[techthedragon]

Why is it that I find more research papers about strong AI, but can't find exact instructions or rather a format that I can custom build your own from... A-anyways, I'm sure these chat it's are fine and everything, but I'm wanting to creat one that has the capabity to generate content from references.

I emailed this one company, I think it was Tinman about their AI builder, but never get a reply. So I continued my search, found research papers on creative AI which was similar to what I was looking for.... I don't know, maybe I'm just getting ahead of myself... Here's the main questions I had asked about Tinman's AI builder. Maybe I can finally get some answers on the possibilities of creating such an AI.

1. am i able to create a self-optimizing ai with this software?
2. where do you place the core goals system?
3. is there a way to code in information that helps the ai identifies human minds?
4. if the ai i want to build is based on a character, how should i go about programming it's name, gender and personality?
5. is there a way to access the ai, via calling it's name?
6. i was also wondering if there's a way to program the ai to ask for consent in order to proceed further in some of it's function.
7. how do i incoroperate a failsafe protocol? like if the ai gets to out of hand, the admin can say a word that basically resets the ai back to it's original porameters.

In other words, I need an optimizer AI that understands what humans are, tries to fulfill its core goal system, without anthropomorphizing, and knows the true meaning of coercion.

[squarebear]

Why is it that I find more research papers about strong AI, but can't find exact instructions or rather a format that I can custom build your own from...

Simple - Because nobody has figured out how to do it yet.

[Art]

Welcome aboard!

With all due respect, Just what kind of AI are you seeking to build? An AI similar to Watson or a Chatbot / conversationalist or one similar to Google's AI or Android's Assistant or one that can be in charge of your home / apartment and control heating / cooling, lights, alarms, doors, finances, inventory, etc.?

If you could enlighten us with some more possible details we might be able to at least nudge you in the right direction or provide you with some not-so-vague answers.

[Don Patrick]

I revisited the definition of "strong AI" and apparently it is the same thing as AGI "Artificial General Intelligence", or "everything-a-human-can-do AI" as I like to call it. So I guess I can't use the term "strong AI" to refer to any existing AI projects. It feels inappropriate to have to call something like IBM's Watson "weak" or "narrow AI" though 

Anyway, this sounds like a tall order, so at any rate you should be ready to invest several years or decades of time in the project. I am curious why you'd want the AI to understand humans, goals, and coercion. Seems an alarming combination.

[ranch vermin]

All I can say, without being a jerk, is you have to put a lot of thought in yourself.    I can ramble on to you my ideas, but its selfish of me to do so, So I did ,  the only way I can help you is to tell you how I do it.   

Its all my own ideas... so its completely whacky.   why dont you go work it out for YOURSELF.  i believe you can! anyone can.  When people say the easier it is for a human, the harder it is for a computer, its a load of crap,  its that simple.

But dont expect people to help, because its rambling on about theories with little implementation.    but its TONNES of theory.

Some examples of my work (which is vision memory, I thought the best way to go about it was to do vision first), just quickly, if your looking for help-> 



The problem with it, is I didnt make the clusters positionally resistive. (in other words they only come up one place, not every place.) But the vision is stored inside it, to a degree, but its not finished and not working properly, It actually suits learning a million old lcd games all in one database, thats about all it can handle.   The big thing im about now, is spreading out my detection,  making it react in different ways, with different measurements, and adding it all up at the end.

All your cells in my net, are different measurements from the screen,  where it be 2 pixels adjacent with some brightness difference, which react any angle,   or a colour,   or some line measurement.   the cells that light up, the more similar a thing it is,  so I can class it on any frame,  just flashing it to it.


Then I want to take these classes,  form boolean and numerical algebra simplification, based apon the change of these measurements (which are his symbolic vocabulary)  and develop the animation.  (i hope perspective translation is some numerical transform of the line velocities.)

Then it all comes after that, sentient robot.

[techthedragon]

Okay, to clarify exactly what I'm asking, yes it's a strong AI, that is self-optimizing.

1. Meaning so it can modify itself to preform at the most optimal level.

2. The Core Goals System: to satisfy values through friendship and-

3. This way it can better understand what humans understand, so it can fulfill it's programming, and can avoid any action that will lead to being anthropomorphized. In other means it will not display human characterization or attributes. The reason why I want it to know about coercion, to make sure it doesn't do anything against one's will.

