Archived

This topic is now archived and is closed to further replies.

j8l5s

Life like AI??

Recommended Posts

Is it really posible to make an AI program think like a human? I ask this because as far as I know human thoughts and actions can refelect many difernt and random things that that preson has encountered in there life. People also have habits and personalitys that develope from many difernt source in there life. Like say a kid is badly brund as a child that child could grow up scared of being brund agin or could simply forget it ever happened. You also have to take into acount the mental development of a person. There is an infant amount of things that could happen at anytime given time in a humans life. There is also a lot we still dont know about the human mind and the way it works.Now that I said all that, what do you guys think, is it possible??

Share this post


Link to post
Share on other sites
It''s difficult to say before its been done, but I reckon yes. I believe creating complex behaviour that externally seems like it is influenced by personality is in fact much easier than creating ''perfect'' intelligence. Human-like behaviour is mostly defined by the mistakes made (but also the ability to learn from them ...) That''s possible with simplistic learning technology.

You can fake personality and personal history. We don''t know how human brains work from the inside, but from the outside the seemingly strange process can be mimicked when enough factors are taken into account

I''m working on this at the moment, I''ll try to keep you posted.



Artificial Intelligence Depot - Maybe it''s not all about graphics...

Share this post


Link to post
Share on other sites
You make a good point, but think about it people often dont do the best possible thing the often take the easy way out even if your not spose to. Lazyness is something that would have to be considerd to, but if you think about the demolises the whole point of a computer. lol. Anyways I really doubt that true AI will be here anytime soon, are computer are still not good enough to handle that much.

Share this post


Link to post
Share on other sites
theoretically, i guess.
however, it might be rough setting it up. unless you cut corners (as alexjc suggested) and only "fake" it, you will have to set up the "training" to simulate a humans early life. if you cannot somehow give the AI sensors for sight, pain, et cetera, you at least would have to simulate these somehow (i.e. get the AI to "think" it has had such experiences as burning itself). i cannot think of a good way to do this, granted, but i''m not an AI expert.
just my 2 cents.

Share this post


Link to post
Share on other sites
An interesting question!
I thought many times about it.
I think the main-thing is the learning intelligence.
-Do anything
-Was it good or bad? (remember(situation and "flags"))
-
If same or similiar situation occurs do this(or not).
I''ve heard about a forumula about this. (Brown I believe)

Other thing I thought allready is:
You enter for example in a room.
The room is closed, no light from outside and simetrically.
I mean there is the same on the left an on the right.
Why do you look left or right first???
Have you the feeling to look on this side??
Or this: Say a random number!
Why can a human say a random number?
I mean a "computerrandom" is never a real random.
(some formula with current time if i''m informed right)
Example learning AI''s in games: Black and White.
And creatures. I''ve heard they made a test with the creatures of creatures ( ) in a flight simlulator.
(if crash: Au!) And they learned to fly really good!

Hope this can help or brings any ideas! )
Greets: C.Ruiz
(I know, i know my english...)

Share this post


Link to post
Share on other sites
Now there is a really good question. Why left or right.
Well taking in to acount a lot of difernt things I''ve come to the conclusion that it is not possible to have ture AI on a computer as we know it. The main resons I think this is:
1. A computer is designed to do only as its told, you cant give a comutper indepent thought.
2. 90% of all choices we as human make are based on feelings, and you cant give computers feelings.
3. You cant tell a computer to make a desion on its own. It can only tell it to read a program and do exacly what it tells it to.
I''m not saying it cant be done, but the only thing at this time that can think like a human is a human.
I would love it if some one could prove me wrong, I just cant see it happening.

Share this post


Link to post
Share on other sites
quote:
Original post by j8l5s
Now there is a really good question. Why left or right.

this is the same as the random number thing. unlike cruiz, i do not think humans can pick a random number; likewise, the direction isn''t random either. this seemingly random decision is actually based on far too many variables for us to predict, thus it seems random. but, hear me out here: computers generate "psuedo-random" numbers (based on some algorithm or another), but even though these numbers are not truly random there is still nobody that can accurately predict the next number. so, it doesn''t matter if the computer can pick actual random numbers/directions/whatever to be intelligent; humans can''t even do this. (go ahead, tell me i''m wrong; it''s just a theory)
quote:
1. A computer is designed to do only as its told, you cant give a comutper indepent thought.

i will take my argument to yet another level: neither can a human have "independent thought." i''m not arguing against free will, but personally i think that any imaginative idea, no matter how seemingly original, is really just a compilation of what the thinker has learned and experienced, thrown together in a new ([psuedo]random?) way.
quote:
2. 90% of all choices we as human make are based on feelings, and you cant give computers feelings.

i hope i''m not being too absurd yet... feelings are a concentration of various chemicals in the body, that alter the way the brain works. this can be simulated (although that would admittedly be a lot of work).
quote:
3. You cant tell a computer to make a desion on its own. It can only tell it to read a program and do exacly what it tells it to.

but you can tell it "play with this batch of input data and learn which combinations are good". this is a massive understatement, when talking about something like this, and "good" is hard to define, but theoretically it can be done. i think.

