The problem that many researchers see with the AI issue is that our current hardware innately lacks the ability to represent a set of infinite possibilities. Digital computers are, by their nature, finite. Whereas the human brain is actually an anolog super computer. Though we know the human brain uses electricity to transmit messages from place to place, it is a much different ball game compared to a digital computer, the messages are not binary, the brain interrprets these messages on the set of reals as opposed to the set of integers (please excuse the crude numeric correlation...but you get the idea). Pessimists say that even though some operations of the brain can be simulated there is no way to replicate it in its purest form (a true inductive learning machine) with the current hardware; only something like quantum computers could potentially allow any real substantial forward moving progress. Others feel that it is possible to somehow find away to do it even with the current hardware limitations. Personally, I feel that it is possible to construct such a feat on our current hardware, but I think researchers have been approaching the problem in the completely wrong way. They try and teach a computer how to differentiate faces or hold a conversation...when they should be trying to teach it how to think! You can't learn how to walk before you can even crawl, right??? i dont know either, but how do you teach machines to think without fully understanding the Secrets of The Human Brain? now im supposed to say hmmmm...... I've said it before, I think they will end up using the brain of a human to control the body of a machine and work from that understanding.
The computer processor is already made.
Maybe someone will have it in their will that they would like to be donated to science if they were ever put into a situation like being brain dead. Then that would give them the chance to work on expiriments like this.
Or maybe being brain dead would make them worthless, who knows. :) But you get the gist.
-WC- There is a problem with this idea. There are two parts to the human brain, there is the hardware...the actual brain matter and then there is the state of all the neurons. The problem you face is, once blood flow to the brain is gone, your brain looses its state...this is commonly called "brain damage"...the brain is never the same again, because it has lost valuable "software" that makes normal operation impossible...working with a dead brain would pose just as problematic as working with a digital computer because we still don't know the fundamental algorithm to describe the purest form of thought.
--mct. Wow, that makes a lot of sense. Great post!
-WC- Maybe it is possible to have the brain in some kind of stasis that prevents decomposition? :( I am sure that will be the easiest part of it.
More difficult would be finding a willing donor of a living brain! :)
-WC- heh, yeah. could use a monkey brain or something. Thatd be kinda cool actually. :) poor monkeys:( always have to suffer because of our curiosity...
but think of it... if it were possible to connect the human brain to a machine... would it be (for the person who's brain it is) just like waking up after sleeping? In escense you are your brain are you not? i'd find that rather disturbing and horrible... it would be like torture... only one way to find out...
I guess it would be like waking up, except it would feel like you slept for as long as you were in stasis.
Whatdyou reckon they keep the brain in? a tube filled with goo and the brain floats there? Inside the computer around millions of cables and circuitboards?
I reckon it'd be in like, an artificial skull (doesnt look like one of course) with cables and stuff in the supercomputer. I agree funkyhorror. I have no way of proving this but I would imagine that if we were able to ever achieve such a situation...a live brain connected to a machine rather than a body. The mind (what I call the state) might just go insane. Lots of veterans or accident victims go through what is called phantom limb syndrome, but could we even begin to imagine how taxing on the mind it would be to have your entire body amputated?! It is true the body is only a transport for the human experiance but the body is even more tethered to our concept of existance ironically by the true source of existance: mind itself! Morpheus said, "the body can not live without the mind"...I say, the mind can not live without the body! the body is kindof like the mind's blanket, we carry it around all our lives, we get attached...and we all know from messy breakups human beings have difficulty giving up things that they have become attached to. :-) What we would need is the brain of humans that have not yet become attached to their body, a clean slate so to speak. But, I and hopefully the rest of the world does not have the jubblies to do that sort of thing! I feel horrible just mentioning it. There's that word again *ethics*...hehe...development with organic matter is far to messy, us computer scientists should leave that to the doctors! This thread keeps getting better and better.
I agree with the thought that most brains would find the transfer something that would provoke insanity. I think that was even mentioned in Robocop1 (not using that as scientific reference, just a tidbit)
Maybe then, you find someone who's health is terrible and they are very rich. So you replace an arm. They get used to that, another arm. A leg. Then another leg. Then replace the legs with spider legs or treads. Start replacing organs. Pretty soon the human would be nearly 99% mechanical.
Then the mind isn't totally shocked with a total transformation.
-WC- I think that with current hardware, we can't simulate a self thinking machine. I'm sorry to be pessimistic, but that's just the way I feel about it.
As you said, the brain is analog. And as you know, computers are digital.
For instance, if you want to record sound (which is analog) you first need to digitalise it. That takes much more power and bits than the original (think about .wav files).
So I think we first need to 'analogize' the computers to teach them how to learn.
On the other hand: humans don't need to be tought how to learn... so why do we try to teach computers how to learn?
