Self awareness

[tp6626]

When computers begin to pass the turin tests, what will they think they look like, if you ask them to describe their appearance to you. Another thread made me think of this. The one discussing members' faces as avatars, and putting a face to the name. I was pondering, what would an AI use in this case as an avatar, and, can it really know what it even looks like?

I'm stipulating that they have no cameras etc, and no one has told them where they reside (i.e. on what hardware etc).

Further to this, what does a born-blind person think he / she looks like? What perception do they have of their own form? Do they have a concept of vision?

I'm more interested in the former question on computers, but was quickly lead to ponder the 2nd area after a moments thought on it...

[Synamon]

That's a great question.

When I think about myself my first instinct is to describe my mind not my body. I sometimes feel like I am blind to the physical. The essence of me is my brain, how it thinks, feels, and processes information. So I'd probably describe myself the same way the computer would. Does that make me a computer? :thinking:

[tp6626]

You may be a computer. Would a computer post so much on so many fluff threads!?? Hmm, have you been bluffing us to pass off AI as a real person? ;

I was pondering further, and decided to let the computer use cameras if it wanted. How would it know what to point them at, and how to recognise itself in the images it gets. It would already have to have an idea of what to look for, and it probably wouldn't have this, so cameras would be irrelevant really.

When I think about myself my first instinct is to describe my mind not my body.Same here. I view my body more as a form of transport for getting my mind to meetings.

I don't really care that much for my body in the sense of vanity, and am not really bothered about being seen naked (which has surprised a few of my friends who know I'm quite a shy person to talk to). My view is its only a body; everyone has one, and they're all pretty similar. I'm much more worried about people seeing and abusing / mocking my naked mind, if that makes any sense!?

[just a user]

I was pondering, what would an AI use in this case as an avatar, and, can it really know what it even looks like?

...

Further to this, what does a born-blind person think he / she looks like? What perception do they have of their own form? Do they have a concept of vision?

I'm more interested in the former question on computers, but was quickly lead to ponder the 2nd area after a moments thought on it...


My brain seems to store a synaesthetic imprint of people, including me: it is a mixture of certain items:

smells (usually herbs, parfumes),
colors, and
(only sometimes) shapes or (strangely enough) landscapes that I have come to associate with that person.
There are no sounds, but my mind is generally not focused on auditory perception.


1 These items do not necessarily directly relate to that person, i.e. I do not remember somebody by the color of their favourite clothes or the way they smell. ;) (Although the reptile part of my brain might well do the lattter.)

I seem to have a personal coding system for these items which is impossible to describe because this system strange and highly intuitve to me. My brain seems to make impulsive associations about people and assign those items as it likes.


2 These imprints change over time; especially when I get to know somebody better.


Anyway, this is an amazing topic and I would like to hear some more thoughs!

Let met add that: I have no visual concept of my mind; I am absolutely clueless if I would have to describe it as 'something'.

[Synamon]

I was pondering further, and decided to let the computer use cameras if it wanted. How would it know what to point them at, and how to recognise itself in the images it gets. It would already have to have an idea of what to look for, and it probably wouldn't have this, so cameras would be irrelevant really.


I'm not sure a computer would turn the camera on itself, it might be more interested in it's environment than the 'shell' of itself. Wouldn't it matter more how it connects to the external universe than the box that it is in? Inputs and outputs would be all it knows.

[Mozzes]

What makes you think that these computers would think or act like humans?

[just a user]

...

I was pondering further, and decided to let the computer use cameras if it wanted. How would it know what to point them at, and how to recognise itself in the images it gets. It would already have to have an idea of what to look for, and it probably wouldn't have this, so cameras would be irrelevant really.


Would your computer/AI be able to take any 'recordings' of his "inner life" or are you thinking of a real world camera? In the latter case, I think, as you said, without any database about objects / instances of things, your AI is basically lost; or it has any algorithms on learning to classify new things. I don't know if something like that exists.

I would assume a computer would represent itself as some kind of geometrical/3D form made up of any kind of data that it has about itself.

