 |
|
|
 |
|
|
|
|
Pennywise |
| | | | | | posted 7/1/2003 18:06 |  | | | | | | | | | | |
nettron2000 wrote @ 7/1/2003 4:56:00 PM:
In my opinion before we can build an artificial "mind" we first have to build an artificial brain. It appears to me that todays artificial entities, chatbots included, are just plays on words. Real thinking doesnt consist of simply taking a sentense apart, analizing its different constituants and searching a database for an appropriate reply.It goes much deeper than that, several neural modules are involved in cognition.
| |
I agree with most of your comments, especially those about the apparent lack of promise of chatbots.
One thing it might be beneficial for you to understand is that the mind and the brain are two separate entities. I believe that eventually, neurology will pull through and show how the brain causes the mind, but the odd prediciment we're in now is that we can observe both the brain and the mind and we don't see the precise relationship between the two.
This divide between the mind and the brain gives rise to several different sciences with basically the same underlying goal. I would say that the big three are:
Neurology -- Studying the brain and trying to understand how it works.
Psychology -- Trying to understand the mind and how it works.
Artificial Intelligence -- Trying to rebuild various parts of the mind and eventually a whole mind.
What's the part that's missing? Apparently the science of building an artificial brain, as you point out; however this ends up falling under Artificial Intelligence. If you assume that the brain is the reason for the mind, then one way to create a new mind is to build something similar to the brain. This ends up being the connectionist camp of Artificial Intelligence, the people responsible for such contributions as Neural Networks. Unfortunately, the connectionists can't move much faster than the neurologists do without inserting their own speculation as to how the brain works. And in the end, when we finally do understand how the brain works, there will probably be some much more efficient way to implement it than with trillions of software or silicon based neurons.
Personally, that's why I find AI so interesting -- Neurology moves too slow (reverse engineering a trillion neurons is going to take a while) and Psychology is fairly subjective (assert input, observe output, make some assumption of what happened inbetween). AI allows us to build our own mind and see if or how well it works. Essentially guess and check. Not ideal, but better.
| | |
| | |
nettron2000 |
| | | | | | posted 7/1/2003 21:29 |  | | | | | | | | |
| | | Last edited by nettron2000 @ 8/8/2003 9:04:00 PM | | |
| | |
zyvex |
| | | | | | posted 7/2/2003 00:02 | | | | | | | | | | | |
nettron2000 wrote @ 7/1/2003 4:56:00 PM:
[1] In my opinion before we can build an artificial "mind" we first have to build an artificial brain. [...] several neural modules are involved in cognition.
[2] Im not a programmer but i beleive the Brain analizes the context in which words appear, the "why" of it ,if you will, in making its decision for a response. And "context" covers quite a braod spectum in cognition and could very possibly involve all brain modules, emotions included.
| |
[1] What are the "several neural modules involved in cognition"?
[2] Context is not necessarily a causal hypothesis; the guess people make about the cause of an event -- "the why of it". Context is broader than that.
| | |
| | |
nettron2000 |
| | | | | | posted 7/2/2003 15:46 | | | | | | | | | |
| | | Last edited by nettron2000 @ 8/8/2003 9:01:00 PM | | |
| | |
Pennywise |
| | | | | | posted 7/2/2003 20:13 | | | | | | | | | | | |
nettron2000 wrote @ 7/1/2003 9:29:00 PM:
I know its bin said that the mind and the brain are two separate entities, the so called mind- brain barrier.
[1] But think of it this way; can the mind exist without the brain?
[2] And im not sure if you have read or heard that the brain is the "wetware" on which the mind runs. I dont agree with that, i beleive the mind comes about as a result of the brain,as outlandish as it might seam.
[3] The processing principles behind the operation of the brain are completely foriegn to how it is done in digital computers. I think connectionists are on the right track its just that they are on the wrong train. They can remotely be likened to someone trying to build an airplane with flapping wings.
[4] As for the so called "big three" there is really just one and thats neurology. Artificial attempts at replicating intelligence ,namely neuromorphics, just follow what ever breakthroughs or discoveries that are made in anatomomical studies of the brain(Neurology). Many attempts at guessing the underlying "secrets" to intelligence have bin made over the years, but you must agree they are pretty lame.
| |
[1] In theory, yes. Ask your local priest what happens when you die. Pay attention to how your soul (and your mind) don't cease to exist.
Perhaps you should read about dualism. The following links should help:
http://ai.ato.ms/MITECS/Entry/crane2.html
http://cognews.ato.ms/1057126544/index_html
[2] Can you clarify the difference between the brain being hardware (or wetware) with the mind being software and your concept that the brain gives rise to the mind? It seems like both are just ways of saying that the brain is physical and the mind is logical.