--- krez (krezisback@aol.com)

Share this post


Link to post
Share on other sites

Thinking very abstractly: Most people think of humans, human intelligence and thinking as something vague and unpredictable. This is not true: Human thinking is just a huge neural network. There are some factors which make understanding the brain so hard:
- Our brains are made up of billions of neurons
- There are different kind of neurons which all react differently
- Some behaviour (many call it ''instinct''...lol) is hardcoded into our brain (crying when you need attention)
- There are different kind of networks involved (for example memory is a kind of bidirectional associative memory and recognizing some kind of neural network)
- and the main problem is: all this is interconnected in our brains.
- all the above makes our behaviour so complex that some of it seems random.

I believe we can create ''human'' thinking as soon as we have enough understanding, computer power and memory. For now we can''t come near it.

Edo

Share this post


Link to post
Share on other sites
In a course in Complex Systems I learned that the human brain is about 1000 times more powerful as the computers of today... If the computer developement continues in the same pace as today, we will have computers with computatioal capacity similar to the human brain in around 25-30 years. Well, most of us will probably be alive then,

Share this post


Link to post
Share on other sites
You all make good points, but it sounds to me like every thing is theore. How far have people gone so far on this. This reminds me of the aids virus. You have thousands of people working year round tring to solve it("cure it") but barley getting any where. I would think that if it was posible some one would found it by now, but anything is possble.
Here something else to think about, how could it chose a favorit color or food? How could it become atracted to some one?

Share this post


Link to post
Share on other sites
quote:
Original post by j8l5s
You all make good points, but it sounds to me like every thing is theore. How far have people gone so far on this. This reminds me of the aids virus. You have thousands of people working year round tring to solve it("cure it") but barley getting any where. I would think that if it was posible some one would found it by now, but anything is possble.
Here something else to think about, how could it chose a favorit color or food? How could it become atracted to some one?



The question you ask all come down to an essential difference between the computer-model (that''s it: model, capacity does not matter whether it''s an average 500MHz Pentium or Sun''s latest number cruncher) we use today and the way our brains function.

We are rapidly closing in on the computer in memory capacity, we''ve passed in reaction speed, a brain neuron fires more slowly than a processor crunches data, the essential difference is, that a PC has one CPU, a big one has maybe 64 but they are very carefully designed to work together (and remember, 64 coworking processors do not equal 64 CPU''s working for themselves, there''s a law of diminishing returns).

So our brain can take millions of data-units in every second and process them in parallel. That''s what makes our brain very complicated, and then I''m not even mentioning the backfiring in our brain (anyone who has looked at the mathematical theories concerning neural nets that loop and feed the output from turn n as input on turn n+1 knows what I mean)

So if you would give a computer the possibility to have millions of tiny CPU''s that work in parallel, it might develop intelligence. The only thing we can do to mimic this is simulate parallelism on a small scale (every PC does this today to some extent) and hope we can create a small fragment of intelligence in a very small field.

But in the end, we''ll get there... And then we can all grumble when our computer refuses to boot because it does not feel like it :-)

Greetz,



******************************
StrategicAlliance

On the day we create intelligence and consciousness, mankind becomes God.
On the day we create intelligence and consciousness, mankind becomes obsolete...
******************************

Share this post


Link to post
Share on other sites
My answer to a post of krez some posts before...
I don''t think too that a human can make random decisions.
My theory is nothing really nothing is random.
Everything has somany variables that you can''t calculate!
Yeah right.
But if you think so a human must "experiance" all this variables.
Feelings? A good question. A result of this variables like I said.

Another thing: I''ve heard a few days before:
I read in a magazin how scientists try to explain how human-brain
works. An interesting article. Try to find something like that in the internet. It''s also about memory.

Sorry i''ve now school I must run!
cu next: C.Ruiz

Share this post


Link to post
Share on other sites
quote:
Original post by cruiz
Or this: Say a random number!
Why can a human say a random number?
I mean a "computerrandom" is never a real random.



Actually they cannot. Humans are hopeless at generating random numbers. You might try and suggest that they pick a number at random from the set of possible numbers, but even then, they do show bias. If you want to test this, here''s a very quick thing to try. As fast as you can (and being as honest as you can not to think about what sequence of digits to write), write out a very long integer (say 50 or 100 digits). Now perform a quick frequency count on the digits. You''ll find some are highly favoured while others are not!