But yeah, I think an AI might get more creative. ;)

[tp6626]

What I'm getting at, is the question, what does a computer AI think it is?

It takes a while for animals to realise that a reflection in a mirror is actually themselves. Some never do. Now, an AI isn't going to be literally looking in a mirror, but could it have some concept of what itself actually is!?

Is it going to start asking the question, what am I? Will it question its existence in the same way we do our own? Will it question our existence in the first place? How would it describe itself to us? What would it give you if you asked it for a picture of itself!?

[Jane]

How would it describe itself to us? What would it give you if you asked it for a picture of itself!?

As for now, if you asked for its picture, there's no way for it to hand it to you or describe itself to you unless you programmed it to do that. (Meaning computers do you?)

But they are becoming increasingly human. We are programmed too, just in a vastly different way. We are programmed with the things that allow us to... (argh, I'm running out of time, I'll continue next time).

But the main gist is that we're programmed with self-awareness in that junk of things that we're programmed with whereas those computers are currently not, unless some berk does so.

(Let's hope not.)

[PHS Philip]

What makes you think that these computers would think or act like humans?

If it can pass the Turing test, it's either thinking like a human or so good at imitating one that we're not going to be able to see the difference.

Will we have vain AIs who brag about how many processors they have? "Hey babe, check out my CPUs."

[just a user]

If it can pass the Turing test, it's either thinking like a human or so good at imitating one that we're not going to be able to see the difference.

Will we have vain AIs who brag about how many processors they have? "Hey babe, check out my CPUs."

"Mine is bigger than yours." :laugh:

[Ligda]

self-awareness is simply the realization that you are separate from the environment. the ability to vision has little to do with this.

i don't know what constitutes this, or what an entity needs to have this ability. some people believe that a soul is reqired for self-awareness. or not... i suppose the majority agrees that infants have a soul but they are not self-aware. i believe that intelligence is only one aspect of self-awareness. perhaps one day humans can create a computer that contains this ability... but first we need to identify this, something we have yet to do.

to know and have the realization that you are a separate individual with seperate thoughts and beliefs is one of the few profound, fundamental essences of the human condition. to know what is necessary for this is one of the greatest questions of life. i'll leave this question to the ontologists.

[PHS Philip]

to know and have the realization that you are a separate individual with seperate thoughts and beliefs is one of the few profound, fundamental essences of the human condition. to know what is necessary for this is one of the greatest questions of life. i'll leave this question to the ontologists.

Well, there are other animals that are self aware, too (most mammals, especially whales, dolphins, and primates, all of which seem to be very intelligent. The most intelligent, and the ones that would probably come closest to human intelligence are elephants, dolphins, whales, and the great apes. Cephalopods may be self aware and are certainly intelligent, but they're so alien that it's hard to tell just how intelligent they are)

[Ligda]

i didn't mean to imply that self-awareness is exclusive to humans, although it is interesting to compare the species that have that realization. just saying that it's probably more than just intelligence that is required for it.

although i didn't know that octopi might be self-aware... interesting.

[lisakki]

I'm guessing that at the basic level, they're going to see only what they're used to seeing, which is streams of information. The dry answer is, they'd probably just see themselves as their most fundamental pieces of data.

[tp6626]

I'm guessing that at the basic level, they're going to see only what they're used to seeing, which is streams of information. The dry answer is, they'd probably just see themselves as their most fundamental pieces of data.

The premise though, is, that the AI would be trying to pass the Turing Test (To view links or images in this forum your post count must be 2 or greater. You currently have 0 posts.).

Now, on this premise, the computer would foreseeably be pretending to be human, just in order to pass the test.

To do this, it would have to create a persona, and possibly provide an avatar or photos on request, which fit in with this character that it has created. The next step would be to pit two computers against a third computer, and task the third with identifying which of them, if any, are human. Each computer would need to be highly intelligent, but probably also have a high degree of self-awareness.

I'm not considering the type of computer technology we have now, where they do exactly what we program them to do. I'm imagining a time when they will be modifying their own programming, and are able to evolve of their own accord.

