When does an algorithm turn alive?

NQ · 2006-01-26T19:42:22

EDIT: This first post is outdated. Please ignore this first post and read the corrected version at page 2. Follow this link. Hey, I'm about to write an article which relates to AI, for a magazine. Before I do this I want to check my ideas with the people on this forum, to check for inconsistency. You guys usually know what you're talking about. =) Before I start explaining the topic, let me state some key information: - I'm posting coz I would like some help to check for contradictions, faulty logic and similar errors. - This is not the ACTUAL article, only a summary of the thinking I base it on. - Enjoy! When does an algorithm turn alive? ----------------------------------- One reason to why I like AI is because it forces us to face the really big questions regarding our existence. Say I'm working on creating a self-learning AI. I build a robot which learns about it's environment and how to deal with certain problems. One generation after myself, somebody else creates an even better one. And then somebody else creates an even better one. We end up with an autonomous AI robot, which is very hard to seperate from a human. It looks like a human, it walks around like a human. When you talk to it, it responds like a human. In the end you can't distinguish it as non-human, because it's built to produce the exact same behaviour as real humans. But it's still just a machine. An algorithm which is built to produce certain output given certain input. A very advanced one, but an algorithm none the less. This scenario have been proposed many times before, such as in the movie 'Blade Runner' from the 80's. In the movie, the main character must pick out AI machines within the actual human population. He does this with psychology-tests, since the AI is incapable of having feelings. As the famous thought experiment 'The Chinese Room' (wikipedia) tells us, this would be true in real life as well. The AI could still look happy, smile and wave its hand, but it's not really happy. A smile, a frown, a shiver... to an algorithm they're just movements of facemuscles. A means to produce a certain effect. The effect could be to appear human, or to communicate, but the desired effect doesn't matter. In the end, the smile is just movement of face muscles, with nothing underneath. An algorithms output to a given input. It's obvious that an AI - even though it's learning from its mistakes and displays intelligent behaviour... still lacks something compared to real biological brains. This 'something' is really hard to decribe, perhaps even impossible. Call it life, counciousness, spirit, creativity, feelings, sapience or whatever. That is the realm of seriously deep philosophy, and I will avoid the problem altogether by just summing it up with 'what AI lacks compared to real humans/biological brains'. For the sake of simplicity in the following text, I will refer to this thing as a persons 'ghost'. This is a term I borrow from the japanese manga 'Ghost in the Shell', which suits the topic quite well. It is obvious that an AI is a created being. Whatever it does, it does because somebody designed it that way. Somebody programmed it just like that. However, both 'Blade Runner' and 'Ghost in the Shell' suggests the idea that... once a system with powerful enough capacities is created, a ghost spontaneously spawns within the system, wether you want it to or not. Perhaps not immediately, but after a while, like a child trying to learn to read. Suddenly the 'blockade' is no longer there, and she can no longer choose to NOT read something she's looking at. 'Blade Runner' suggests this could happen with autonomous robots. 'Ghost in the Shell' takes it one step further and suggests it could happen with... the Internet. Sounds quite improbable, right? If you regard the Internet as a massive array of nodes, each with a vast amount of information, constantly expanding both its storage room and its contents and each node is communicating with all nearby nodes with blazing speed... then it suddenly seems less improbable. If you're imagining the Internet 'entity' like computer code being run, then I have to correct you. That would take somebody to write the code, wouldn't it? Rather, the processing would be the actual traffic of the net. Somebody requesting a pageview. Somebody opening a new server. Switching IP adresses. This system takes on the shape of a