Timkin

Share this post


Link to post
Share on other sites
There's a tonne of good discussion in this thread, which is great to see... just a couple of thoughts from me...

One way researchers judge the current accomplishments of AI is to compare the abilities of the AI to a particular level of human development. Many consider the current status of AI to be around the 3 year old mark. That is, consider the abilities of an average 3 year old child and these can be reproduced using current AI technology.

Emotions are possible in artificial agents... unfortunately there is (and has been for hundreds of years) a raging philosophical debate about whether other beings can experience emotions as we humans do. Consider this though: there is a very strong correlation between emotional states and neuro-chemical levels in the brain, suggesting a causal connection. If emotions are governed by the state of the brain then it is reasonable to assert that emotions in other agents are also governed by some internal state. If the outward behaviour of the other agent shows a correlation with our own behaviours while we are in certain emotional states, then it is reasonable to assume that while displaying these behaviours the other agent is experiencing an emotion and it could be labelled based on the behaviour of the agent, just as we humans label the emotions of each other.

Of course, there are complications. Consider edotorpedo's suggestion that crying is hard-coded into the brain to get attention. Actually, when you're an infant, crying is the physiological response to discomfort, particularly internal discomfort (stress) brought about by things like hunger, fatigue and certain emotional states. Parents and carers have a certain response (be it physiological or learned) which is to try and ease the infants discomfort (which requires the carer to pay attention to the infant). As the infant grows, it learns the correlation between crying and attention and throughout the following years it often uses this correlation to gain attention by crying when it chooses to. This suggests that the child may not be experiencing the same emotional state it experienced when it was an infant but that it is decieving the parent as to its emotional state.

One final thought on the issue of choice:

If you want to ask how an artificial agent might make a choice, ask how you make such choices. They are generally selected with a bias. Humans tend to make choices based on their ability to justify the 'correctness' of the choice, given input information. We would like to think that we are rational agents (selecting the action that maximises our expected utility in that situation) however humans are often irrational (possibly because we can maintain paradoxes in our internal logic and still perform reasoning - something AI struggles with). Additionally, correctness is something we judge based on our experiences.


Ultimately, to create an artificial agent that has the same faculties and behaviours as a human, you would need to create that agent as an infant and have it live in our world, interacting with humans. If it had sufficient internal mechanisms to allow it to behave in a manner indiscernable from other humans then you might consider that you had succeeded in creating a human-like agent. Then of course, you're stuck in the moral dilemna of deciding if it should be considered human and have the rights of a human!

Cheers,

Timkin

Edited by - Timkin on January 8, 2002 8:34:17 PM

Share this post


Link to post
Share on other sites
quote:
Original post by Timkin

Consider edotorpedo''s suggestion that crying is hard-coded into the brain to get attention. Actually, when you''re an infant, crying is the physiological response to discomfort, particularly internal discomfort (stress) brought about by things like hunger, fatigue and certain emotional states.


I think what edotorpedo meant was that the physiological response *is* the hard-coded behaviour. As far as I know, infants'' brains do cause crying. If so, then it must be the case that their brains are initially wired for crying in certain states.

With regard to the initial topic, I think it highly likely that AI programs will eventually think like humans. But you probably wouldn''t want them to. Humans are dumb and often nasty.

Share this post


Link to post
Share on other sites
quote:
Original post by Argus
With regard to the initial topic, I think it highly likely that AI programs will eventually think like humans.

heh heh heh... the first time i read this through i thought you said "AI programmers "... heh heh heh..

Share this post


Link to post
Share on other sites
If you have 2 identical doors, symmetrical layout and so on, most people will choose the one on the right . (I read that in an article in New Scientist about a few habits of humans, in some fancy department stores, people pay extra to have their products on the right hand side )

Anyway, I think one of the big things which is lacking is the actual interface. Most people would try to keep things simple by just having say a text input/output to start with, but what that means is that the computer probably won''t learn a thing.

Take a human baby. After its first word, it receives praise and attention from people, even though it doesn''t understand what the people are saying, it still must be able to understand that it is being rewarded and not punished. There needs to be some "hard coded" things in the baby, things like pain which aren''t learnt. Otherwise, you might get people growning up who believe that pain is a reward or whatever. But, if you try to hard code those things in, you must still be low level enough. Taking the text example, if you program it so that the sight of the phrase "Don''t be stupid", is a pain statement, then whatabout when it is used jokingly? The concepts of pain need to be very low level, and then learn upon, you can''t skip steps.

