Your Mom

Miscellaneous Forums/General Discussion/Your Mom

If you created an AI intelligent enough to, or strictly for the purpose of, grow and have 'emotional' relationships with others, how would it define its parents/mother/father? Would it ever attempt to define a person as one?

This is a rather odd thing that popped into my head recently and since I'm no good at AI (although somehow I doubt very few people here have any good knowledge of it either).

Basically, would an AI that develops emotional (bear with me here) relationships, would it ever consider a person as one of its parents? Would it consider a specific group its family? How would those attachments be made if not hardcoded into the AI and told to do so? If you never told it that it has parents/creators/whatever -- would it ever reach the conclusion that it has a creator or parent?

Would it just consider those individuals it favors most to be its closest 'relations'?

What if it was told that it had a parent and it didn't like the parent?

Seeing as even the most celebrated chat-bots are still just shambolically disguised databases I reckon you're about - ooooooh - two hundred and fifty years too early with those questions. Though, of course, the answer is (d)Kill All Humanz.

No reason why we can't speculate as to the answers.

I would think that the AI would define its parents or parent as it's sole creator... maybe

I think they would have given up making human-like AI, and instead achieved computers that act as the ultimate servants.

However, I think it would occur almost automatically with a well-written system. The "parents" of the AI character would be considered as its original suppliers of knowledge, etc. Thus, since they would also be present from the beginning of the program's life, they would become the most commonly referenced characters for various topics. (I think Antony's neural net stuff covers this branch of AI rather nicely...)

It depends on your definition of mother/parent: Do you consider this the person(s) responsible for its physical creation, or those nurturing it durings its upbringing, and emotional development?
With an AI that could be anywhere from the people responsible for the design of its circuitry, the people responsible for creating the software, the maintenance techs/psychologists/support staff helping if make sense of all the new information and learn how to deal with things, etc.

Since any functional AI is insanely far into the future (each current AI application really isn't anything more than a glorified database. A human brain has trillions of neurons and synapses, and we are not even scratching the surface of finding out what makes it tick, any true AI will be the product of hundreds or even thousands of people cooperating to design one, eons from now.
Until you can grab a DIY-AI kit from the shelve at circuit city, the whole 'who would it consider to be its parents' seems kind of a non-issue...

Ok, I'll give this a shot with some very hazy logic...

If AI is to truely be AI, it must have an idea of its own existance and therefore be a form of "life".
This in turn means we have to look into what defines "life".
Now, looking at this in a cold, harsh way, it is fair to say that self propogation of a life form is one of the, if not the most important componet of life.

A key reason why humans have family structures like parents is because we share a large proprtion of our genetic makup with them, 50% from each parent and 75% in the case of a brother or sister. It is in each of our interest to make sure that we propogate our own genetic material above others. So, if you work together, you are more sucessful therefore your more likely to raise kids to a point where they in turn propogate and thus your genetic material is more sucessful. The cycle is up and running.

Now, if an AI has to live by the same sort of rules we "proper life" do (i.e. born, breed, die), then I see no reason why an AI couldn't see itself as having a family structure, because it could be in the AI's own "biological" interest to make sure there are as many AI's similar to it as posible .This in turn gives rise to the emotions required to keep such a complex way of breeding working, ways telling and knowing family from non family etc.

Like I say, a littly hazy, anybody got better ideas?

Edit: typo

If AI is to truely be AI, it must have an idea of its own existance and therefore be a form of "life".
If this hypothetical AI achieves self-awareness, it would no longer be artificial.

if this hypothetical AI achieves self-awareness, it would no longer be artificial


It would still be a manufactured, non-biological, artificial being... So I guess it depends on your interpretation of the term 'AI' as well.

Our emotions are hardwired mechanisms which have evolved over millions of years. They serve many purposes, a lot of which are irrational. Rational thought can exist without emotions. I think that when we first create machine intelligence, it will be nothing but cold, rational thought; we'll likely spend ages afterwards experimenting with machine emotions.

I would guess that machine intelligence would form only relationships of familiarity with other intelligences unless we explicitly added socializing emotions into its mind. Although, if the machine intelligence needed to be trained, perhaps it would form a stronger relationship of respect and understanding with its trainers.

