...

Ie : proposals, opinions, reports , expository, loosely associated statements, warnings, advice

Recognising the sentence type and choosing which action to perform accordingly is part of the decision making process ... And response generation ...

ANN: for looking at Anns and regression I use SAS / IBM spss / or weka .... I would like to know how to code these types of techniques ... I know it can be done in netbeans with Java ... Which as soon as I learnt java my brain just dropped it ... Although the techniques remain .... (Closed minded programmer) as vb is my tool ....

[Snowman]

My guess is neither, as the rock is statistically insignificant, yet that means it would always fumble at the rock.

Well, I would probably stumble at the rock too... The whole idea of ANN is that it is founded upon the human mind. Technically, you know about the NN just as much as anyone else does. You have a BNN (Biological Neural Network) . So, my question is, would you stumble on the rock? If so, then an AI would too. Sometimes I trip, and look back to find nothing at all   . Sometimes it sucks to be a BNN.

I think that as the human mind is developing, the BNN makes room for quicker learning. We already know that the human mind has its own Operating system. Mine is called Aaron, yours is Don. We know that this Operating System abides within the Network. Therefore, it would make sense that there would be certain sub-networks that are dedicated for quick learning. In fact, our NLP makes for highly developed learning. A fully developed ANN should develop a complete NPL over time... after millions and millions of iterations. (woo!)

If your mental stamina isn't up for prolonged tedious, you might want to look around for existing grammar parsers

I'm going the long way round. I'm doing this because I need to understand the NLP fully. I do this because I know that most likely I will create something unique and imaginative. Also, no one should be able to legally lay claim to my work, whereas, I would have to pay others for their work. On top of all that, I actually enjoy this way too much, even though I wear myself out. Thanks for the advice though. At least your looking out for me.

Spydaz, thanks for that post. That was very informative. I'll probably refer back to it while doing research. I've been absorbing details about ANN in hopes of understanding it enough to not only build one, but make one that fits my needs perfectly. It would be ideal if somehow a perfect NLP could be written within an ANN. (not saying it should be done, but could be)

There are many different types of  NNs and there are many types of learning algorithms, some slower than others. An ANN needs to be carefully built with everything in mind. Either way, I need to get most of the English language grammar rules fresh on my mind to properly design things. Most of the time a program is firstly visualized, before it ever gets programmed.

In three days, I will be 35. 
The world will end and someone will get terrible indigestion.

For some reason that made me think of a movie quote.

"One day there will be a boy named mistake, born to Ruth and Loyd Worrel, and he will catch mumps and measles again, and catch on fire, and fall off the edge of the world, and mash his finger, and die before he is twelve. Nighty night mistake! "

The quote is at 6:30.

[ranch vermin]

I just had this new idea.
What if the computer has to match the intent and cause to the speech of the user.

so  USER->     "hello computer, is there anything complicated about speaking at all?"

this is actually the clue to the query which the computer carries out to the CAUSE & INTENT of the speech

OC,  when it was training,  it would have to learn off an explicit cause & intent dataset,  because the way computers interpret symbols is very explicit, and is exactly like programming, the only difference from programming would be special algorythms which make it explicit, from gappy/missing information, confusing order, symbolic meaning & simalee phrasing.  (its not easy is it.)

If you wanted to give yourself a break,  you just get rid of simalees, and just do it string matching.


Then we take the answer of this query->
CAUSE&INTENT OF USER    "user is disgruntled with simplicity"&"asking the solution of the problem"

and put it in another query to find the computers intent->   (another NLP database of a very specific dataset.)

COMPUTERS INTENT-> "answer with solution to problem and bring him up because hes pissed off about life"


Then one final query to find the exact written text.   (the final NLP extracted database.)

COMPUTER-> "Cheer up buddy, look at the bright side, at least you dont have to run text out of a chain database."


Why would this be better?  its a data conversion tactic, the computer predicts things about what you said, before it answers.
I bet it would be simple sounding in the end product,  but it could be interesting if it works at all.

[Don Patrick]

would you stumble on the rock?

If I were as blind as the bouncing AI in the video, then yes I would trip over it. But I would likely quickly develop a subroutine to overcome the obstacle from that singular encounter, whereas it is an established fact that neural nets require bucketloads of encounters a.k.a. training. I don't subscribe to the popular notion that neural nets work just like the brain. They operate by simplified neurons, yes, and bricks are made of molecules, but those same molecules don't necessarily form a house.

I'm going the long way round. I'm doing this because I need to understand the NLP fully. I do this because I know that most likely I will create something unique and imaginative. Also, no one should be able to legally lay claim to my work, whereas, I would have to pay others for their work. On top of all that, I actually enjoy this way too much

Yeah, I get that (most grammar parsers are open source though). It's just that after I went through that whole process, it turned out (for me anyway) that most of the standard grammatical "part-of-speech tagging" holds up, it's what you do with the tagged elements after that where things get inventive. On the other hand, maybe it is necessary to work on language for a long time before one gains enough insight, and I have been coming to a point where I wonder to what extent I still need the grammatical phase. Good luck with it.

[Snowman]

I don't know about you, but I learn like that stick figure video. I go over things over and over in my head until I get it down. I test and re-test and check it against the results. Sometimes it takes me ages to learn something. Case in point: Every Sunday I show up for church. I set a book down at my feet because I have few choices where I can put it. (Its my personal song book.) And just about every time I get up I will trip over that stupid song book. People laugh at me, but for the life of me, I can never remember there's a song book at my feet. Most of the time, if something isn't natural for me, it takes years for me to learn. I can play the guitar which takes a lot of skill, I can read an article about Relativity and virtually memorize it, but I can barely remember people's names, even if I here it said everyday for months on end. The speed of light is 186,282.396 mps. I learnt that in 5th grade when I seen it in an encyclopedia. If I was walking on an empty field which occasionally had a stone in it, my toes would get mighty sore.

We have the benefit of an imagination. Most of the time our brain carries out scenarios, and mostly we are unaware of it until the brain is ready to show us it's conclusion. It is our Eureka moment. Suddenly we understand something, but perhaps we don't know how we figured it out. But if we had access to the lower levels of our brain then we could see how we came to that conclusion. I've spent years trying to get to the bottom of why I think the way I do. I've learned to pay close attention to my thinking processes. I've learned that most of my brain power is focused on trying to find new knowledge out of old knowledge, and every new thing I see I have to carefully absorb as to not come to an erroneous conclusion. I've also learnt that people typically will come to conclusions far to quickly and therefore end up believing something strange and usually very wrong. In my opinion, If you don't know something, then either speculate or just keep quiet. I actually learned that from the Book of Job.

I found that program that ranch's video was showcasing. I'm playing with it now. I think the documentation said something about how the input is connected to the output. I guess that would create a loop, testing every scenario, with the knowledge of passed fails, until the task is finally done. Hmmm, sounds like my Athena programming techniques

Download at your own risk.
https://dl.dropboxusercontent.com/u/51621154/____LEARN_2Dsticks_V12.zip
https://dl.dropboxusercontent.com/u/51621154/____LEARN_2Dsticks_NEAT.zip   'Experimental Version

[Art]

<...Also, no one should be able to legally lay claim to my work, whereas, I would have to pay others for their work....>

That is why we have Attorneys, lawsuits, Patents and copyrights (depending on your product / creation).
Of course you would pay others for their work. You own / drive a car don't you? Yours but you need a license to drive it. You purchased a software? NO. You purchased a license to operate that software. Ownership remains in the hand of the creator regarding software, for the most part and not particularly applicable to Open Source which is released into the wilds of population. You are still the originator / creator.

I pay a roofer or plumber for their services. Programmers get paid for their services although for some who program for others or other companies, their efforts are proprietary and instantly become property of the company for which they are employed.

Just some thoughts....
Lots of turns and twists as we go further down the rabbit hole....

[Snowman]

What we want to happen and what does happen are two different things.

Consider the following scenario.

You create some software, but you are poor and have very little notoriety. But, somehow you manage to patent it anyway. 
Someone rich wants your software.
They steal your software and start selling it.
You can't sue them because you are poor.
You try to sell your own software to others.
They sue you, even though it is your software.
You get thrown in jail because you loose lawsuit.
You meet a felon named Tickle who isn't very funny.
You become his "special" friend.
The End.

Consider Scenario 2
You use someone else's software in your program, but you are poor.
They see your program is successful so they want a part of the profits. 
You say, ok. So they now get 10% of sales. 
They say, "I want 20%" 
You say, No. 
They sue. 
You can't fight them because you are poor. 
You go to jail and get reacquainted with Tickle. 
The End

I have to admit Art, money is not my forte.
I was told once by an officer that I would never survive in jail.

[squarebear]

Which is why I include Easter Eggs in anything I write which displays my name and shows my ownership.

[Snowman]

Yes, it's pretty tricky business if your aren't as rich as they are. It's best to first be invisible, work with your own hands, and only present your product publicly when everything is mature enough.