[Art]

I have read that book and many others of a similar nature. While I do get the gist of the author's message, I'm not sure if it will be happening in the very near future, as described in the book. Many people do bond with a lot of non-organic items. Some have named their cars, computers, guitars, etc. Most boats have a (sometimes female) name on the stern. I also do not think that such bonding is always of a sexual nature either. It is of a different fulfillment or satisfaction or joy that comes from using that item (guitar, car, boat, etc.).

These android / robot sensuality books should never replace human compassion and love but I dare say that it sill exert it's influence in some circles..much like adult venues have proliferated in the past.

To form a "mutual" attachment or bonding between a robot / android and a human would certainly require (or one would think so), a degree of sentience or advanced AI within the bot or droid along with a suitable range of emotional constructs to emulate human behavior.

This is sort of getting askew of Ultron7's posting about forming a "Bond" with various items but he did initially refer to a robot so I'm safe.

Whether such robots can "evolve" to form some rudimentary emotions would largely depend upon the programming. If the programming wasn't too specific and allowed the bot to exercise a good degree of freedom to explore and learn on its own, then that might lead to a completely different outcome.

Those interested in such topics as the above paragraph should read the Free eBook from a talented Sci-Fi author, Mike Deering, entitled: Janice - Robot Maid. (why did his robot act differently than other ones from the same factory? What did he do or not do that influenced the outcome?) Read!
http://www.feedbooks.com/userbook/11578/janice-robot-maid

[Ultron]

This is sort of getting askew of Ultron7's posting about forming a "Bond" with various items but he did initially refer to a robot so I'm safe.

LOL well its not like I'm going to tear you apart...
http://imageserver.moviepilot.com/-4db72405-26c2-4f66-9d07-0763249b368b.jpg?width=1440&height=595
...from the inside...
http://imageserver.moviepilot.com/picture-2-i-m-gonna-tear-you-apart-from-the-inside-the-new-age-of-ultron-trailer-is-here-and-it-s-badass.jpg?width=1440&height=598
(Don't you just love those Marvel references? )


Jokes aside, I was actually discussing about our stance towards intelligent life forms - the 'bonds' might have been a bad example or I may have put it the wrong way. The initial discussion was what is ethical to do to robots and what is not. This does not mean the robot has to form a bond with us, but no matter how advanced they are, they are in fact going to one day become the predecessors to an intelligent race.
I believe the most valuable thing I could say right now is that it does not matter who or what is in the shoes of these robots, yet the actions we take against this entity. Again, our actions are a reflection of who we are.

If I must, I shall write an example... What does a kid who bullies others tell you about his character? Or yet, what does a man who brakes glasses in a bar tell you bout his? It does not matter what the subject is, it matters that we are the action takers and that we are described by the verb.

Whether such robots can "evolve" to form some rudimentary emotions would largely depend upon the programming. If the programming wasn't too specific and allowed the bot to exercise a good degree of freedom to explore and learn on its own, then that might lead to a completely different outcome.

I do not believe they could achieve this with classical programming and hardware. For one, the software would need to be dynamic - that means change according to the needs (much like our mind 'edits' its neural pathways and creates new cells). This would probably mean self-reprogramming hardware or to somehow create a new language that has a fixed structure which takes values from a table (values being some sort of instructions) which could be easily edited by another program monitoring the needs of the machine and finds better ways to execute certain jobs (as a background program / processor).
Or the alternative is to create some sort of dynamic (organic?) hardware like in the movie Automata, where the robots possessed something called a bio-kernel which could evolve and re-structure itself. In this case the software might not have to be dynamic but it would still be hard to make a program that can control something that is not constant and ever-changing (Do we need dynamic drivers?).


I guess I am trying to say that with current technology, a limited robot could in no way grow over it's original limitations. If we program it to be free and learn - yes, but an RC car would never turn on you.

[ranch vermin]

I guess a robot to me,  an actual makeable one, could be a model of someone, this model is more external than internal,  and is only superficial compared to person that it was modelling.   But depending on how well they predicted the persons actions with machine learning, the more internal of this external representation would come through.

Sorta like this ->

Except its also motor cortex reading, as well as the eye,   then you could form a model of the person,  but its not that dangerous, as if it were a real copy,  its just a simple version of the real thing.

This way, you could go into a magnetic resonance imaging device,  then go into a robot that has cloned you, (It took a fair a bit training - or calibration - to make this clone)  and then you can then inhabit and explore/move inside a virtual world, whilst sitting on a comfortable couch.

[Ultron]

We are decades away from such feats. We don't understand the brain yet and exactly how it functions. This research is interesting however at this time it is pointless since they are skipping a few important steps. For one, how does our memory work? You are going to make a machine that can memorize our dreams but you don't even know how we do it?

For one, we might only be remembering certain parts of the image we find necessary or unique to it, and later (when needed) reconstruct it part by part, and those parts very previously memorized and not introduced or recorded by or from the latest image we saw. I cannot really find an easy example for this but say you were tasked to draw a robot - you remember a pattern you have associated with it (let's say, as not to open another topic, a generalized pattern) which tells you it consists of motors, metal plates and shaped like a human body. You first start drawing the humanoid shape, then on the joints you add motors (whose exact look you find in your 'database') and so on.

It would be memory inefficient to remember all details about every image, when you can just remember a pattern and further associate it with other images, thus saving space on your 'hard disk'.