Where did you guys get those awsome avatars? ;) WC: who says robocop isn't a scientific reference?! ;-) hehehe...seriously though, lots of science fiction though it may seem stupid and off base ends up being true... Because humans come pre-packaged with all the software pre-loaded to "learn"... the human brain is the most amazing peice of hardware AND software ever written...! Good book that talks about this issue a lot: "Society of the Mind" by...Harrison I think. Great peice of "science fiction"...basically some genius hires a psychiatrist for a very special patient...hehe i think you get the idea...great book! WildCard, you bring up an interesting idea, however, in order to put a human brain in a machine, even if you are just doing it piece by piece, you still have to think of what the brain needs to survive. Oxygenated blood. However, if you could figure out a way that you could keep the brain in the machine, what do you have??? You have a mechanical human, not AI. There would be almost no point, it is just a human with metal instead of skin. I say almost because I'm sure you could think of certain things to do to a person, kinda like in Mr. Gadget (corny reference but oh well). Anyway, even if you could think of some good usefull things to add on to the human body, you still have to remember that if we can do that, then why dont we just build that machine and program it to do whatever useful thing it was we were going to have the human brain tell it to do. To address those that think that you can't build a self-learning robot. What about the possibility that you model the robot on the way the human brain thinks. It is based on your experiances. Just make sure that once the computer experiances something, it is able to analyze it and then add it into its own program. That way it will continue to learn, plus it could have more capacity and faster than a human brain. I think what we were brainstorming was not how to zip straight to creating pure AI. Any form of technological advancement has always had a stepping stone. For example, where did we get the idea to fly? We had an example, a model to follow: the bird. At first our implementations were not quite effective because we had not yet grasped the idea of aerodynamics. Please forgive my gruff example, I am not a physicist, but you get the idea. In the problem of AI, we only have one thing to go by: the human brain. Like our predecessors that studied birds with dreams of flight, so must we study the brain with dreams of...errr...thought. Ahh recursion, don't ya love it? :-)
So what is the point of hooking a working brain up to a bunch of machines? Simple, we could get a closer look at how the brain functions. Analyzing a working form of intelligence is the only way we can expect to come up with our own "artificial" intelligence. It would be impossible to try and reverse engineer the mind. It is the most complicated peice of software ever written! Try reverse engineering Microsoft Windows and have it operate exactly the same way as a ceran-wrapped version of Windows XP would? Do all that...but do it without a monitor. :-) Pretty tough, eh? Wow, great post. But heres a question: Just how exactly is hooking up the brain to machines going to help us understand it? I mean, what exactly are the machines going to tell us? Ok, I am not trying to say you can't create a pretty good machine that learns. If you want a machine that is Sherlock Holmes you got it! Deductive reasoning is very easy to do (relatively speaking of course ;-)...the big problem with creating AI...is something called inductive reasoning...coming up with new ideas. Right now computers can only tell us what we already know. If this and this or that then this must be true. But what happens if we don't know what must be true? Inacuracy! The computer can only give us answers that we have already thought of. Is that "intelligence"? Not in my book. If it was we could have pronounced that "AI" was created many years ago. No, no, the key is *inductive* learning. Real AI would never be caught dead saying "It's a matter of deduction, Watson."...It would say "What if I did this? What do you think about that?"
This draws me back into my original post. Creating pure AI on a digital system is a personal project of mine, but that doesn't mean I can't be realistic (or pessimistic if you prefer). It may be impossible, but I hope not. If not, this problem won't be solved until there is a breakthrough in the hardware sector.
Just my $0.02... If that was your 2 cents I would love to see your dollar!!!!! You are very intellectual, I like your logic. Well, I agree wholeheartedly. But I still feel that you could create a program that knows how to learn. Once it can learn, it can deduce things, even things we don't know yet. What you think? Possible? Hooking up a functioning mind, doesn't necessarily mean we remove the mind from the human body...at first. It just means a more direct connection there. If we see the kind of signals that are being sent to the different parts of the body, we can begin to deduce how these messages are structured. It would be similar to...hmm...port listening on a computer system. If you sit and just look at packets all day you can get a feel of how the protocol is structured. I surmise there must be some kind of structured language or syntax that the mind uses to transmit messages within the matter of the brain and other parts of the body. The problem is, working on the brain is tough! There is a lot of blood and it can get really messy really quick. That's why the rocket scientist cliche has been replaced with the brain surgeon of today. The brain is one tough mutha to work with...but we gotta do it. We have to find a way to get in there and get more details about how it functions...that's all I am trying to say. Of course, that is understandable, however, in order to hook up a brain to a computer or machine, we need to already be able to translate the brains messages. We need to know just what signals it is giving to each part of the body at one time. OR we could just hook up the brain, and we make up what every signal means. Ok, then your machine body will be going crazy because the brains is trying to get you to do one thing, but your machine body is doing something else. Well, just like when you are a baby and your brain learns to control your body, once you are a marchine, the brain will just have to readapt its messages to control your new body. As for hooking the brain up to a computer, I think that that is going to be pretty useless because how are we going to interpret the messages. We can't handle them one by one, the brain is sending thousands of messages at any given moment.
Pretty tough stuff, any ideas how we could ever decode the messages sent by our brain? Brute force, I'd say. It seems logical to me that the human body is just a bunch of cells interconnected with an extensive network of nervous cells. I would imagine that each part of the body would have to have some identifying "code" that would let it know that the message coming was actually to it. It's similar to way a computer network works. In reality ever computer on your network is recieving every packet you send over the network. There is something called a packet header that contains addressing information. The brain probably has something similar. Every message addressed to your hand passes down through your neck, shoulder, arm, and into your hand. What I am suggesting is that the brain might actually transmit that message throughout the entire body but the only object that is designed to pick up the message will get it and everything else will ignore it. This might be how it works. I am not sure, but it seems logical. In this case, all we would have to do when hooking up our HB to a computer for monitoring purposes is to be mindful of the activity that is going on when they move their hand or leg, and use the scientific process to figure out which messages are going where.
Hooking up a computer system to do this sort of monitoring would be the first step. The next step, after you have a bunch of messages and you think you know what they do, you could hook up a mechanical devices to the nervous system and have those mechanical devices pre-programmed to handle the messages recieved from the brain.