If you never told it that it has parents/creators/whatever -- would it ever reach the conclusion that it has a creator or parent?


Humans have a tendancy to make that assumption irrationally all the time. I suspect that it's because of an anti-depressant mechanism in our brains. Conversely, we'd probably (accidentally) leave plenty of real clues for it to reason out that it was made by us.

If you never told it that it has parents/creators/whatever -- would it ever reach the conclusion that it has a creator or parent?


A completely "cold, rational thought" machine could definitely deduce this if it was at the level where it could think something like this:

I think; therefore I am.
If I exist, then I either have always existed, or I have begun existing at some point in time.
My memory banks have recorded record thought only for a finite period of time, therefore I must have begun existing at some point in time.
Therefore something must have caused my existance.


etc. etc.

And if this intelligent logical system had the capability to ask questions and have curiosity, it would ask itself "Who or what created me?" and "What is the meaning of life?".

Our brains are simply quantum devices. Let me explain: our skulls do not contain enough matter to result in "our" perceived intelligence and abilities. If each cell could store a discrete bit of info, then our craniums have room for about a trillion bits. Hardly impressive.

Therefore, my theory is this:

The brain is like a two way quantum radio. It is tuned into a universal consciousness that exists quantumly everywhere all at once. This helps to explain deja-vu, past life experiences, thinking of a song seconds before it plays on the radio, everyone seeming to have the same idea at approx the same time (some act on them, some don't) etc ...

My experience of life is defined by the fact that I am riding a temporal wave through the Universal conciousness. Many other "brains" are on this same temporal wave and we are sharing a unique experience. I believe that because my brain is quantum my consciousness has existed everywhere all at once. What is "me" is simply a temporal manifestation of the physical.

Therefore artificial intelligence will not be created. We will create quantum computers and thus provide a means for intelligence to manifest from the universal consciuosness with out the need for a human brain to manifest the physical.

Wierd, maybe. Am I crazy? Or just quantumly rotating the wrong way? God knows. Speaking of God ... :)


Allan Hambrick

Lots of people have no parents due to being adopted or accident, so perhaps the AI could cope with having none?

If each cell could store a discrete bit of info, then our craniums have room for about a trillion bits. Hardly impressive.


An arbitrary unit as a 'bit' per cell seems very low, especially since it has been recorded that the brain uses many different chemicals all over the place to store information -- the capacity should be significantly more than just a bit per cell. Researchers also found that certain stimuli will only activate a few dozen brain cells to 'deal with it'. A lot of very complex things are going on on a ridiculously small scale.

In addition to that, in many ways the brain does act as a large database -- rather than storing the same 'objects' over and over and over again, it appears that in many cases the information stored in your long-term memory 'references' other memories. If one gets damaged somehow, it can affect many other memories with the same false information. You also make a (subconcious) selection of what information to store -- your brain doesn't store full motion video at 100fps day in day out. Much information deemed insignificant will not be stored in your permanent memory at all, and many other things will be very fuzzy -- you can recall the basics, but not the details. Nice garbage collection system, too.

A human brains start out with about a thousand trillion synapses at birth (connection points between the various signal cells), and estimates for adults range between a hundred trillion to 500 trillion. How many data lines & connection points do the current generation of computer chips have?

yeah multiple connections is where it's at.

Hambone: Each nueron in the brain has on average ten thousand connections to other nuerons. Connections are formed, strengthened, diminished, and destroyed as we learn and grow. The strength of these connections is analog, which affects the thresholds at which nuerons will fire. It's the strength of all these connections as well as which destination nuerons the connections are pointing to that hold the information for any one nueron. You can't model that with a single bit - that's a huge amount of data!!!

I have a great book to recommend to you! Carl Sagan's The Demon-Haunted World: Science as a Candle in the Dark

P.S. You have no idea what you're talking about.

"Agreement is not necessary — thinking for one’s self is."

And I have an interesting DVD for Octothorpe

"What the Bleep Do We Know!?"

http://www.amazon.com/gp/product/B0006UEVQ8/102-4288867-0912131?v=glance&n=130

Always be open to varying points of view.