massively paralell processor. It's processing completely random things, producing completely random outputs, in a massive environment. Is it so unlikely that somewhere, sometime, a cluster of nodes are arranged with such properties that they exhibit the behaviour of self-preservance? If such a thing ocurred once, then it would remain, wouldn't it? Couldn't you even say, if you sum up the entire Internet into One entity, that it shows the behaviour of self-preservance? If an ISP is knocked out, the little white bloodcells (the network admins) work around the clock to get it back up running. Without competition, it's spreading across the world. First settling on the most lush lands, the wealthy and highly developed countries. When that is done, it's gradually spreading into Africa too, though that land is not quite as suitable habitat, so spreading there goes slower. If there was a competitor to the Internet, they would try to defeat eachother. Claim eachothers territory, like Macintosh and Windows. Either one would win, or they would settle into an eco-system. Sounds like the behaviour of a living being, doesn't it? But that wouldn't necessarily mean Internet has got a ghost, would it? Nope. Not at all. Just maybe. One can argue that other systems such as the country France, if summed up into one entity, behaves like a living being. If attacked, it will retaliate or request help from friends. The resources it's dependent upon are either self-produced or imported. If suffering from a lack of one resource, the hunt/desire for it grows higher and higher, possibly starting a frenzy or an attack on another country which has the resource. This sounds really silly, right? Systems cannot have ghosts, can they? After all, why would they? But how do you explain that we have ghosts, then? Evolution is a heavily backed up fact. Given that we originate from a bunch of goo in the ocean, at which point did we aquire these 'ghost' thingies? Of course, there's lots of ideas about what a ghost is and were it's from. I can't tell anybody what way it is. Nobody can, at least presently. All we can securely agree on, is that we DO have it RIGHT NOW. Everybody agrees we have got ghosts (whatever that is). Evolution happened. We have ghosts. This gives us three possibilities: 1. Our ghosts was given to us. (this implies an origin such as a higher being, God, or aliens if you believe in that) 2. We aquired them ourselves. (this implies that ghosts appear spontaneously within something qualified to hold it) 3. We had them all along. (this implies that even atoms have ghosts) Who am I to say that either of these are true? I can't. It would be based on my personal belief, and would thus be for me and me alone. But based on this, we can make some rather interesting conclusions. The list above would mean either of these things: 1. A higher being exists. An idea which in all versions of it has countless and severe logical flaws. It would mean logics aren't worth much. 2. This would mean that ghosts are something which appears when a system is configured the right way. That means we can create AI which is just as alive as ourselves. It would also mean that it's possible for other systems like the Internet, the stock market, families, and the country France have ghosts. Which is straight out silly. 3. This relates heavily to quantum physics, and would mean things such as the pavement you walk on has a ghost (but perhaps no intelligence). As well as your pencil, your shoe and your toupé. It would also mean we can create AI which is just as alive as ourselves. Important would also be things such as the concept of 'Gaia' (the collective 'ghost' of planet Earth). Which is just weird. Given this, no matter what option we choose, we can conclude that the universe is a far more weird place than it first seems like. It is not as simple as 'a place where physics happen and stuff'. Don't think we know everything yet. There's major stuff still to be unveiled. ------------------------ End. It's meant to inspire the reader to further thinking, and smash the belief that 'nothing matters anymore - all exciting ages have already been in the past, with frontiers, swords and heroes and stuff'. I want to make people see this age is not any less exciting than the rest. There's still frontiers. Any comments? Merry Christmas! =) (or whatevers appropriate to say where you come from!) [Edited by - NQ on December 29, 2005 10:06:28 AM]