[3] I totally agree with the flapping wings analogy... I've used it myself. What finally got us flying machines? A wind tunnel. When are the neurosciences going to get their wind tunnel? Who knows, but if it's not in this century are you just going to stand by and idlely wait?
[4] Actually neurology, psychology and artificial intelligence (and other cognitive sciences) all learn from each other and teach each other. I agree that neurology is more of a hard science while psych and AI are more subjective, but to think that neurology is going to be the ONLY answer is a bit premature.
Perhaps an analogy would help. I hand you a hard drive and say "figure out what this program does". There is no documentation for what's on the hard drive and there is no dicernable text on the hard drive. You may spend decades going through the hard drive bit by bit, learning how all those bits inter-relate and what they mean to each other.
Now imagine that I give you a computer with that hard drive in it. You can see that it's running windows, what happens when you move windows around, what happens when programs are opened and closed, etc. You don't know how a window on your monitor relates to those bits on the hard drive, but you'll now know to look for bits that give you information about opening windows and launching programs. Perhaps you can even seek out hard drives with specifically corrupted bits and see how that effects the programs that are running. But maybe you could go one step further -- you could create your own CD that attempts to be like the hard drive. There would be problems (you can't write to CDs) and it would take a while (good luck creating an entire OS and applications over night), but eventually you would come to understand how to create that hard drive. And then maybe you could sell hard drives for a living.
Obviously the hard drive is your brain, the software is your mind and/or behavior, tumor testing and abnormal psychology is corrupted bits, and creating CDs is artificial intelligence.
I can see how your comments could be inflamatory and lead to long debates. Your brash and uninformed disregard for points of view that you don't yet understand is bound to get people riled up. Rather than asserting your own point of view, you should spend time trying to understand both sides of the story -- no matter how painful that may seem in the begining, it will be worth it in the end. You do not truely know that you are right, until you know why everyone else is wrong.
| | |
| | |
nettron2000 |
| | | | | | posted 7/3/2003 01:14 | | | | | | | | | |
| | | Last edited by nettron2000 @ 8/8/2003 9:05:00 PM | | |
| | |
Ribald
[Guest] |
| | | | | | posted 7/3/2003 01:30 | | | | | | | | | | | |
nettron2000 wrote @ 7/1/2003 4:56:00 PM:
I posted a comment to a few AI sites awhile back in which i received alot of flak. I thought i'd post it here just to see if the responses ( if any ) would be similar.
In my opinion before we can build an artificial "mind" we first have to build an artificial brain. It appears to me that todays artificial entities, chatbots included, are just plays on words. Real thinking doesnt consist of simply taking a sentense apart, analizing its different constituants and searching a database for an appropriate reply.It goes much deeper than that, several neural modules are involved in cognition.
Im not a programmer but i beleive the Brain analizes the context in which words appear, the "why" of it ,if you will, in making its decision for a response. And "context" covers quite a braod spectum in cognition and could very possibly involve all brain modules, emotions included.
Ok let the flaming begin!
Net.
| |
You are painting in rather broad strokes. When you say "mind" I am assuming that you are meaning an intelligent entity that is functionally similar to a human?
I believe that there is a big difference between constucting an artificial intelligence, and building an artificial human mind. To emulate a human mind we would have to understand the human mind enough to model it. I agree that will necesitate understanding of the brain.
I also agree that the digital computer metaphor is prone to misleading conclusions.
But building a "brain" also requires knowledge that we do not yet posess.
Once something is understood it can be modeled. The medium at that point is not so important. If you are trying to discover how something works by gross replication, then yes, the medium is important.
But I don't think the real endeavor is to recreate a human mind in digital medium, is it? I think an artificial intelligence of autonomous functioning level would be adequate for some of us.
| | |
| | |
Pennywise |
| | | | | | posted 7/3/2003 05:55 | | | | | | | | | | | |
nettron2000 wrote @ 7/3/2003 1:14:00 AM:
[1] So you are saying that the "software", the mind , is the soul?
[2] You must be a programmer, am i right? You are trying to compare the brain to a digital computer, how that the software in a computer can be separate from the hardware on which it runs and how this can be likened to the thought processes in the brian being a separate entity.
[3] im sorry but you are wrong. I 'll explain later
[4] the "wind tunnel" for the neural sciences is the brain. BTW the "wind tunnel" for flight was birds. Read up on the works of Lilenthal which is what the Wright brothers and others did in thier analogy of the mechanics of flight.