I, for instance, hold to no religion. Nor do I believe in religion. That said I recognize that humans are spiritual beings and that EVERY race and society of man on record has believed in something larger than themselves. Might be crap, might not be. If I did not have the capacity to appreciate alternate points of view my life would not be nearly as interesting. Note I said "appreciate" not embrace.

P.S. If I always knew what I was talking about, how dull I would be. :)

Philosophy and Meta-physics is fun. Complete waste of time, of course, and much better appreciated with beer.

Allan

in many ways the brain does act as a large database -- rather than storing the same 'objects' over and over and over again, it appears that in many cases the information stored in your long-term memory 'references' other memories.


Annecdotal evidence: I've lived in the US for the past 7 years, but grew up in the Neterlands. I 'remember' much of my childhood with English dialog now, even though I know for a fact that the conversations in question were in Dutch. The brain is a funny thing indeed.

And this:
The strength of these connections is analog, which affects the thresholds at which nuerons will fire

is a very good point as well. Digital machines like todays computers have much lesser capabilities than analog or non-binary digital systems.

The only reason computers use binary (0-1) is that things like tertiary (0-1-2) and up is too difficult to accurately present, since a '0' and a '1' aren't actually absolute values, but constitute a value withing a certain range of signal strength. For tertiary (or above), you either need to have smaller tolerances which require a more accurate power system, or you need to deal with higher voltages and longer switching times between extreme values which either leads to more heat and/or to slower switching.

The brain, in contrast, is a self-modifying, self-repairing, self-replicating, self-programming, and very 'fuzzy' analog machine. All of the above is near impossible to replicate in digital form, and building something fro scratch in analog form would be orders of magnitude more difficult thanks to the lack of discrete rules that binary needs to make sense out of anything.

But even if you can make a technically fully functional 'brain', the really hard part will be to actually turn it on and get self awareness. A brain from a dead body won't really do much for you, even though technically it all seems to be there.

Hambone: I love it when people think rationally for themselves! Personally, my favourite thing to think about is Occam's Razor. Fun stuff! :)

> Your Mom

Really, this is just a test to see if this topic will be deleted isn't it? :)

What if it was told that it had a parent and it didn't like the parent?
I saw that movie!!!

On a serious side I don't think AI will ever come about. One of the current popular models requires the networked individual intelligences to become "aware" much like that bundled mass of nerves did in some proto-simian skull 500,000,000,000 years ago...

no

I have seen diagrams for Turing machines that include an ego component. the problem is with the amount of input available to these machines nothing really happens to stimulate them.

Do you remember ELIZA and RASTER??? Parsing was one of the tools hoped for in AI.

LISP created nets or rules... the problem is that to create a net of rules as large as our tiny brains would crash a CRAY... and run slower than ENIAC.

NIOL (Not In Our Lifetimes) I think... forget about creating Data (ST tng) and make me a flying car instead!!!

forget about creating Data (ST tng) and make me a flying car instead!!!
I'll second that. Flying cars and jetpacks are way over due! We know the US government used them to fight the Nazis over 60 years ago, so were are they all at? Surely a civilian version would have been made by now! :o>

you ever see that video of the flying "dustbin" that you stand in. Was pretty cool. Metal Gear Solid 3 featured them too, very funny shooting them down.

Really, this is just a test to see if this topic will be deleted isn't it? :)

Artificial Intelligence is game-related.

Flying cars and jetpacks are way over due! We know the US government used them to fight the Nazis over 60 years ago, so were are they all at?

Interesting. Where can I get more info about this?

If you are reffering to the hydrogen peroxide rocket-packs, I think they never could make them efficient enough for more than 30 seconds of flight. Still, even 30 seconds would definitely be a life-saver if you were surrounded by an overwhelming enemy force :)

Really, this is just a test to see if this topic will be deleted isn't it? :)


Actually, I wanted to see if I could make a thread called 'Your Mom' and not have it deleted. So I picked some random thing that came into my head that would fit the criteria.

I think, by chance, it's actually a rather fun thing to discuss.