This is mainly theory I guess (which as usual I strongly believe in) but the point is that we are not sure how we function therefore I find projects such as these a waste of time - when you build a house you start from the foundation, not the roof!

[ranch vermin]

Ultron!  listen to me!

Machine learning works without having an understanding of what its doing.  (do you think your computer understands what its doing?)   I said the model is superficial and lacks indepth likeness,  but on the surface, it does actually work a bit,  and its prooven that it works.   (Did the predicted pictures look similar?)

I wouldnt stress how important machine learning is, if it didnt work.  It works,  and its actually the best way to go about forming a model of something, you dont actually understand, but youd like to mimic its behaviours.

The best motion capture system in games, (like GTA V, looks great doesnt it.)  is taking the uninteractive skeletal record, and then making it interactive by finding correlations for the movement. (not the cause, the cause is internal! and impossible to find.)  but with external correlations you can get something similar, as long as you layed the blame on something closer to the truth, but nowhere near it. 

Then youll be able to bring back the mocap animation, in a more interactive order,  than just the initial frames you taught it with.

The first autonomous vehicle. (I think it was a white van)  was interactively backpropped in the 80s!!!    You need to catch up to things man.     If you want to be one of the mugs involved, you have to do your homework.

And if its decades away, its frighteningly close to now.

[8pla.net]

Ultron7 said, "I cannot really find an easy example for this..."

I can.  You are referring to: Closure.

Example: http://www.ministryoftesting.com/wp-content/uploads/2014/11/closurelaw-sm.jpg

[Ultron]

I am not sure myself, 8planet. There may or may not be a connection between what I was talking about and that picture - my head is currently in a state of internal mess so I can both find contradictory and accepting ideas which link or divide these two things.
This is probably because I am not sure I can explain to myself how I identify a triangle within that image and because I don't entirely understand my own concept I previously brought up.

This is complex and most definitely mind-blowing stuff, but it helps make a nice and useful discussion.

Regarding machine learning, I understood you ranch. I just forget half of my thoughts before finishing a reply. I do not like machine learning, and I will explain why with an example...
I have previously mentioned I am a member a Bujinkan (use google please), and I have gained a lot of inspiration from my training. One thing I can relate with this topic is the method of learning techniques - beginners are told to try to mimic the teacher as closely as possible, in order to achieve better results and a better technical understanding of the exercise. But you must also learn the 'spiritual' or psychological aspect of the kata - for example you must understand when it is used best, how to look at the opponent and how to make them feel prior to the attack.

So what I am saying is just mimicking human behavior doesn't make a robot human - and achieving this (or even surpassing humans) is the ultimate goal of A.I. right? We won't learn anything from machines who mimic us - they also need to understand what they are doing something and why, so they can later at least improve it or use this knowledge to teach others.

All my research so far has been focused on understanding how we understand - 'knowing what it means to know' and how to implement this into artificial creatures. So I have neglected the possibility of using such methods.

[Art]

When a machine / robot equipped with a suitable AGI and was capable of machine learning, it can then learn how to adapt, evolve, better itself.

There have been many machines / computers that have been capable of creating better machines than themselves. They realized their weaknesses and decided to fix them. Kind of like a person studying extra math courses in order to pass a certain level test in order to become better, more capable.

In order to become self learning there must be understanding...knowing what you know you don't know...then learning it. It's a twisted path to enlightenment even for a budding robot!

[ranch vermin]

Of course mimicing isnt the real thing,  its like a little toy corvette compared to a real one,  then a supercharged one.

I dont think even making the real corvette is easy,  or even possible by myself, let alone the supercharged one. (which would be the singularity.)  But the little toy corvette, if you can make it, will revolutionize everything already.

Thats why im trying to build a toy corvette,  not the real thing.   And you can show off your fancy algorythms and all the kids want to come over to play with your invention already, because its a step up from what we have now.

Dont feel let down by that,  its great fun, and general automation, in my mind, isnt the same thing as general intelligence,  like I see elsewhere, but they are all talk and no walk.  But general automation hasnt happened yet either..  and will serve for all the exciting things to come out in the next 10 years.   Not intelligence.

[Ultron]

But the little toy corvette, if you can make it, will revolutionize everything already.

I agree, ranch. I have not once underestimated the difficulty of making a machine that can accurately mimic human behavior. And you are right, that is already a step up the stairs. But you can turn a computer with average peripherals into a human mimic just like you can't turn your phone into a laptop. To mimic human behavior it would need to have so many sensors (existing and non-existent) so it can at least 'mimic' our behavior when hungry, thirsty, hurt...


If we are to create a copy of an organism's 'intelligence' or mind we must first replicate the hardware it normally controls. This could be theoretically done by simulations (virtual body parts) controlled by a program, I suppose.


I support and am often amazed by such research, except it feels wrong when it is referred to as AGI or intelligence at all. I believe these programs and systems would be more accurately referred to as Artificial Intelligence Mimic (AIM), as it is only mimicking intelligence and is not 'intelligent' on is own.
This is also caused by lacking a universally accepted definition of interconnected terms - I for one disagree with the 'official' definitions because they put futuristic androids and roombas on the same level.


My personal goal is a real marionette as you called them. I believe it holds much more potential (we do agree) but I chose this mainly because nobody else is aiming for the same level and I also believe that we can skip right to it.


And it is also worth mentioning that we might be starting to go slightly off topic - the original discussion was about the ethics of robotics as well as how they will become a part of our society and everyday lives. Cheers to the good discussions!