For the computer to become intelligent, it needs to have a full set of senses, so it can learn to interpret tone of voice, body language and so on. Without that, it will also not be able to learn, if you yell at a baby, I don''t think it would like it, and it would be a punishment. The computer would also need to understand that type of thing.

Finally, remember that a human isn''t born literate. While they are intelligent, we don''t really have any way to see it. Since a human takes so long to get to a level when they can talk (especially in computer years ), I think that we will get together more computing power than is in the brain, and then begin to teach it, so once it reaches human levels of intelligence, it could then continue on, and be even smarter than us.

And when that happens, does it get treated like a human? If we can''t tell the difference between a computer and a human, then should we treat the computer as a human?

Trying is the first step towards failure.

Share this post


Link to post
Share on other sites
The only way to prove thatsomething is possible is to do it. This hasn''t been done, so it hasn''t been proven possible. It hasn''t been proven to be impossible either. I wouldn''t really waste my time worrying about this though. If it''s believable, does it matter how it works behind-the-scenes? Should we really bother trying to differentiate between pseudo-intelligence and true intelligence? The way I see it, it doesn''t matter. In some philosophy class, I see why you''d discuss this. But game AI is not designed to be abstract and academic; it''s designed to work. If your computer player plays the game as a human would, then there''s nothing more you need to do!

Share this post


Link to post
Share on other sites
quote:
Original post by ragonastick
Anyway, I think one of the big things which is lacking is the actual interface. Most people would try to keep things simple by just having say a text input/output to start with, but what that means is that the computer probably won''t learn a thing.
...
For the computer to become intelligent, it needs to have a full set of senses, so it can learn to interpret tone of voice, body language and so on. Without that, it will also not be able to learn, if you yell at a baby, I don''t think it would like it, and it would be a punishment. The computer would also need to understand that type of thing.

ok, i know none of the details, but i have to say it anyways: in my younger sister''s "intro to computer science" book (i was bored and i wanted some light reading ) there was a sidebar about a robot they built in some university that has "eyes", "touch" sensors, "ears", and a neural net controlling it. the little article said that it was progressing, as they let it wander the lab and interact with things an people, in a similar fashion to a human baby (much more slowly of course). i wish i could remember the name of the project, or even the school that is doing it, for a reference... perhaps someone else has heard of this?
quote:
And when that happens, does it get treated like a human? If we can''t tell the difference between a computer and a human, then should we treat the computer as a human?

my feelings on that one change depending on the last sci-fi movie i saw that has androids in it

--- krez (krezisback@aol.com)

Share this post


Link to post
Share on other sites
quote:
Original post by krez
there was a sidebar about a robot they built in some university that has "eyes", "touch" sensors, "ears", and a neural net controlling it. the little article said that it was progressing, as they let it wander the lab and interact with things an people, in a similar fashion to a human baby (much more slowly of course). i wish i could remember the name of the project, or even the school that is doing it, for a reference... perhaps someone else has heard of this?



It is quite likely that you are referring to the work done at the MIT AI Lab. They have done some fantastic work along these lines.

Cheers,

Timkin

Share this post


Link to post
Share on other sites
quote:
Original post by cruiz
My answer to a post of krez some posts before...
I don''t think too that a human can make random decisions.
My theory is nothing really nothing is random.
Everything has somany variables that you can''t calculate!
Yeah right.
But if you think so a human must "experiance" all this variables.
Feelings? A good question. A result of this variables like I said.

Another thing: I''ve heard a few days before:
I read in a magazin how scientists try to explain how human-brain
works. An interesting article. Try to find something like that in the internet. It''s also about memory.

Sorry i''ve now school I must run!
cu next: C.Ruiz



Your right, the randomness in human brains is a very complex algorithm(or something like that) which chooses an answer, it may seem random but must follow logic (lots and lots of it!)

Even your random numbers in games are not random, they''re pseudorandom, numbers chosen from a large list of numbers, the list is so long it would take ages to start again, and therefore seem random.

So yes, nothing on earth is random.

Share this post


Link to post
Share on other sites
Human thought can be approximated by computers but only to the extent they were programmed to(by a Human). A computer cannot think like a human unless it is somehow given self awareness. Self-awarness would allow a computer to think about how it thinks and improve upon it.

Also, we think of creating artificial thought by way of emulating brain functions. One school of thought on this says the mind is part of the soul and the brain is only an interface to it. How does one code a soul in C++? :->

t = brain->GetThought(&mind);
brain->ProcessThough(t);
brain->EnableESP();
brain->TransmitThought(t, PersonX);
brain->Forget(t);

hmmmm... maybe

Share this post


Link to post
Share on other sites