Anywho, point is, I picked a thread title and filled in the content after.

Still, even 30 seconds would definitely be a life-saver if you were surrounded by an overwhelming enemy force :)


Problem is that if you're 30 seconds up thattaway, what happens when you run out of those 30 seconds? Splat. ;)

You press re-wind. :P

Problem is that if you're 30 seconds up thattaway, what happens when you run out of those 30 seconds? Splat. ;)

:)
At least you have 15 seconds to make your escape, and another 15 seconds to come to a stop. That's still a considerable distance.

I don't think it's conisderable enough for a group of soldiers to not shoot you either mid-flight or when you land. Or they could just shoot you on sight. Saves the trouble of aiming at you mid-flight. ;)

Or they could just shoot you on sight. Saves the trouble of aiming at you mid-flight. ;)

That's the situation you're trying to avoid; the jet-pack gets you off the ground and into the air where you're harder to hit. From there you continue to gain altitude and distance yourself from the enemy.

If 15 seconds isn't enough, and you really need 30 seconds of constant acceleration, you would just have to bring along a parachute for your descent.

To respond to the OP:

For an AI to recognise an emotional bond with a parent could only be as a result of there being an advantage in there being a bond in the first place. Such bonds in humans developed as a survival trait (and hence was brought about by evolution). Those humans that bonded with a parent and maintained close physical contact with a social group had an overall better chance of survival.

Our brains have thus evolved so that the behaviour has indeed been 'hard-coded'. However, I would consider the task of skipping evolution (with an AI) and hard-coding such traits by editing by hand would be as meaningless and impossibly difficult as trying to do the same with a real brain.

Such a complex system as the brain (or an AI) has only 2 ways of learning. The hard way; responding to external stimuli. The harder way; going through tens of thousands of generations in a reasonably stable environment in order to have a million million chance mutations giving rise to the possibility of evolutionary development.

I would argue that you cannot teach a person (by making them respond to external stimuli) to bond with a parent (if such trait did not already exist).


'Coding' an AI will remain in the arena of sci-fi until the end of all time. The only way to develop a true intelligence would be to have a complex system responding to rules that closely relate to those found in the real world. Whilst our computing power these days is very powerful, I would imagine that a complex enough system could only arise if we had several orders of magnitude more primary storage (RAM) coupled with a great deal of CPU power, or purpose-built fairly simple neuron emulating processors.

Making a single AI 'grow up' is probably a useless idea. You would need a community of such systems in order for them to develop. Intelligence is shown to be linked with compex social groups (primates, dolphins).


Within our lifetimes? Possibly, in some very limited way (simulating an ant colony perhaps). True AI I think is a very long way off, assuming we overcome the lack of oil to power our civilisation through the next 25 years and then again through the next 25 centuries etc.

You would need a community of such systems in order for them to develop.


I would think you'd need a community alright, but not necesarily a community of identical systems.

Just like an asian baby can adapt perfectly fine to growing up in a western household where many things would be very different to where it would naturally have grown up, they do just fine.
As long as the basic needs for interaction, intellectual stimulation, physical safety, and acceptance/love are covered, I'm sure they'll develop just fine.

wooo found the flying dustbin! This thing is for real. Saw it on tomorrows world once.



Also just watched I Robot. OK movie, but really just an action movie. replace the robots with soldiers and VIKI with a "bad boss" and it's pretty standard really. Checked out Wiki on the movie as I read all the Asimov books as a teenager and the movie didn't groove with my memory of them. Wiki explain all, it's a fake movie basically. Shoulda remember that I Robot is an anthology of short stories, not a single novel like Caves of Steel etc.

Also some futurologists are blabbing on about AI more powerful than humans in 40 years etc. Thing is this was all said 40-60 years ago with old sci-fi and it never happened. One thing's for sure, the Internet will surely have a much bigger role in any AI now, it could be it's memory/reference bank. Google AI? Also how come microsoft hasn't started making robots big time, maybe this is next?

> Also some futurologists are blabbing on
> about AI more powerful than humans in 40 years etc.