>"are you just going to stand by and idely wait?"<
[5] Im an avid circuit designer.
[6] I agree. Premature? and you dont think that present day attempts at programming AI on digital computers is premature?
[7] I thought your analogy with taking the computer apart would have opened your own eyes.
[8] I think i would rather take the thing apart and study it and ask questions like; what processing priciples is it using? is it using the same processing principles that i know? this hardware looks unfamiliar and i cant replicate it can it be modelled? whats this wire for? heh. See what i mean? atleast i have a chance.
| |
[1] Again, read up on dualism. I can only lead you to the water, I can't make you drink.
[2] Yup, but that doesn't negate my point. My terms, physical and logical, may be directly from the computer domain but the concepts apply perfectly to brain and mind. This doesn't mean that the brain is "hardware and software", I understand neural plasticity and the modularity of the brain which don't relate well to a digital computer analogy. This only means that the brain is of substance -- it can be touched, held and is physically tangible -- where as the mind is not.
[3] Telling me I'm wrong and not telling me why is going to make for a rather boring conversation. Keep in mind that I could say the same to you.
[4] The "wind tunnel" for controlled flight was the "wind tunnel", invented by the Wright brothers. I understand that pigeons influenced the twisting design of the wings, but they would have never been able to verify the design without a wind tunnel. (Especially without killing any more of their investors). I guess that's an entire other debate.
[5] VHDL or Verilog? I've actually done some circuit design myself, but only on a XC2V2000E FPGA.
[6] You seem to think that AI is attempting to surpass neurology -- and sometimes it might be. But many parts of AI are meant to augment neurology. For example, in this study where AI has shown neurologists how the brain could understand emotion:
http://www.newswise.com/articles/2003/7/COMPVISN.OSU.html
(Other examples are left as an excersize to the reader.)
[7] And thus the problem of argument by analogy... You've completely overlooked the similarities because you're so blinded by the details. I'm partially to blame for using "computers" in the analogy when I should have known that you would rally against that one word. Feel free to substitute "computers" for "complex system X", where we can see both the implementation and the result of that system. Perhaps a UFO would be a better example -- your argument is that we should only study a UFO by examining the only one that has ever crashed on Earth. You don't think that we should ever attempt to fly that UFO or build duplicate UFOs (at least not until we _fully understand_ that first one... at which a point we're acting as a photocopier, not a learning mechanism). You're overlooking the value of the other forms of experimentation.
[8] The chance of understanding every bit on the hard drive without seeing those bits in action is exactly as good as seeing the bits in action without ever seeing the bits, given that the number of bits is significantly large.
To put it another way, if I were to disassemble some modest piece of code, not even something large like an OS, and strip it of it's symbols it would take you years to figure out. I would suggest that goes up exponentially with program size. If I were to do the same thing and actually tell you what the code does, it might only take you months.
But really, dismiss Psychology and Artificial Intelligence if you want to. It's at your own loss that you throw away your tools.
| | |
| | |
Pennywise |
| | | | | | posted 7/3/2003 05:58 | | | | | | | | | | | |
Ribald wrote @ 7/3/2003 1:30:00 AM:
But I don't think the real endeavor is to recreate a human mind in digital medium, is it? I think an artificial intelligence of autonomous functioning level would be adequate for some of us.
| |
That's a good point too. Human level intelligence isn't necessarily the goal of AI. Factory automation and good games (aka - weak AI) are goals as well.
| | |
| | |
Ribald
[Guest] |
| | | | | | posted 7/3/2003 06:21 | | | | | | | | | | | |
Pennywise wrote @ 7/3/2003 5:58:00 AM:
That's a good point too. Human level intelligence isn't necessarily the goal of AI. Factory automation and good games (aka - weak AI) are goals as well.
| |
An AI of the level of a dog would be a smashing success. Certainly would provide all of the financial rewards you could want if placed in a clever robotic body.
Also, I don't believe I have seen any more robust headway gained in the ANN hardware realm than has been attained in digital software.
But please correct me if I have fallen behind the times again...
| | |
| | |
AI heretic
[Guest] |
| | | | | | posted 7/3/2003 20:45 | | | | | | | | | |
Ribald wrote @ 7/3/2003 1:30:00 AM:
But I don't think the real endeavor is to recreate a human mind in digital medium, is it? I think an artificial intelligence of autonomous functioning level would be adequate for some of us.
That's a good point too. Human level intelligence isn't necessarily the goal of AI. Factory automation and good games (aka - weak AI) are goals as well.