Artificial Intelligence Programming

Started by NQ December 23, 2005 07:31 AM

87 comments, last by Timkin 18 years, 2 months ago

Iftah

413

January 22, 2006 05:37 PM

Quote:Original post by Daerax
No, you simply misunderstood me. I simply meant that he is not correct. He works only, from a suppostion among many. The burden of proof need not be proof of infalliability of theory but rather that it is your task to show that your theory has correctly made accurate predictions. What he stated was not fact nor has there been any experimantal evidence in its favor and is thus mere conjecting. Occam's razor is a common sense principle and any scientific theory which requires it is inherently flawed. Intepretation of measurement in Quantum mechanics for example. I do not see what the existance or lack of, of God or evolution figures into what I said.

Note also that a theory which makes many correct predictions and then one wrong one is not necessarily declared to be wrong or thrown out. In fact no physical theory can be *True* or correct (where correct strictly means Not Wrong) since they are in essence nothing more than approximations since a~b => a <> b

Maybe I did misunderstand you, I thought you are attacking his theory of "brain is a machine" and not just the experiment he suggested, if thats not the case then please ignore my response.

I think you focus too much on the experiment he suggested and ignored the idea of his theory (which is what I tried to protect). While the experiment to clone a human atom by atom is problematic (to say the least) the idea he tried to describe is that the brain is just a machine, made up of matter just like any other machine; and that the feelings and such are just states in the machine.

That theory may not be tested by copy pasting a human, but it is supported by many other experiments, for example the brain malfunctions or interaction of medicine or CT scans and such. They all support the theory that the brain is a machine (made of atoms and nothing more) and the spirit or feelings are just states in the machine. So I would say there is plenty experimantal evidence in its favor.

I dont have any knowledge in the spirit world so I dont know how they explain the effects of chemical medicine on the state of the "spirit" or how they explain a physical damage to the brain affect the nature of the "spirit".

What I said about evolution is that it explains how this brain machine came to be without any need for a God or for a special spirit in us humans. Again, it has nothing to do with the experiment he suggested, only with the main idea he described and I tried to protect.

edit:

Quote:
You cannot answer for me: Are all subatomic particles truly the same? Yes? Then beyond arrangement we are no different, you and I. We are entirely indistinguishable at small enough scales. Why, are we even seperable?

I dont know if all subatomic particles are truly the same, but I dont see the problem with idea that we are indistinguishable at small enough scales, or with the idea that beyond arrangement we are no different. Is that so impossible? beyond arrangment a truck and a spaceship are (almost) the same yet you have no problem telling them apart. Not to mention that a computer running windows and a computer running linux are almost entirely the same but are seperable. I think you agree that beyond arrangment my tricep muscle is the same as your tricep muscle (well, sort of) so why should the brain be different?

Iftah.

Daerax

1,207

January 22, 2006 06:33 PM

But you see, the key is not in the possibility of atom by atom copy but rather in whether or not such is a sensible thing to imagine and if so, if such an action would result in the same person with the same memories. It seems such an assumption would be safe but it is only that and too often are assumptions shown wrong. When it comes to the subatomic we are more mere chefs with recipes than chemists with reactions or scientists with understanding. So such an offering in support of brain is machine argument is doubtful at best.

You cannot say the brain is a machine until you define what a machine is. Regardless of what you state, it is likely that the extracted essense would be similar to a requirement of a Completely deterministic system in the classic sense. You take a view that the sum is no greater than its parts. You make an assumption and so what you say is not true nor is it a fact. And volition, where does it place in this connective analogy? Please keep in mind that I have not stated my opinion on the matter. For or agaisnt, it is more useful if i keep it apart from this discussion.

Your last paragraph! Filled with so many intricately connected and complicated concepts whose unraveling would be most tasking. What do you have to say on this: Not only is everything indistinguishable but it is also inseperable.

Also Iftah, look again at this statement of yours and consider its implications in its entirety.

Quote:I think you agree that beyond arrangment my tricep muscle is the same as your tricep muscle (well, sort of) so why should the brain be different?

[Edited by - Daerax on January 22, 2006 6:33:21 PM]

Timkin

864

January 23, 2006 12:50 AM

I really wanted to stay away from this discussion, but like any good black hole, it eventually grew to encompass that portion of the universe that included me!

Anyway, I wanted to offer a response to something Sneftel mentioned a page or so ago...

Quote:Original post by Sneftel
Humans are no good at perceiving absolute sound volume, light intensity, temperature, etc... we are informed only by recent changes in sensation.

Certainly evidence from machine learning shows us that adaptation requires a persistent excitation signal and that learning (in dynamical systems) with the inclusion of differential information is more efficient than without it. However, what is most important for survival is our ability to detect change, which is more easily obtained by sensing or inferring differential signals than absolutes. For example... let's say you notice a 5% increase in the temperature in your bedroom. Just knowing the absolute values isn't very informative. What if you noticed that this increase occured in less than 1 second. Now THAT is informative (and suggestive of danger). If it happened during a 12 hour period though, we wouldn't worry so much (particularly when the daily temperature variation might be as much as 50% or more).