That's already happened since computers were first invented pretty much. You'll never be able to do math faster than a 1960's calculator even if you spent your whole life training for it.

It all depends on what you mean by more powerful than humans. In many tasks they're already overwhelmingly better than us (hence the reason why they keep replacing people in industry)

Sure that's not 'striclty' AI, but all AI is made by the user. Even when you simply tell it to do some math.

That's already happened since computers were first invented pretty much. You'll never be able to do math faster than a 1960's calculator even if you spent your whole life training for it.


I personally won't, but I do know at least two people who can do major addition/subtraction/multiplication/division in their heads faster than most people can punch the buttons on their calculators... Don't underestimate the brain. :-?

I personally won't, but I do know at least two people who can do major addition/subtraction/multiplication/division in their heads faster than most people can punch the buttons on their calculators... Don't underestimate the brain. :-?


I think he's referring to the speed at which the calculation itself is done, not the speed at which the input is entered.

Here's a sum I randomly created.

(((((((10 + 856) * 25) * 4) - 1024) / 65.2) * 153.2) / 22.2226)

Now find someone who can work that out quicker than me typing it on my windows calculator (22seconds approx it took me to type in)

32... did I win?

The question itself is very easy, once you obtain the scientific definition of the word "parent".

Calculators and other processors are faster in math becose you can design them to do that particular job. On the other hand, human brain was not designed (evolved..) to calculate math fast or at all, and the brain is not that good on learning new "software", so its pretty slow on tasks like these.

/me thinks for 3 seconds:

9048.3269146387701593121996680658?

when I said AI more powerful than humans, AI was the keyword. I meant sentient AI that can think and do stuff like a human, yet can do other amazing stuff too, i.e. more powerful, I didn't mean calculators or GPUs, I wouldn't call them AI would you?

> I wouldn't call them AI would you?

Depends what you mean by AI and whether you consider performing math as a form of human intelligence.

To do maths, a human needs to be able to see, perceive the numbers and symbols, interpret them into it's own understanding, work out an answer (somehow and people do this differently) and output it in verbal or written form, or just hold it in thought I guess. A human has to have the concept of maths and what a mathematical problem is. The calculator knows nothing about the concept of maths or problems and it must be fed the problem in exactly the right format and only gives the output in a single format. I never understands what it is doing or why.

So in short, yes I consider doing maths part of overall human intelligence, but to do it needs other skills and awareness.

As for what I mean by AI, I was extrapolating conventional AI to an entity with self-awareness.

Anyway wiki says:
http://en.wikipedia.org/wiki/Artificial_intelligence

A human has to have the concept of maths and what a mathematical problem is. The calculator knows nothing about the concept of maths or problems and it must be fed the problem in exactly the right format and only gives the output in a single format.
Math is simply a set of rules, with some taking precedence over others. The way you solve an equation, and the way the calculator solves an equation is exactly the same.

If your calculator needs input in a special format, you need to get a better calculator.

You guys are crazy.

I didn't read all of the posts, just some.

AI isn't really intelligence. It simulates intelligent acts. Based on parameters it is coded with or taught through it's logical system.

If all of this was possible today. The AI would do exactly what it was asked to do. If it was told to call someone dad, it would. If It was told to recognise the first female and male to interact with it as it's parents. It would.

It is that simple.

so I was talking about "Sentient" AI, and not just the illusion of sentience either, or do we all have that, hmm.

The way you solve an equation, and the way the calculator solves an equation is exactly the same
that's so 100% false. So you solve equations in binary do you and your brain works in 8,16,or 32 bits? Also it's well known that higher mathematics has several different ways to solve things. Anyway forgetting higher maths, here's an example, 53*11, how do you work that out? I go 10*53 = 530 (I just add a zero, didn't "work" anything out) + 53 = 583. Also when I added 53 to 530, I didn't add the 3 first or the 50 first, I just "knew" the answer. A calculator doesn't work like that. That is what I meant OK.

A calculator needs to be fed a special format of buttons being pressed, this being converted into binary and a sequence of appropriate actions. I can do a sum if I see it written or heard, not the same format at all.

Just watched "Ghost in the Shell" just cos it seemed topical right now.