Of course human level intelligence is NOW not the goal of symbolic AI. Why do you suppose that is?
Symbolic AI has shifted its focus from its original goal of creating a general intelligence machine to a far less ambitious branch of computer science dealing with the creation and analysis of effective data structures and algorithms, yet completely overlooking basic cognitive architecture and the underlying hardware of processing.
They dont need to re-assist old algorithms or creat new ones, they need an whole new outlook.
| | |
| | |
Ribald
[Guest] |
| | | | | | posted 7/3/2003 20:53 | | | | | | | | | | | |
AI heretic wrote @ 7/3/2003 8:45:00 PM:
Of course human level intelligence is NOW not the goal of symbolic AI. Why do you suppose that is?
Symbolic AI has shifted its focus from its original goal of creating a general intelligence machine to a far less ambitious branch of computer science dealing with the creation and analysis of effective data structures and algorithms, yet completely overlooking basic cognitive architecture and the underlying hardware of processing.
They dont need to re-assist old algorithms or creat new ones, they need an whole new outlook.
| |
Who is "They"? I assume that "they" are the opposite philosophical camp than what you consider yourself to be in.
There is no "symbolic AI" club or organization that I am aware of. There are a number of groups and individuals that are still focused on "strong AI" approaches to creating human level intelligence.
I assume from your post that you hold this magic new outlook? Please do share it....
And of course share the examples of success to corroborate that this new outlook is more effective than any other...
| | |
| | |
zyvex |
| | | | | | posted 7/3/2003 22:31 | | | | | | | | | | | |
nettron2000 wrote @ 7/2/2003 3:46:00 PM:
>[1] What are the"several neural modules involved in cognition"<
Heh .You should have put a smily face after that quote. I hope you werent expecting me to answer that seeings how it is still under much research.To put it in the words of Alan" they dont tell me everything " ( or something to that effect)
There are practly thousands if not more modules in the cortical coulmns of the neo cortext alone.
For a better understanding of "modules" heres a page that explains it somewhat.
http://psycprints.ecs.soton.ac.uk/archive/00000420/
>[2] Context is not necessaarily a causal hypothensis; the guess people make about the cause of an event---"the why of it".Context is broader than that<
Could you ellaborate on what you mean?
I didnt say the brain makes a decision for a reply based _solely_ on the context in which it occurs. Please re-read my post.
| |
[1] In context of the given article, you're original assertion doesn't make sense.
[2] Context is a form of partial-pattern recognition. Creating a hypothesis for "why" something happened can be part of context, but this hypothesis is NOT required. Context is more a matter of noticing and remembering sensory correlations.
| | |
| | |
AI heretic
[Guest] |
| | | | | | posted 7/3/2003 22:42 | | | | | | | | | | Symbolic AI is the idea of using an ordinary computer to represent knowledge in the form of some programmed algorithms ( or a set of rules to be followed to arrive at an answer in a finite number of steps).
The other "camp" ,as you put it, are those that use specialized hardware to do the processing, there are no algorithms in the computer sence of the word, processing is acheived via an artificial neural network, in which knowledge is distributed over a vast network and is represented by the "strength" of connections similar to how it is done in the brain.
Do a search on Hugo de garis Cam Brain, NARS non-Axiomatic reasoning system.
These are the types of cutting edge projects that should receive more attention and funding.
| | |
| | |
Ribald
[Guest] |
| | | | | | posted 7/3/2003 23:08 | | | | | | | | | | | |
AI heretic wrote @ 7/3/2003 10:42:00 PM:
Symbolic AI is the idea of using an ordinary computer to represent knowledge in the form of some programmed algorithms ( or a set of rules to be followed to arrive at an answer in a finite number of steps).
The other "camp" ,as you put it, are those that use specialized hardware to do the processing, there are no algorithms in the computer sence of the word, processing is acheived via an artificial neural network, in which knowledge is distributed over a vast network and is represented by the "strength" of connections similar to how it is done in the brain.
Do a search on Hugo de garis Cam Brain, NARS non-Axiomatic reasoning system.
These are the types of cutting edge projects that should receive more attention and funding.
| |
I would suggest that you ignore Turing and Von Neumann's work in suggesting that you can not model the same kind of hardware that you mention, but in software.
"in which knowledge is distributed over a vast network and is represented by the "strength" of connections similar to how it is done in the brain"
What support do you have for this statement? We do NOT know how information is represented in the brain, that is what makes emulating the brain so darn difficult.
I asked you to show me some greater success, but you have not.
I DO agree that experiments like CAM should be better funded.