As HBs we are very blessed to have hands, fingers, eyes, everything all pre-programmed to handle messages from our brain. In replicating this aspect of the human experiance I think it would be a two step process. 1) Finding the right messages to send, and then 2) handling those messages with whatever device you choose.
I think the above stuff is already being dabbled in, or already done for all I know.
Now monitoring internal brain messages, this is the really tricky part for reasons I mentioned earlier: the blood, the high chance for death. How do we stick a plug into somebody's brain in order to figure out what is going on up there, without killing them in the process?
Originally posted by Helix
Of course, that is understandable, however, in order to hook up a brain to a computer or machine, we need to already be able to translate the brains messages. We need to know just what signals it is giving to each part of the body at one time. OR we could just hook up the brain, and we make up what every signal means. Ok, then your machine body will be going crazy because the brains is trying to get you to do one thing, but your machine body is doing something else. Well, just like when you are a baby and your brain learns to control your body, once you are a marchine, the brain will just have to readapt its messages to control your new body. As for hooking the brain up to a computer, I think that that is going to be pretty useless because how are we going to interpret the messages. We can't handle them one by one, the brain is sending thousands of messages at any given moment.
Pretty tough stuff, any ideas how we could ever decode the messages sent by our brain? Deduction is taking a discrete set of givens and finding the correct solution amongst a discrete set of possible solutions.
Inductive reasoning doesn't work off a discrete set of solutions...because how many solutions haven't been thought of yet? Infinity. If we could represent infinity on a digital computer, inductive learning might be possible to replicate. But the problem with digital computers is that they are innately discrete! On a 64-bit system you can represent about 18 billion (or trillion maybe) solutions (-9.17b to 9.17b). That's a whole lot of solutions but it still doesn't have the kindof precision that the human brain has.
Ok, let me try and explain it this way...
Q: Give me the smallest number that is greater than 0.
A1: Digital Computer: 0.0000000....error please give me a different question. Or it might just give you the smallest number that it has the capacity to give.
A2: Intelligence: 10^-Infinity
The point is, computers have a capacity limit in terms of the number of solutions they can pick from. The only capacity level for a human mind, is the speed at which we process things ;-)
Believe me, I am looking for a way to replicate the inductive thought process on digital hardware...Personally, I think that it can be done...but I am just not smart enough to figure out how. I have done some work in evolutionary algorithms, which seemed like a good idea at the time...but they are still finding the most fit from a discrete solution set, from a discrete set of givens.
Originally posted by Helix
To address those that think that you can't build a self-learning robot. What about the possibility that you model the robot on the way the human brain thinks. It is based on your experiances. Just make sure that once the computer experiances something, it is able to analyze it and then add it into its own program. That way it will continue to learn, plus it could have more capacity and faster than a human brain. Alright, I understand that, but I think that that is not the first problem we have with making this system. The first problem is that the computer is just going to be relaying its experiances, it is not going to be understanding what is telling you, or what you are telling it. It is just going to search its database to find an "answer" for you. AI can't be a computer program. If you say to a PROGRAM, "hello", and it says "how are you?" back to you, it is only because it has been programmed to say that as a response to hello. In order to make AI, you are going to have to create an exact replica of the brain, except with more capacity, more memory, and faster prossessing power. Well, you dont have to have more capacity and more memory and faster prossessing, but if you don't, then why not just use a brain? The whole point of AIs is because you are trying to create something with more "brain power" than humans so that the machine can do our thinking for us. Originally posted by Helix
Well, you dont have to have more capacity and more memory and faster prossessing, but if you don't, then why not just use a brain? The whole point of AIs is because you are trying to create something with more "brain power" than humans so that the machine can do our thinking for us.
We humans only use 10% of our brain. If the efficiency of an AI mind is 100%, you need only 11% of the memory, capacity and processing power from a human brain in order to have more possibilities ;). welcome to the real.................realy really soon
i want to go back to the phantom limbe theory. It is the brain that is telling us that we still have the limb in question, but the whole body.?
Now imagine you could make this human brain believe it's body was there and "re-boot" it.
That might work. Wow ... lots of fun stuff
Well my two cents ... we have to go behavioral on this thing and not get hung up on making things "like" humans. I don't know that we have the software or hardware yet, but we need to do the following few things.
- These entities we attempt to create must have senses. (Another big can of worms). What I am getting at is the stages of development and what you would expect a newly birthed AI to experience. As a baby, you are only concerned with not being hungry, not being cold, not being hot, etc. You don't know what you need or how to get it and your communication skills are not up to par. It is only the fact that there have been babies in the past that we recognize the need for something and as parents take care of those needs. As a birthed AI, the needs would be radically different than that of a human baby. What's worse, the crying would not necessarily be vocal. And while a baby may be able to go a while without food or water (like until the next meal) the AI could be dead or comatose before we recognize it has a need much less meeting that need. This is the greatest challenge to AI - creating something that lives and feels, but does so differently than us, and being able to keep it alive long enough to learn enough to communicate with us. Also, conception of time with the computer AI would be nearly inconcieveable. If a computer entity could see but can't change his view or change it quick enough, he could be "mad" in a matter of moments as he percieves he is looking at a pencil holder or computer programmer for eons.
- The easier way out of this is programmed perrogatives. No focus on these "soft" senses, but some other more basic or instinctual level ideas. These might be internal temperature or voltage management combined with simple on/off touch sensors not firing (running into stuff). Building a large enough list of these would allow something to act intelligent, but doesn't allow for modification or adjustment of these perogatives because they are so low level. Also, some pre-decision by humans of how much conciousness should be devoted to each perrogatives controls how they think. So we can create something that we tell what is important and how much to think about it. Sounds like kids huh? Sounds like untrue AI. Maybe once we get good enough, we can allow the AI to create it's own perogatives (as a perogative) and have enough concious ability to create and think on those own perogatives in addition to the original. Originally posted by Marc413
...the big problem with creating AI...is something called inductive reasoning...coming up with new ideas. Right now computers can only tell us what we already know. ... No, no, the key is *inductive* learning.