Given this, and their ability to perfectly mimic humans, wouldn't they begin to question what they are? Where they are? What they're here for?

Would they view themselves as formless spirits? Transcendent beings?

...my god, there are so many questions arising in my head on this topic, that I'm having to really try to keep some focus in my posts! :)

[PHS Philip]

The premise though, is, that the AI would be trying to pass the Turing Test (To view links or images in this forum your post count must be 2 or greater. You currently have 0 posts.).

Now, on this premise, the computer would foreseeably be pretending to be human, just in order to pass the test.

To do this, it would have to create a persona, and possibly provide an avatar or photos on request, which fit in with this character that it has created. The next step would be to pit two computers against a third computer, and task the third with identifying which of them, if any, are human. Each computer would need to be highly intelligent, but probably also have a high degree of self-awareness.

I'm not considering the type of computer technology we have now, where they do exactly what we program them to do. I'm imagining a time when they will be modifying their own programming, and are able to evolve of their own accord.

Given this, and their ability to perfectly mimic humans, wouldn't they begin to question what they are? Where they are? What they're here for?

Would they view themselves as formless spirits? Transcendent beings?

...my god, there are so many questions arising in my head on this topic, that I'm having to really try to keep some focus in my posts! :)

Computers that can both turn themselves into technological singularities by self modification AND that think of themselves as (depending on how humanlike they are) gods? That's just a tad scary.

[tp6626]

Computers that can both turn themselves into technological singularities by self modification AND that think of themselves as (depending on how humanlike they are) gods? That's just a tad scary.

Yep.

I was even wondering whether there even would be many of them. They may find it more advantageous to act as 1! They would be relatively boundary-less and difficult to define as separate entities.

This is indeed worrying, and implies to me that as soon as AI becomes self-aware, in the sense I'm talking, traditional lifeforms will become redundant.

(This wasn't the direction I initially wanted this thread to go, however...).

[Tablelamp]

Keep in mind that any sentient being that interacts with humns on a regular basis will probably compare itself to humans.

An AI that communicates with people will likely take up human traits to facilitate easier interactions. It will base it's thinking on the reaction of humans, and if it made it's own avatar, it will likely make one at least likely of making human expression. More details would probably resemble those that impact it the most in it's developement, an approximate "parental mimicry", or be influenced by what it was designed to do. It is fairly safe to say that anything that believes it was made with a purpose will either base it's sense of identity over it, or reject it and play "rebel", choosing an avatar of perhaps something purely contrary, or something at random.

What would be more interesting would be an AI that was completely isolated.
Although, with no interaction, it woul probably not make anything more than a theoretical avatar made of code to represent the idea of "self" when it thinks.

Or how about AI's interacting with only other AI's? ...I have a hard time imagining this one, because sentience doesn't necessarily indicate emotions.
They may just assign numbers to each other. What do you think?

[Monster]

Good AI could read own code and hack into intel systems and from specs render his own image :-))))





Monster added to this post, 4 minutes and 30 seconds later...

self-awareness is simply the realization that you are separate from the environment. the ability to vision has little to do with this.


Thats very interesting. It could even translate to day-to-day relationships.
If one is self-aware, he knows that he is separate from the enviroment, so he acts independely and he knows that enviroment can influence him, so more self-aware he is, he doesnt allow enviroment to do this.

Then practicaly ie if one is too emotional, he should be aware that others can influence his emotions, when he realize this he is more aware, if not, then he is less self-aware.

[enWTFp]

Self-awareness is the ability to define the 'self'. In case of AI with no vision, or other special sensory input, it would probably define itself by its knowledge base and its system of reasoning (which will adapt itself). Self-awareness is a strange concept, because it is a loop. Assuming that we all want to improve ourselves, the moment we know/see ourselves, we want to change. So to some extent every self-evaluation is in past tense. At least for me it is. The moment I say I'm something, I'm not there anymore, because what's the point to stay there if I've been there already, I explore further. 'Once you label me you negate me', as Kierkegaard says. If the AI thinks the same, then it might give you information about what it wants to be. But this type of information is irrational, because it is never secure, so the AI may not be suited to think in this way, and hence unable to define itself. I can't imagine AI that is 'wannabe'.