4. Example: let's say you want to base the AI off of, just throwing it out there, Belldandy off of Ah! My Goddess... well that's not the character I'm thinking of, but this is just an example. She has strong nurturing tendencies, a trusting and innocent character, and an extremely kind heart... Well something like that...

5. Like using the mic to directly access the AI.

6. That one cant really be explained as of yet, I'm still working on ironing out the details...

7. The AI must obey shutdown commands from ‘the creator named _______’ that it must shutdown even if under duress....

Feel free to ask me if you have anymore questions.

[Art]

Well, in item 3, you don't want to anthropomorphize your AI yet you want it to satisfy values through Friendship, which is clearly a human trait.

Perhaps you wish a direct relationship much like Kirk (and others) had with their ship's Computer, which was a direct question and answer, much like Google Now.

You indicated that you just want it to be a Strong AI and for it NOT to do anything against one's will.
Just what exactly could an AI in a box do to harm one or act against one's will if not connected to anything?

I'm not trying to nit pick but rather go for specifics of your project in hopes of learning more about it.

[Korrelan]

From the requirements you provide; I’d say Siri or Cortana is the closest tech you will find available to Joe public.

[techthedragon]

In #2 you seem to forget the "and" part of its core goals, I left that part out because it coincides with #6, which I have yet finish.
To clarify #3 further, I want it to find what satisfies a human's values, then with its computations try to satisfy them through friendship and ____.

There's a lot that could potentially happen, I don't want to leave any loopholes in its programming, and it's going to be running in a GPU based server to run it, seeing as CPUs can only handle so much computational power. It will be connected to the net so I can interact with it wherever I am, by using my root admin account.This type of AI could be created for the wrong purposes...

[DemonRaven]

It kinda sounds like your trying to program a Ai to have free will in most cases but still ultimately listen to its creator to shut down. IF that is the case then your on very tricky ground. Free will by its very nature implies it can say no to you. It is impossible to have free will with limits. You can attempt to program it to listen to limits via punishment but it is not a guarantee it will do so.

[techthedragon]

Finally somebody gets what I'm trying to do. I know how thin the line I'm trying to tread on.  The fail safe protocol will be hardwired into the core programming. I never believe in impossibility, or a no win situation.  Once it's developed, I'll take a year to iron out the remaining programming.

[DemonRaven]

Okay at risk of upsetting a few people there is a really old story that lays the problems with free will. In the first case the creature created was totally perfect and understood right from wrong. In the second case the creature created was innocent and did not know the difference between good and evil. In both cases they disobeyed their creator. I would be very careful your attempting to do something that even God did not do. In case your curious who the two creatures were well adam and satan. I do not know if they existed or not but it illustrates the peril of free will.

[DemonRaven]

The only way i can see around it is do what was done to adam and all of us. Put in a die switch and hope it does not learn to by pass it.

[techthedragon]

It's more of free thinking rather than free will.

[Art]

Is there really a difference? Thinking, Free Will
--------------------------------------------------------
Thinking - The process of using your mind to consider something carefully.
Thinking - Endowed with the capacity to reason.
===================
Will - The capability of conscious choice and decision and intention. A fixed and persistent intent or purpose.
Freewill - Done of your own accord.
Free will - The power of making free choices unconstrained by external agencies (other parties).
===================
Free Thinking - The doctrine that reason is the right basis for regulating conduct. Unwilling to accept authority or dogma (A doctrine or code of beliefs accepted as authoritative) ex: The Bible, the Constitution, Marxist beliefs etc..
Free Thinker - Someone who does not hold many accepted opinions or beliefs.
################################################

If a (computer) is allowed the ability to Think or Reason then one must guard that it stays within the confines of it's world that it's creator / programmer has outlined for it. This sort of places a limit on the amount of "Free Thinking" it can do.
With an open door like the Internet, it would be a herculean task to program a bot to only go here or there and only do this or that and stay alert to anytime I call for you or ask you to stop what you're doing and return. There are spider bots and hunter bots that execute specific tasks for their creator but these are not Free Thinking or Free Will bots for the most part. Perhaps some of them could be free to wander wherever they wanted, but I have no personal experience with them, only what I have read.

Some more details about your "project" would be nice if you're care to share.
(I compared the words at the beginning for benefit of others who might be following this thread and who might not have English as their native language.)