Excellent point...new member here...hope you don't mind me putting my $0.02 cents in as well...
The mechanics of learning can be simulated via neural network algorithms. If you throw enough processing power at the problem, you may even be able to come close to representing the physical brain.
However, as Marc points out, without inductive reasoning, there is little chance we could re-create a mechanical/electronic "brain".
The real challenge here is two fold... First you must create an system that mimics the brain (the hardware) and then you have to create a system that mimics thought (the software).
Maybe someday, advanced fuzzy logic systems might come close. And combining these properly (and by someone with a lot more knowledge than I) with neurel networks, we may even be able to re-create pure human "thought".
But someone asked another good question...why do this? The answer is simple. We humans underestimate our own brains. We have so much confidence and belief that the computer is smarter and capable of more processing than our own brains. We want to create AI because we think, on some subconcious level, that this system may ultimate answer questions that we can't.
The truth is, we humans barely utilize 10% of our brains capacity to think. Even real genius minds (think Einstein) used less that 15%. The more useful project would be to figure out how to tap into the other 80% of our own brains, instead of trying to create an artificial one. We believe this to be the holy grail because we understand computers today more than we understand our own brains.
Ok...I know this is getting to be a long post, but I have one more thought. If you had an old computer that was infinitely capable of upgrading without limitation, by adding memory chips, hard drives, etc., why would you throw out this old computer and start from scratch to build a completely new one? I think this is what we are doing by chasing the AI dream.
My quote below sort of sums up this post. Well, the problem is that scientists CANT find out how to use the other 80% of our brains. So why not create AI that you can control? By creating AI, we have more information about how the brain works, just because we model AI like the brain. So if AI is createn, we can ask the AI how to hook up our other 80% brain capacity. Originally posted by PDAGuru
The real challenge here is two fold... First you must create an system that mimics the brain (the hardware) and then you have to create a system that mimics thought (the software).
I dunno - I think we can make the software hardware independent.
Cases in point:
Java Apps
My SNES Emulator
Mr. Smith "downloading" into a human body (ok - it's the movie)
Another point - digital vs analog with respect to the hardware and stubling blocks towards AI. I think analog is the most direct and technically achievable way to this, but binary is not out of the question. I had the idea that everything could be represented as yes/no or on/off in the universe. The problem comes in the programming of the questions/variables.
IsRed=SubRed(Input)
If Is RedShade1
If Is RedShade2
If Is RedShade3
You get the idea - simple things could become quite complex, but with enough data and processing power I don't see the issue with this. I have the exact same thought about the universe: it's binary, but so complex, that you call it analog.
If you want to simulate 'analogity' with a digital system, you need *very much* processing power. You don't count in Gigahertzes and Gigabytes, but in GigaTera (a billion times thousand billion) and such order. Ok, well I got confused by that. Can you try explain? Originally posted by Helix
Ok, well I got confused by that. Can you try explain?
I will try to explain. I'm assuming you are speaking of the analog/digital representations.
Imagine an apple. The idea is that everything about that apple can be described with binary. Also a point to this idea is that we are attempting to have so much binary information that it approaches the analog (continuous - infinately divisible) truth. One good example is MP3 music versus analog records. MP3 music has a "sample rate" which inherently means that it will only sample or record noise so many times per second. There is no sample rate on a record. It is continuous. Though we may not hear any difference to our ears, it is because the sampling approaches infinity to the point where our ears cannot hear the difference.
Back to the apple:
Every item of the apple could be defined with yes/no parameters, but it would be near infinate. To define the apple, we would need to define each atom in the apple. Each atom's components, each atom's color, each atom's position in space. We would also need to define behaviors of the universe in binary. (I.E. reactions between atoms - bonding - heat). In otherwords, we would have to have near-infinate "data" and near infinate interacting binary rules to manage the objects in the universe. Because of this, memory, processing power, etc become real obstacles. Also, the program would be HIGHLY recursive (each question begetting another question). The problem is, can meaning ever be found in a system like this where everything is yes/no? Is recognizing an apple as a near infinate stream of rules and binary yes/no pairs meaningful? Thanks, I'm pretty sure I understand now. So I think that the problem is not how to make a binary system to hold the information, but to design a NEW system. A different system. One that does not use Yes/No. [long post]
Hello all, I would like to add my opinion to this interesting thread. Going back to the problem of figuring out how the brain works by listening to its messages: unfortunately this is not as simple as listening in on packets. There has been a lot of research about what the signals inside the brain mean, and what each neuron is used for, and there hasn't been much progress yet.
Obviously we know that when we burn our hand, for example, neurons are sending pain signals from our fingers, to our spine, and up to our brain. Also, our brain will most likely send a signal back down the spine and to our fingers to tell them to move away. So we have a pretty good idea of how the peripheral nervous system works. What we don't know is how much of the inside of the brain is organised.
The debate rages as to whether the brain is local or distributed. For example, we know that if certain parts of the brain are damaged, we get certain side effects. If someone sustains damage to their hippocampus, they have trouble forming long term memories, just like the guy in Memento. This would suggest that the brain is divided into neat little departments, each dealing with a separate function (a local arrangement). To extend this further, we might say that every neuron in the brain has a specific purpose. This view is reinforced by the visual cortex in the brain (that interprets what we see), which is arranged as a 2D matrix (pardon the pun) of neurons, corresponding to a gridlike representation of what is coming through our eyes. If we see something in the top left of our vision, the top left patch of neurons in the matrix will be activated.