In actuality though, we have a combination of differential and absolute sensors, each designed to give feedback on information important to our survival and function.

Cheers,

Timkin

Mizipzor

247

January 23, 2006 06:13 AM

The things Timkin said got my brain started. Up until now Ive always designed my nets to take absolute inputs. Absolute speed, height, angle and so on. If I also gave it inputs with the changes, the change of speed in the last tick, change of height and angle, it would probably be able to respond to changes much better. For example, a flying craft compensates by thrusting upwards when its on its way down. Im gonna try it out as soon as I can... I just have to get the back propagation algorithm for my net to work first, lol. :P

Anyone else thought of this? Have anyone tried it? Or maybe its the standard way to do it. :P And I just have done it wrong to this point... hehe...

Iftah

413

January 23, 2006 02:17 PM

Quote:Original post by Daerax
...
You cannot say the brain is a machine until you define what a machine is. Regardless of what you state, it is likely that the extracted essense would be similar to a requirement of a Completely deterministic system in the classic sense. You take a view that the sum is no greater than its parts. You make an assumption and so what you say is not true nor is it a fact. And volition, where does it place in this connective analogy? Please keep in mind that I have not stated my opinion on the matter. For or agaisnt, it is more useful if i keep it apart from this discussion.

Im sorry but Im having trouble understanding what it is you try to say (my English isnt perfect so I choose to blame my bad vocabulary).

The way I see it there are only two options:
1) option One: the brain is a machine, meaning its built only from matter (like any other physical thing in the world) and the mind (including feelings, consciousness, volition and all) are just states in the machine. Sort of like a software+hardware computer where the feelings are states in the software (except in the brains case the software and hardware arent clearly seperated). The self awareness I feel is hard to understand but I think its just a state in this brain machine.
2) option Two: the brain is a combination of a physical machine and some spritual out of this world element. This element is what defines me as a person and is what my self awareness is all about.

option two is what often suggested by religious people and is what I am fighting against. They dont provide any tests to check this theory or even define it well. This option is what I call a bad theory (using Ocassm's razor). Surprizngly many (intelligent) people trust that option two is the real thing only because we have this subjective feeling of self awareness that we cant believe a physical machine can produce.

Self awareness is something subjective and hard to measure but I can imagine being less self aware or more self aware. I believe evolution pushed us to being more and more self aware (but we are still not aware of many things).

If you think about it, it makes sense this self-awerness illusion makes the brain machines act better (survive more). I know Im just speculating and not offering scientific tests here but just think about it.

Quote:
Your last paragraph! Filled with so many intricately connected and complicated concepts whose unraveling would be most tasking. What do you have to say on this: Not only is everything indistinguishable but it is also inseperable.

what?? of course things are distinguishable and seperable! And if we are indistinguishable at very small scales so what? we are certainly distinguishable at bigger scales. Can you distinguish two programs by a single statement or expression? maybe not. can you distinguish between two programs by seeing 1000 lines of code? probably. I imagine our brains as a computer program with 10 billion lines of code. Certainly two different brains are distinguishable and seperable.

Quote:
Also Iftah, look again at this statement of yours and consider its implications in its entirety.
Quote:I think you agree that beyond arrangment my tricep muscle is the same as your tricep muscle (well, sort of) so why should the brain be different?

Why should it be different? its built the same way cell by cell when we were embryos (excuse my spelling, Im too tired to check the dictionary).
The brain is maybe a billion times more complex but its what Im trying to say its still just a bunch of cells.
If you somehow plug out my muscle and plug it into you arm you will get my strength (or rather weakness [smile]) and my speed and my flexability and whatever characteristics the muscle contains.
If you somehow plug out my brain and plug it into you head, you will (I know its not proven or testable, but this is what my belief suggest) gain my personality memories and self awareness. Or rather *I* will gain your body.
Yes, its not possible to test with current technology, but why should it be different then the muscle? These are just organs grown when we were tiny ball of cells floating in our mothers stomach.