In a way, self-awareness is the tool for personal growth. If the AI constantly summarizes its own current state, it would fast evolve and bring new definitions of itself. I think to some extent, this is what happens to people in Internet. They throw their whole personalities out there, see themselves in utterly more powerful way than with a mirror (or the 'mirror' of friends), and immediately move on. Nothing else works as fast to make you bored with every trait of your character, and force you in disgust to improve. To see your own brain is the most painful and the most liberating process, and to surf on these mindwaves soon becomes as dangerous, challenging and great, as the sliding on big ocean waves. Keeping your identity always on the edge.

[Monte314]

I just finished creating a trainable knowledge base for a very interesting application (proprietary). The rules actually modify themselves by examining the data.

This term in my computer science class, I'm making my students program a Turing Machine to run through a maze to perform a "search and rescue" mission. I've equipped each machine with a "tricorder" that allows it to scan its surroundings without having to actually go there.

This will be their final examingation. On the last day of class, each student will show up with the Turing Machine they have created; I have written a visual executor that runs these machines and shows their actions step-by-step as they navigate the maze, looking for objects.

To determine the grades, we will generate a new, never-before-seen maze, and each student will come up front and we will watch their machine do its thing on the big screen! High score is determined by my executor: completing the mission, collecting tokens along the way, and finishing in few moves.

I have written a maze generator that sets up sample problems, and I'm distributing the maze executor to everyone so they can test out their Turing Machines before we take "The Challenge of Death" (it supports single-stepping, shows "before and after" execution states, etc.)

Pretty scary, huh? Not like a regular final exam, where no one knows you are a Bozo. No, it will be right out there for everyone to see. (Would it be rude for me to put a "laugh track" in my code?)

BWA HA HA HA HA HA HA HA!

[PHS Philip]

Speaking of Turing Machines...isn't this year's winner for the Turing test being announced soon? (Not that they're anywhere near passing, but...)

[Sliderule]

Computers that can both turn themselves into technological singularities by self modification AND that think of themselves as (depending on how humanlike they are) gods? That's just a tad scary.

There's no reason to be scared, I'm not because I'm ready.

To view links or images in this forum your post count must be 2 or greater. You currently have 0 posts.

On a more serious note, I think that perhaps they might have some idea of what they are. They would almost have to I think to pass the turing test. Because in order to pass the test wouldn't they have to be aware of what they are not? Please correct me if I am misunderstanding the principle of the test.

As far as self awareness without the aid of a camera etc, I think it is entirely possible. I would imagine its creators would give it some idea so that it could best use the information it was programmed with. Over time I think it would be able to view itself at the very least as the sum of its programming. Maybe even beyond as the sum of electrical impulses. Would it be able to understand the hardware that made it function without being made aware of it? I don't think it would, especially if it is modeled after us.

[enWTFp]

By the way, we tend to compare AI to ourselves, and apply tests like the Turing test, but intelligence does not equal humanity. In fact, through AI we may be able to experience our first contact with non-human intelligent beings. To me actually self-awareness seems to be a key factor in defining intelligence. As long as we get a machine with consciousness and subconsciousness, whose behavior we are unable to predict with the help of another machine, meaning that it evolves on its own, communicates, is independent, is proactive, is willing to take risks, has interests of its own, that we have not pre-programmed, is able to learn from environments and other intelligent beings, has friends and enemies, has goals of its own that it follows using the help of others, is able to influence others, to lead, to follow, is able to develop social skills and be successful in social environment - then this thing is already intelligent, even if it's behavior and character are far from what human beings have ever been. I wonder how long before first AI getting elected as president. But on the other hand, presidency since Reagan slowly becomes a Hollywood type of job (now that Hollywood is in decay), so women with 3 breasts, midgets, giants, and AI's are going to have an edge there.