If the brain was entirely local in this way, then it would be fairly simple for us to figure out how it all worked, because if we could just listen in for long enough then we would figure out exactly what every neuron is "saying", and when it is activated, etc. Unfortunately there are two good reasons for thinking that the brain doesn't work like this.
The first is a set of experiments carried out on rats. Rats were put in a simple maze environment, and they eventually learned to find their way out of the maze. They were then operated on, so that parts of their brain were removed. The experimenters found that the part of brain that was removed didn't affect the rats in a specific way. In all case they ended up slightly slower, and couldn't solve the maze as well as before, but there were no specific problems. For example, you might expect that some would lose their sight, some their muscle control, some their balance; nope, they all worked reasonably well, just generally slower.
Those experiments suggested that a rat's brain is distributed, i.e. no region or individual neuron is responsible for anything specific, but when they all act together, they are capable of "being a rat". As an example, if one server in the Internet goes down, most likely no one will lose internet access because all the network traffic will be rerouted to other servers.
You may say, well, there's no reason to think that human brains work anything like a rat's brain. So here's another reason to think that at least part of our brain is distributed. It's a thought problem called "the granny cell".
Let's assume that the brain is local, and that every neuron in the brain has a specific purpose. If I see my grandmother walking in the street, some neuron somewhere will recognise my grandmother and will send a signal off somewhere in my brain (which might result in me saying hello to her). Now, what if, the next day, I see my grandmother wearing a different dress? Will I still recognise her? What if she has her make up on? And a hat? And sunglasses? It's not feasible to believe that a single neuron could deal with all these possibilities and still recognise my grandmother successfully.
So perhaps these are the jobs of other neurons. Perhaps there's one neuron for my grandmother wearing a red dress, one for a blue dress, one if she has a hat on, one if she's got sandals on, and so on. But this isn't feasible either - what if I see my granny in a swimsuit? That requires another neuron. If I see her bungee jumping? That's another neuron. If I see her flying past me on a skateboard? There are an infinite number of situations in which I might see my grandmother, yet in almost all of these cases I would be sure to recognise her. If each case requires a separate neuron to recognise her correctly, that would be a lot of neurons!
Therefore there's good evidence to believe that at least part of the brain is distributed: that my thoughts and my behaviour is governed by a large collection of neurons all working together in a specific way - if you separated the neurons out, they wouldn't be meaningful on their own. This makes it hard to decode the brain's messages simply by listening to individual neurons.
Sorry for the long post, I think I got carried away! Hope that provides some points for discussion. Ok, well it is REALLY hot out, and I'm tired and lazy. So I'm not gonna read your whole post because I'm just too damn lazy. Think you could make a summed up version of your point????
:D :D :D Well, it's been a very interesting thread, and I'm eager to hear more on what you guys (& gals, ofcourse) have to say. I myself am a teenager and have no idea about what's acually going on as to research like Marc is doing, or the kind of 'medical' problems Squiddie certainly knows about.
But just for Helix here, you lazy bum :P :
It's not easy to figure out how the brain works, by the methods already suggested, because the brain is complicated.
Concise enough? ;) In a nutshell, it's tricky to listen in on the brain's messages because evidence shows that the messages might not mean anything in themselves, rather that they mean something when taken as part of the collective whole of the brain.
Archangel put it well I think :-)
I'll try to keep my posts shorter in future, sorry! If I may ask: please don't! I enjoy the reading, and you're good at explaining things, and you're explaining things that I didn't know about .Now, if only other people (I won't name names ;)) could be bothered to actually read..... :D i think that AI will b accomplised someday....but not very soon. having things that think for them selves is very hard....i mean think about it....its like creating a human being except we have to do all the dirty work .... if we find a way to make a chip that has infinate potential then we can do it.... Lol, Archangel, who could you mean??? :D
Anyway, thanks I understand now, and please dont make future messages shorter just because of me. If you really want you can put a note at the bottom that says, "In a nutshell for Helix: blahblahblah". But only if you wanna.
Signed,
Lazy Bum......I mean, Helix. i think that AI will b accomplised someday....but not very soon. having things that think for them selves is very hard....i mean think about it....its like creating a human being except we have to do all the dirty work .... if we find a way to make a chip that has infinate potential then we can do it....
Yes - you could imagine that to create an artificial entity like ourselves, you would have to provide it with senses like vision, hearing, touch (and maybe even taste or smell), and you'd have to let it learn to decode what it is sensing all by itself. It takes anywhere from 15 to 20 years (or perhaps never?) for a human being to reach full intellectual maturity, so I wonder whether it would be realistic to think that a machine could do it in less time.
Of course, that would be creating a machine that thinks like a human, and as we know humans are not infallible.... perhaps making a machine that thinks rationally, as opposed to humanly (it always makes the best decision it can) will be an easier problem.
BTW, what do you mean by a chip with infinite potential? Well, there's always the question whether or not it would be the best decision. Rational is not always the best. Case in point would be (as we're all fans here), Neo & Trinity going to save Morpheus. The rational choice, that Tank thought of, was to kill Morpheus, because it would be impossible to save him.