Iftah.

klayAlloy

150

January 23, 2006 05:05 PM

Quote:Original post by HemoGloben
Uhh...Isn't a more likely option to the whole internet seemingly alive, france being alive thing the fact that both happen to be in large part run by humans that are alive? I mean, ...meh.

Ok, focus on the france thing. It acts like a living entity because it's run by living entities. Doesn't that make more sense than trying to pretend that france itself is a living entity? Your entire thought experiment attempts to grapple the idea of seperating acts like an entity and is an entity, but somehow manages to miss acts like entity because it's run by entity.

Meh, again.

I apologize for the incoherence, but I think my general thought is there.

A quick thought, if in fact France is a living entity, which is made up of little entities, meaning that that make up is possible. Look at what a human is:

a human is a collection of several entities all working together to form a human. We have several cells(entities) in our make up that are all different yet work towards the same result; a human entity. You know what i'm saying?

What is a collection of humans then? France? The internet? God?

Daerax

1,207

January 23, 2006 08:55 PM

This thread has motivated me to try again on my modern approach to physics series of articles. I think it is time again that physics remembered its ties with philosophy and cognition. Iftah, nothing you say is incorrect, at least not on the surface. But the reality is far more detailed than the image you sketched. Your last paragraph though... the first thing you learn in science is never to speak definately. Never say This Will behave like That. Instead, say, it is highly probable, based on the current state of knowledge that X might behave as such given conditions Y. [smile].

silverphyre673

454

January 26, 2006 02:42 AM

I haven't read the entire set of posts, but I just wanted to point something out about the whole "what makes it up determines whether it's alive" debate. In biology, things display properties that are more than the sum of their parts. These are called "emergent properties," and are essentially what allow us to function. On our scale, a tree, iron nails, and glue aren't really good for much (well, except maybe for air [grin]). Put them together in the right way, and you have a chair to sit in. You can't sit on a tree, nails, or glue.

This may be a slightly weak analogy, but I think that simply because an entity is made up of something other than organic compounds doesn't mean it couldn't exhibit many of the same properties (in this case, life, or intelligence). Robots could reproduce themselves, if they figured out how to mine, refine materials, and put them together properly into a new robot. This is, in essence, the same process we follow: eat, break down the food and turn it into new, useful compounds, and then use instructions coded inside us to make a new person.

The debate shouldn't really be over whether a machine can truly be "alive," since the scientific definition of "alive" is already defined and can clearly be met by sufficiently advanced machines. Nor should it be over whether a machine can be "intelligent," because if a machine can cope and react to all the situations a human can to the best of its physical abilities, it would have to be intelligent. Where the territory gets rougher is when you get into the philosophical/semi-spiritual realm of emotions and soul, where the facts and the beliefs are impossible to separate. Debate that all you like, but it is hard to come up with a conclusive answer.

All my two cents. Nice post, for the most part, though.

my siteGenius is 1% inspiration and 99% perspiration

Timkin

864

January 26, 2006 07:42 PM

Quote:Original post by silverphyre673
if a machine can cope and react to all the situations a human can to the best of its physical abilities, it would have to be intelligent.

...and yet a lot of very dumb human beings manage to live and procreate (my cynical side emerging ;) )! Higher intelligence clearly isn't needed for life... what we need though is to recognise the spectrum that the word intelligence is applied to and define intelligence based on the complexity of interactions with ones environment and the ability to manage them. Thus, the average human is probably more intelligent than the average cat, because the human's interactions with its environment include a whole range of social interactions that cats don't share. When it comes to self-propelled flight (without external apparatus), the common house-fly would be considered more intelligent in its ability to execute flight, but the human (not the average human mind you) would be more intelligent in their ability to explain flight (which, again, is a social interaction! ;) ).

Basically I agree with silverphyre673 though.

Cheers,

Timkin

When does an algorithm turn alive?

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

When does an algorithm turn alive?

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines