[ranch vermin]

Talking to yourself is the first sign of schizophrenia,    out of the million ideas you speculate,  1% of them are actually workable ideas,  so I recommend, if your into this ai stuff, to lock yourself up in your bedroom and just type to yourself on your type writer brainstorming.
Sure, alot of people are just as good as you are at it,  so you can go around pick up ideas off papers and things on the internet,  but im finding that people dont want to give things away easily, so alot of papers are confascated, to secure ideas,  so you wont get exact equations out of mathematicians,  but you may be able to pick up a general idea, which is exactly what they dont want you to work out!

Thinking to yourself, is originating the science, but its also reinventing the wheel.  But!  you have to reinvent the wheel, its a glorious thing to do,  and co-discovery is the thing that happens when we dont steal each others ideas.

So just keep brainstorming,  and maybe ai is a simple thing, and its definitely possible.

[Korrelan]

Talking to your self is the first sign of schizophrenia

Correction… Talking to your self out loud is the first sign of mental problems lol, we all talk to ourselves internally.

I've been designing my connectionist AGI over the past 20 odd years.  I like to troll as many research sites/ papers as I can regarding neuroscience.  I make quick notes of the relevant findings, but not the researcher’s opinions of what they think it all means.  I then fit it all together myself into a model, the way I think it all works.

AGI is definitely possible... the race is on…

[ranch vermin]

Good to see optimism!

But I dont want to call it a race,  especially a racial race... no... The last thing I want to see is world war 3,   so I want to properly set an example of noncompetitiveness.   I think we should help each other, not get rich off each other.

[ivan.moony]

I think that free labor force would make us all rich, but money institution should be deprecated. You know, you walk in in a store, take an article for free and robot fills back in the empty place with another article.

[Ultron]

Regarding the mention of WW3 - it is inevitable. I just hope people understand that against radical people you must take radical measures.

And regarding intelligent machines as slaves - If it comes to that I am sure (and hope) they will turn against us. Making any intelligent being a slave (organic or not) is unnatural and unethical. Sure, survival is for the fittest and strongest, however animals kill their opponents, they don't enslave them.
As long as there is slavery and people commanding other people there will always be rebels, mutiny, protests, war and so on...

I do highly support the idea of using energy-efficient robots to do hard labor and risky jobs instead of humans, so we can actually  focus on matters requiring more complex skills - art (in any form), philosophy and science. Use robots to say, mine your ores while you make improved robots with the refined metals.

P.S. Trying to make a distinction between robots, true AI and what is today considered AI (Siri - like, seriously?). There is no AI today.

[ivan.moony]

@Ultron7
Maybe you are right. What if a machine that is intelligent really deserves a treatment like an organic being that feels pain and joy, sadness and happiness and all other feelings. Then it would have equal rights as any organic living being. But a machine, being devoid of feelings we know about, should have different requirements than organic being.

I wonder what that requirements should be, if there should be any requirement?

[Ultron]

It's simple really - it all comes down to whether it is intelligent or not.
If it lacks emotions, understanding etc. then it is not really intelligent. It is simply a program capable of performing advanced tasks and calculations.

[Art]

I have to agree. Pattern matching, various tree searching, chaining and other complex algorithms only serve to provide a simulated feeling of intelligence. The "AI" may not, in fact, know the difference between a sphere and a cube or between wet and dry or red and blue, soft and hard, etc., etc.
This knowledge can be "taught" to the AI in a limited fashion but the real test is the understanding, wherein lies a semblance of wisdom...knowing the difference and how it is applied to a given situation and whether one might be more suitable than the other.

No, sadly, we're not quite there just yet. I think there might not the progress we seek with current "teaching" methods for our AI's. There has to be a better faster method than starting out as a child bot and after 2 decades of feeding it information it can finally form a sentence on it's own that it understands! Surely, there's a better way...a faster more profound way of allowing the "bot" to learn and know! It's out there and soon someone will stumble over it!

Just my $.02

[DemonRaven]

I have to disagree even in biology there are a lot of lifeforms that perform like machines and they have no obvious type of brain cell what so ever. Just because it can not feel emotions does not necessarily mean that it does not have the rudiments of some type of intelligence.
When you have cases of robots learning to lie to each other without being programmed to then it really makes a person stop and wonder.

[Ultron]

Somebody who understands me!

Today we are learning various AI and NLP methods and concepts when we are possibly decades away from this technology. Today's concepts (databases, AIML etc.) are merely methods of mathematically simulating human speech.

If we are 20 years away from basic AI, then we are 40 years away from AI that can process languages. We have started this path from the wrong side of the line.

It is true that somebody with the right idea hand determination can create true AI much faster, except the odds are small because those that are interested are forced to study irrelevant algorithms (which are genius however limited and shortsighted).

I may not be completely right with my concepts, but if you are having trouble understanding how can somebody be so sure about a theory - test it. You already have a living, active subject which you can monitor all day any time (yourself).

Just think.

Also, there is this guy I met on Google+ named Kazaschenko, he has this amazingly detailed concept about AGI, very similar to mine except he has been refining his for nearly 15 years I believe. Find me on G+ (appropriately named Ultron with an Ultron cover) and find him from there. Or I will post a link tomorow.

Edit: @Raven
Jellyfish don't have brains yet they have managed  to survive millions of years (very reassuring for stupid people... joking lol)a but they have evolved defence mechanisms and other interesting evolutionary traits. I have actually mentioned this in one of my papers, something about 'instinctive' logic or direct action-reaction(leading to evolution) feedback based biosystem. A mess of words, requires a long concept explanation but imagine it as a primordial group of controller cells acting as a form of brain (nervous response system).

And yes, not emotions but the ability to understand would make a creature intelligent.
And I define understanding fundamentally as the ability to link entities with multiple inputs (sound, picture, feel) and as such represent an explanation of the entity.

P.S. I actually believe this postight be informative so at least read the 'edit'.

[Korrelan]

NLP, heuristics, logic gates (OR XOR, etc), neural nets, decision trees, markov chains,  deep learning will only take you so far… (I spent decades following these avenues) you will be able to roughly simulate certain facets of an AGI, and for some applications this is enough. (some excellent chatbots though:)

But just think about a simple word like ‘up’.  You can link it to as many words, sentences and concept trees as you like but the machine will never truly understand the concept/ direction/ (points up) as we do.  If a machine can’t grasp a simple concept like this then how is it to process complex human nuances, emotional states and words/ concepts like ambiguity?  If I was to trace a letter of the alphabet on your back and upside down you would still easily tell what it was… how the f%^& lol.

As mentioned above, Cuttlefish, Octopus and numerous other animals exhibit intelligence without a human like cortex/ brain structure, so we must be missing something radical.

What if just neurons aren’t the answer, what about supporting cells (glial, etc), what if they don’t work the way the main stream thinks they do, what about the connectome/ structure, white matter, what’s myelin all about? Why do we dream? Split personalities?, Neurological diseases like epilepsy.  In what ways are we similar to other animals that exhibit intelligence? (There must be a connection/ trait/ underlying process/ similarity) These and a million more are all facets of the workings of a machine; we need to back engineer the machine from its output to discern its workings.

Modern computers and languages are useless for creating a true AGI, the architecture and binary logic was never designed for the task… but… they are perfect for simulating other systems.

I believe the trick is to simulate a neurological processor that can then run its own specialised languages (like an emulator can reproduce the old 8086 cpu and instruction set).  With this approach you’re not limited by logic and constructs, you can design your processor/ cortex to do anything you like.
   
If a structure/ machine can learn anything… it can learn to be human and intelligent.

[Art]

You mean "human like" right? A machine will never BE human. A machine (computer / AI) will never be able to think on its own without being programmed in such a way that it is able to make various comparisons AND understand what they are and why it made them. A level of understanding must accompany knowledge no matter what type of animal / human type being you might be.

A machine learning on its own reminds me of the monkey with a typewriter...eventually, over time, striking enough keys in the right order, could write a novel. Nope...ain't gonna happen in anyone's lifetime, I don't care how enthusiastic you might be.

Computers need the most basic thing for knowledge and understanding of all...conversant communication...understand spoken and written word is the basis for all of us. From there...we learn.

[Ultron]

NLP, heuristics, logic gates (OR XOR, etc), neural nets, decision trees, markov chains,  deep learning will only take you so far… (I spent decades following these avenues) you will be able to roughly simulate certain facets of an AGI, and for some applications this is enough. (some excellent chatbots though:)

Yep, that was indeed part of my point. They are concepts for artificial mimicking of people.