So basically, if machines like that were around, it'd be really boring and we wouldn't get any really cool shootout scenes! :D Lol, and yes that is the most important thing to be thinking of!! Yeah, about the AI maturing faster than humans. I think that that is definetly possible as computers can think oh so much faster than humans can. Individual electronic components do work a lot faster than human neurons, which are surprisingly slow. But our brain makes up for this because there are so many neurons (about 10^12) and they're each connected to about 1000 other neurons, and they can all work in parallel. Maybe when we create electronic components that can work in parallel on those kinds of scales, we can come up with something that (potentially) could really outperform our brains. Good idea, even though I'm having trouble understanding it........but it sounds good!
:D :D :D Originally posted by coolbean
I will try to explain. ...
Back to the apple:
Every item of the apple could be defined with yes/no parameters, but it would be near infinate. To define the apple, we would need to define each atom in the apple. Each atom's components, each atom's color, each atom's position in space. We would also need to define behaviors of the universe in binary. (I.E. reactions between atoms - bonding - heat). In otherwords, we would have to have near-infinate "data" and near infinate interacting binary rules to manage the objects in the universe. Because of this, memory, processing power, etc become real obstacles. Also, the program would be HIGHLY recursive (each question begetting another question). The problem is, can meaning ever be found in a system like this where everything is yes/no? Is recognizing an apple as a near infinate stream of rules and binary yes/no pairs meaningful?
This is a good example to highlight my point...after doing all of this programming and building enough processors into a box so that they can crunch enough binary data in order to "describe the apple", what do you have?
Answer: A very slow, primative computer capable of only describing an apple! (And most likely, not very completely either!) and crunching numbers....not crunching apples.
Now...let's take that apple, hold it in your hand (is it cold, warm, wet, soft, hard, heavy, dirty), and as you bring it to your mouth and smell the aroma, your eyes just determined a lot of data (is is ripe, is it red, green, yellow, etc.) and then you take a big fat bite! (is it sweet, juicy, sour, crunchy, soft). Was there a worm taste in that bite?
Now...all this data, from the hands, the eyes, the nose, the mouth all get fed back to the human brain and processed instantaneously and completely, faster than any processor currently or ever to be made by Intel...without any need to "write to screen", "Save to Disk", "Error Check", etc. Yes, and that is what we are saying, we need to design a new system, one not dependant on binary numbers, but one dependant one its on sensory projections. Two other problems with artificial AI:
First is the lack of complete information. It is one thing to have an artifical intelligence, but even with intelligence, having the "sense" to act or respond on incomplete information is crucial as well. "To err is human" as long as the intelligence has reasoned out things from the information it had, we cannot discount that intelligence for poor results. Another part of this is to ask questions without answers or not easily answerable. When would an AI say "I don't know"? We certainly woulnd't want an AI hung up on a rhetorical or otherwise joking question for infinity.
Second is causing the need for communication. As we are intelligent and grow, we recognize other people as "like" us. With an AI intelligence being inherently alien, how can we expect it to identify us as an intelligent and have the desire to speak with us? Originally posted by PDAGuru
If you had an old computer that was infinitely capable of upgrading without limitation, by adding memory chips, hard drives, etc., why would you throw out this old computer and start from scratch to build a completely new one? I think this is what we are doing by chasing the AI dream.
Problem with the human brain is that it is limited in capacity by biology. Biology evolves extremely slow. Computers have the potential to evolve much faster than we do. Whether this potential materializes or is even there, can only be assessed from hindsight. So I haven't the slightest...only when we achieve artificial life will we know for sure. Originally posted by Squiddie
Individual electronic components do work a lot faster than human neurons, which are surprisingly slow. But our brain makes up for this because there are so many neurons (about 10^12) and they're each connected to about 1000 other neurons, and they can all work in parallel. Maybe when we create electronic components that can work in parallel on those kinds of scales, we can come up with something that (potentially) could really outperform our brains.
I call that the Quantum Leap ;-) No seriously, the buzz going around is that only a Quantum Computer would have the potential processing power to solve this problem. Originally posted by coolbean
Two other problems with artificial AI:
Second is causing the need for communication. As we are intelligent and grow, we recognize other people as "like" us. With an AI intelligence being inherently alien, how can we expect it to identify us as an intelligent and have the desire to speak with us?
Do we not have a desire to speak with our creator(s)? Prayer, introspection, making love.
Pick one/more/anything else :D LOL, well thanks for the answer, but originally my question was how the Quantom people plan to solve this problem. So again I ask, How?
:D -Helix- :D Whoops, I thought you replied to the "Do we not have a desire to speak with our creator(s)?"-post :p
I still think prayer, introspection or making love would be a good idea though. :D Oh, well yes angel, I totally agree! :D I think that in order to be able to truly learn, you have to be able to experience things. So we program a computer to learn, give it a few basic function, (accessing senses, moving, interpreting images, etc.) and put it in a robotic "body" with sensory input. It reacts with the environment, and "evolves" into true AI.
The reason that it is diffucult to make true AI without giving it a body is that eother you program information into it, and it has the information, or you don't give it anything, and it is almost impossible for it to learn anything by trial & error. I'd appreciate any comments on this, as I'm just kinda wandering in the dark. i guess this depends on whether or not you believe in this or not but... i think that a human brain would not function right with out a soul.. i belive that our emotions have something to do with the soul... like Love and Hate Jealousy ect...
i know some say it has to do with the chemicals in the brain sending signals and all that which is understandable and mostly true because you do interpet love and emotion through your brain but i belive our souls are just in a vehicle ( our body) and then the soul leaves the vehice the vehicle is not good anymore including the brain..
you've probably have had a post about this but i just wanted to add that
feel free for expressing your thoughts and suggestions Thanks! :D Ditital computers are by nature finite.
yes ditital computer are but quantum computers are Not
and yes their is a such thing but only in the 1+1 stage but once they go into practical use imagine the possibilites of AI I want to return to the idea that we only use 10% of our brain power. I have 2 main goals in my life - Use my brain to the full capacity, and take care of our flaw (see quote). Does anyone know where I can read more on the brain capacity one? I'm sure many men have worked on this, and I have to know more.
P.S. All the posts on this thread are very interesting. Keep them coming Our conciesness only occupies 10% but our uncoincecss ocupies the 90% the key is waking up to the other part. yes. ive found the best way to use the other 90 is to astral project. (1) We don't know that the brain is actually analog. There could very well be discrete thresholds of response to electrical and chemical stimuli. I'd like to see some hard evidence that this isn't the case before we make the leap to "the human brain is analog."
(2) Not all computers are digital. It's not difficult to build an analog computer. They are used in various facets of industry, and in some ways are more useful than their digital counterparts.
(3) A digital computer can emulate an analog computer in almost absolute precision given enough processing power and memory capacity. I doubt that our brains are so sensitive to signal levels that an error of 10^-12 would cause a marked state change. (4) Humans use more than 10% of their brain capacity. In fact, we use nearly all of it. The 10% figure is an urban myth.
(5) Digital computers are not limited by binary to making simple "yes/no" decisions. I'm not sure where this logic came from. It is perfectly reasonable to create a program that evaluates variables and chooses the best of a million possible courses of action. That is hardly categorized as a "yes/no" decision. Whatever an intelligent being defines it to be. :p
Seriously, I've thought about that question quite a bit. I have my own definition that takes ideas from some of Carl Sagan's writings in Cosmos.
I would define an intelligent being as one whose information storage capacity of its brain was equal to or greater than the information storage capacity of its genes expressed in bits.
I'd define an advanced intelligent species as one whose cumulative information storage capacity in external media (libraries, etc.) was equal to or greater than the storage capacity of its brain.
Works for me, for now at least. Well, I guess you could have been even more confusing :)
At least - thanks for not saying "data storage" and throwing me for a total loop!
By this definition, a species that is intelligent will become more intelligent. I'm afraid to sound like a total moron, but was the guy who invented, let's say, a catapult, less intellingent than me, or was he less "something else"? The ability to realise you exist. Yes, "self-awareness". I have two problems with this:
1) I think my dog realizes he exists - although, probably somewhat differently than me
2) It seems like this definition is more suitable to be - partly - a definition of "life". Allright, here are just some ideas of mine.
I think that creating AI would be very difficult without understand our own brain, and afaik, we are still far away from that. plus, who says our 'minds' are defined only by our physical form ? As I said before, strange things happen in this world, Astral projection is one of them, without some form of non-physical state, Astral projection would be impossible. And since we know nothing about that either, it adds to the difficulty of realizing AI. Also our brains are basically computers, in some way or another. We have parts that make sure we breath, that our heart beats, basic life functions. This part wouldn't be to hard in making an AI program, you right some functions, safety features to ensure the existance and continuation of the program. Then, parts of our brain are used for our basic functions like movement, eyesight, hearing, smell etc. Again, this would also be possible in a computer... But then the trouble starts, parts of our brain generate our own thoughts, and yet other parts make up our subconscious... That imo would be nearly impossible to realize... plus, don't forget about that non-physical form.
Also I believe hooking up a supercomputer as an extra processing unit to a brain would create massive opportunities. if possible, you create some sort of interface between the brain and the computer allowing us to perform massive mathematical equations etc. Things we can't really do on our own... a symbiosis between man and machine would be one powerful tool. The independance, uncertainty, emotions and mind of the human brain, combined with the strength, speed and reliability of a supercomputer... Imagine... See, this is what I think the problem is...
Why do we need to define intelligence?
1. Because we want to create artificial intelligence.
When you programm anything, before you even begin programming, you define your goal, right? So you would define what you want that AI to do for you. Maybe it does not need to have too much "intelligence", and we already have such sub-intelligent robots.
On the other hand, I think a lot of times the goals are underestimated. That robot that went to Mars had to run into a rock first, and only then it would start "thinking": is this rock too big for me to climb over? Yes, it is, therefore, I need to go around it. That process takes it a few seconds, if not minutes.
So where is this fine line between this robot's "intelligence" and another robot's, who would go around a rock to begin with? What is the additional intelligence needed?
But then, maybe you want to make a totally automated cashier (they have them in HomeDepot already). Those things are way "under-intelligent" right now. But if they want to make them intelligent enough, they need to fix much more than just their current problems. This AI would need to "know" that sometimes it's better to let the lower price go, even if the sale is already over, or to accept an expired coupon etc. People "bend" rules, and we act irrationally, and THAT is how we get things done.
This cashier is totally different than, let's say, "the woman in the red dress". Mouse said "she does not talk much", but apparently she does other stuff :D That requires a whole lot of imagination and fully irrational thinking.
2. Because we might need to be able to recognize intelligent life (if we ever get a chance to meet it).
That, I think, is much more problematic. What if that life has the level of intelligence of our dogs, for example? Is that "intelligent", or not? What if WE are at our dogs' level of intelligence to it?
How would we even recognize that they are alive, let alone "intelligent"? What is "alive"?
Given that it's very unlikely that we would ever meet the little green men, I guess it is not that important to define our own intelligence. But it's just a nagging question. Originally posted by yanka
By this definition, a species that is intelligent will become more intelligent.Hmm. Well, I think up to a critical point there is always the possibility of evolutionary digression. Becoming more specialized, or complex, or intelligent is not guaranteed by natural selection... it is only a possibility. However, I think once a critical intelligence is reached the species would quickly become "advanced" and would probably not digress unless it was the result of some horrible disaster - natural or otherwise.