As Yaki likes to say, the proof is in the pudding. To date all of the existing AI based tools are a product of software.
If there were any projects out there that were based upon a real model of the human brain I would be more supportive. The fact is that our ANN models do not reflect the accurate working of neurons in the brain because we do not know what that is.
But this shouldn't be a zealotry issue, and only becomes one when we go around saying that we know the "right way", and everyone else is wrong.
If shown proof that anyone has the "right way" I will be the first to jump on board.
All I see in looking at approaches like CAM is that they are adherents to the "more" philosophy. More connections, more nodes, more processor power, more rules, more this more that. It all boils down to the hope that more will suddenly become what less didn't achieve.
| | |
| | |
Pennywise |
| | | | | | posted 7/4/2003 00:59 | | | | | | | | | | Suddenly this all makes sense. This isn't about AI vs Neurosciences, it's about the classic hardware vs software battle, and someone is a little bitter because software is getting all the funding.
The reason people aren't trying to implement AI in hardware yet is because they haven't figured out how to make it in software yet. Who wants to spend at least a quarter million dollars in mask costs alone everytime they come up with a new idea?
It's been time proven that concepts start off in software and as they become solidified and economically feasible, they get moved to hardware.
| | |
| | |
zyvex |
| | | | | | posted 7/4/2003 01:09 | | | | | | | | | | | |
Pennywise wrote @ 7/4/2003 12:59:00 AM:
Who wants to spend at least a quarter million dollars in mask costs alone everytime they come up with a new idea?
| |
There are some pretty fast, capacious FPGA's on the market now.
| | |
| | |
AI Heretic
[Guest] |
| | | | | | posted 7/4/2003 01:38 | | | | | | | | | | >I can see how your comments could be inflamatory >and lead to long debates. Your brash and >uninformed disregard for points of view that you >don't yet understand is bound to get people riled >up. Rather than asserting your own point of view, >you should spend time trying to understand both >sides of the story -- no matter how painful that >may seem in the begining, it will be worth it in the >end. You do not truely know that you are right, >until you know why everyone else is wrong.
>Pennywise<
Do i sense some hostlity here? Couldnt that statement also apply to yourself?
I must tell you that ive bin studying and researching AI for quite some time. Tho i dont consider myself a programmer, ive done my share of programming in the past but after gaining some understanding of the complexities involed in cognition i dropped the idea realizing i needed something more ,alot more.
No ive have never used VHDL, Verilog or FPGA. I was thinking more along the lines of circuits that emulate the function of a real biological neuron.In particular, Adaptive threshold elements and floating gate technology used in the new field of neuromorphics. Of course i dont have access to facilities to build neuromorphic circuits in VLSI, most of my work is in VLSI CAD drawings and simple electronic threshold comparator demonstration circuits that i use to verify certain parameters.
But thanks for askin.
| | |
| | |
AI Heretic
[Guest] |
| | | | | | posted 7/4/2003 02:26 | | | | | | | | | | >The reason people aren't trying to implement AI in hardware yet is because they haven't figured out how to make it in software yet. Who wants to spend at least a quarter million dollars in mask costs alone everytime they come up with a new idea?<
You are just alittle bit uninformed. The reason that more advanced versions of neural nets have moved to hardware is because large nets cannot be simulated on computers. The learning time for a large simulated neural net, beyond a few hundred nodes, explodes and becomes unmanagable.
Neural nets involving thousands of adaptive and trainable silicon neurons with millions of interconnected nodes are already in the works and some have already bin fabricated in VLSI.
Why do i have to point this out doesnt anyone search the web for info. and i thought i was the one that was supposedly "mis-informed".
| | |
| | |
AI Heretic
[Guest] |
| | | | | | posted 7/4/2003 03:22 | | | | | | | | | | "in which knowledge is distributed over a vast network and is represented by the "strength" of connections similar to how it is done in the brain"
>[1]What support do you have for this statement?
[2] We do NOT know how information is represented in the brain, that is what makes emulating the brain so darn difficult.<
>[3]I asked you to show me some greater success, but you have not. <
[1] Thats the basic foundation behind the operation of a neural net. I suggest you do like ive done thousands of times, do a Google search for info on artificial neural networks, im sure you will get thousands of hits and you will learn alot about nets that you never heard before has i have.
[2] Not only do we not know exactly how information is represented in the brain but connectionist have a hard time trying to figure out how information is represented in Nets they built themselves. Shouldnt this tell you something?
[3] www.google.com ,its a free download and you can add it to your browser window as i have.
| | |
|
|
|
|
|