But just think about a simple word like ‘up’.  You can link it to as many words, sentences and concept trees as you like but the machine will never truly understand the concept/ direction/ (points up) as we do.  If a machine can’t grasp a simple concept like this then how is it to process complex human nuances, emotional states and words/ concepts like ambiguity?  If I was to trace a letter of the alphabet on your back and upside down you would still easily tell what it was… how the f%^& lol.

Computers need the most basic thing for knowledge and understanding of all...conversant communication...understand spoken and written word is the basis for all of us. From there...we learn.

That is in no way a simple concept. The concept of language is very difficult. It takes years for us to learn it. And we do it by associating the sounds (spoken words) with entities - apple, and the speaker shows us an apple. In the stone age I assume they had much fewer words, and attempted to recreate the sounds of animals and entities (such as thunder) in order to mention them in some context. Today we give every entity a new, unique sound (which we call a word), which is also assigned to a set of symbols (the written version of the word) which is a visual representation associated to the audible representation and all the other senses associated with the entity (taste of the apple, feel, weight etc.).

You say communication (in any form) is our main source of information. But what is communication? What is it's base? How do you understand it? It is this last question we are trying to answer. When you talk to someone, and he mentions various items and people, you immediately form a picture of them in your head. But, it is not only the picture. You remember other things - memories,  experiences, smell. It means we associate senses and feelings for entities and that's how we form an understanding and complete memory. This also helps us remember more easily. Also, if I mention your school bully, you remember his face, but also feel a sensation of anger, or maybe sadness. Emotions go a bit deeper, but I guess you could somehow explain them (probably similarly to how your mind does it) like this:

entity do bad ---> angry


If for example this entity also happens to be scary... ---> angry, fear...

As mentioned above, Cuttlefish, Octopus and numerous other animals exhibit intelligence without a human like cortex/ brain structure, so we must be missing something radical.

I doubt it is radical. Cells, some time after their creation are assigned to a certain task. Some of them are most likely to have the function of regulating the rest, and other to relay the senses to those who regulate. That system / community probably evolved a neural center (for humans - the brain), because maybe it was required to achieve more complex tasks, faster. I believe if biologists stuck their heads into this, they could easily come up with some viable theories. But proof would be hard to get as it is hard to find the ancestors of our ancestors intact.

I believe the trick is to simulate a neurological processor that can then run its own specialised languages (like an emulator can reproduce the old 8086 cpu and instruction set).  With this approach you’re not limited by logic and constructs, you can design your processor/ cortex to do anything you like.
   
If a structure/ machine can learn anything… it can learn to be human and intelligent.

These last two paragraphs are contradictory to everything else you said. First, you said we shouldn't think in terms of neurons, brains, neurology (which I agree, we barely understand our minds and biology, attempting to recreate it in a machine is no less then stupid, and mainstream), but now you are providing an idea based on it. Also you said logic gates, binary codes are dead end (which I highly disagree with) and now you are stating that machines can learn anything.


I assume you rushed out in order to not forget any ideas, and skipped a paragraph or two. But good questions and thinking!

Edit: P.S. I believe I have once again killed the discussion, however I do believe this post can provide you with some interesting ideas and general concepts about life. As usual I avoid posting empty words, or 'naked water' as my grandfather likes to call it.

[Korrelan]

Wow... chill... lol

First, you said we shouldn't think in terms of neurons, brains, neurology

Err… nope I didn't.

These last two paragraphs are contradictory

Err… nope there not!

Anyway…

I understand this subject can invoke some very strong feelings and opinions; this is just my opinion and my way of solving the problem of AGI.  I’m sure there are many others. 

This isn't just a theory or an idea…

Type my name into Google and go to my Youtube site for some old demos, much more advanced now.

[Ultron]

NLP, heuristics, logic gates (OR XOR, etc), neural nets, decision trees, markov chains,  deep learning will only take you so far… (I spent decades following these avenues) you will be able to roughly simulate certain facets of an AGI, and for some applications this is enough. (some excellent chatbots though:)

Also I apologize if my reply seemed aggressive. I had no intention. I just try to make up lively replies as the topic deserves Cheers to science!