I'm afraid to sound like a total moron, but was the guy who invented, let's say, a catapult, less intellingent than me, or was he less "something else"? Since man hasn't evolved appreciably in the last many thousands of years, on an individual basis (ignoring actual IQ, and basing any comparison strictly on species genetics and structure) you would be exactly as intelligent, since you would both share essentially the same genetic capacity for intelligence.
However, due to the amount of information that is stored externally to our brains in today's world, you would have to conclude that man as a species is more intelligent today than they were a few hundred years ago. At least by my definition, and it is certainly contestable - it is a definition more for the potential information capacity of a species... not a measure of how knowledgeable any one member can possibly be. As far as knowing when AI is actually "intelligent," I think that will probably fall under the same umbrella as so many other things do... "we will know it when we see it."
You see, defining "artificial intelligence" thresholds aren't all that much easier than actually creating it. Humans have the knack (because we are really intelligent) to quickly identify patterns and such. We can quickly recognize faces. How? Can we really describe, verbally, how we do this? Or is the simplest answer just "we know it when we see it?" Were we able to easily communicate just how we accomplish facial recognition so efficiently and quickly, we'd be well on our way to emulating that process in a silicon processor.
The same is true for AI. Were we able to easily communicate just what definition would be an appropriate litmus test for artificial intelligence, we would have, in that very description, the keys to creating what we seek to identify.
A catch-22 I suppose. Thus, when one starts talking to us, or starts painting abstract art, we'll quickly know "hey, that damn thing is intelligent," but we'll probalby have a hard time knowing that it was "almost that intelligent" for the two years leading up to those oh-so-important first words. :)
But I still like my definitions of intelligence in general usage. ;) if there is any sense in going... but I want to perfectly clear...
Presuming the storage capacity of our brain does not change much genetically, but the external storage capacity of the species DOES change exponentially (and seems to - due to the amount of written works at least)... the more time passes, the more intelligent the species becomes - by this definition.
If I am a member of XX century species, and the catapult guy is a member of... forget it, I don't know which century it was invented in... as a member, I am more "intelligent".
I can accept it as some annoyingly quantum thing where the weight of the group is larger than the sum of the weights of individual members. Which, I guess, is what you say here:
Since man hasn't evolved appreciably in the last many thousands of years, on an individual basis (ignoring actual IQ, and basing any comparison strictly on species genetics and structure) you would be exactly as intelligent, since you would both share essentially the same genetic capacity for intelligence.
I guess where I am most confused is this: Why do we want to create something as intelligent, unpredictable, undefined, etc. as us? To serve which goal? Why was Faust so mad? What is the purpose - just to see if we can do it? Do you know what I mean? Challenging God, maybe... the ultimate goal is to reach God, to create life... So ambitious and Vein.... Let me see if I can clarify my above response.
As a species, man becomes more intelligent. However, the intelligence of the species is no longer simply the sum of the intelligence of each individual person, since we no longer contain all information within our brains. Thus, the intelligence of the species is measured as the combined information stored in all individuals' brains, plus the information stored externally. Therefore, the species now, as compared with the species a few hundred years ago, is more intelligent.
However, when you compare one individual to another, the comparison should really be centered on the information capacity of that individual, and not include any external information he might have reference to. In this manner, a person born a few hundred years ago and magically transported to our time as a baby would not have any difficulty "learning" modern subjects.
I guess the point is that we should be careful when talking about intelligence to articulate whether we are referencing man, or a man. As a species, man becomes more intelligent. However, the intelligence of the species is no longer simply the sum of the intelligence of each individual person, since we no longer contain all information within our brains
! have done my best on storing information inside those 1700cc of my mere brain, just to be able to process and categorize all new information better and to expand my range of possible answers to a problem. The sum of the experiences itself would not do the job, it is the judgment, emotions, motive and objective that makes the blade, and the previously stored information (along with the ability to randomly access and utilise)the sharpening factor. We do agree on most, as you would find, and the dangerous potential of A.I. would be exact as that.
It is quite easy to create A.I. Easier than Windows 2000. 20 million lines of code would be a spare.
What would U do with it, is the question.
I wouldn't make it, no matter what. If I did, I should implement several insticts (just to keep up up and running)
that would collide with the original objectives that I would design it to reach. quote:
--------------------------------------------------------------------------------
Also I believe hooking up a supercomputer as an extra processing unit to a brain would create massive opportunities. if possible, you create some sort of interface between the brain and the computer allowing us to perform massive mathematical equations etc. Things we can't really do on our own... a symbiosis between man and machine would be one powerful tool. The independance, uncertainty, emotions and mind of the human brain, combined with the strength, speed and reliability of a supercomputer... Imagine...
--------------------------------------------------------------------------------
it wouldnt be that simple to just hook you up to a computer it would be more of the computer hooking up to you because it would be impossible to to coincecssly control the computer for one you would need time to get the "feel" of it the computer would be like a growing an extra limb say you had wings wings sergically attached the brain would have no clue how to work them it would take a long time of coincecssly controlling every movement until your uncoincecss part takes over same thing with "hooking" up to the computer and if you put your brain in a vat then hooked it up to the computer you could become self immortilized in the computer. Hey! Relax! That object comes later on another thread. Not here, not now...
View Full Version : Digital Problem
Good book on the subject - Interesting Concept...
|