Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
Jackie and the Brain

By mindpixel in Science
Mon Jun 13, 2005 at 10:52:24 AM EST
Tags: Software (all tags)
Software

Jackie was a very simple computer program that simulated half of a human conversation which I wrote in Visual Basic in December of 1994 as an entry for the 1995 Loebner Prize. At her heart was a look up table that was built up by having numerous people interact with her converstaionally. The look up table consisted of a stimulus, a response and a number of supplementary indexes to the stimulus. The key to Jackie's heart, and her uniqueness, were her supplemental indexes.


When Jackie was in training mode and she was given a new stimulus not in her stimulus index she asked the trainer for a custom response to that stimulus. Thereafter when she was in interactive mode and she was given a stimulus she had seen before, so long as the match was perfect, she retrieved and displayed the perfect handcrafted response that was given to her by the first person to expose her to that stimulus. This is largely how Richard Wallace's Alicebot and kin work today. Given only a very large number of trainers, both Jackie and Alice appear to be human-like. However, aside from extreme biological implausability, there are problems with this strategy.

The first problem with a pure stimulus/response strategy is that there is no common personality across all stimulus/response pairs. Different users are not aware of how they are each handling specific areas of the life of the simulated character. For example, one person may train the system to respond "Yes, I have a cat named Rufus." to the stimulus "Do you have any pets?" while another trainer may train the system to respond "No. I hate cats." to the stimulus "Do you like cats?" - clearly inconsistent. The solution to this problem is to provide personality guidelines to the trainers, but unless the guideline forces a strictly binary response [This is the case with Jackie's little brother, GAC], the guidelines will always have to be as complex or even more complex than the simulation itself.

The second problem with the pure stimulus/response strategy is what I call "match hardness." If the exact stimulus is not in the index, the system fails catastrophically and must evade the stimulus in an Eliza-like fashion. Such systems are extremely vulnerable to being unmasked as simulations by simple binary questioning about common sense aspects of life. There are two obvious solutions to this problem. The first is to inject a very large number of common sense propositions collected via some other manner [This was tried with some success with Alice and data from the Mindpixel project in the Spring of 2005], or to "soften" the stimulus matching system.

Jackie used a number of stimulus match softening techniques that vastly amplified the number of effective items in her primary stimulus index. The first was simply to convert the stimulus to phonetic codes. One of Jackie's supplementary indexes was based on the SOUNDEX algorithm. This converted each word to a standard code that was insensitive to spelling. The effect of this secondary index was Jackie could find a match to a given stimulus even if words in it were spelled incorrectly. Of course, given a large primary index, many spelling mistakes will be in the index anyway, but this phonetic index expansion technique is vastly more efficient and keeps the first problem of response consistency from creeping up again. And of course it is much more biologically plausable than the biological equivalent of a massive stimulus/response table.

A second soft matching technique Jackie used was an additional index of SOUNDEX codes where all the words in each stimulus were sorted alphabetically. This had the effect of stimulus standardization. With this index, stimuli with slightly different word ordering could still be matched and a response retrieved. This was still imperfect as meaning could be lost or possibly unintentionally created in the standardization, but it is much preferable to evading a stimulus altogether.

Finally, Jackie had a secondary index which was the standardized index filtered of high frequency words from a hand coded list, though in theory this should have been machine generated.

The end effect of Jackie's soft matching systems was to amplify the index footprint of every hand coded stimulus in her primary index - she appeared to know a great deal more about life than what was put into her and her behavior became interesting and unpredictable. In fact, the very first time I exposed Jackie to a person other than myself, she shocked me by responding to something I knew I did not train her on.

At the time of Jackie's first exposure to a person other than myself, she was quite small and fit on a 1.44 MB 3 1/2 inch disk. I would train her at night, teaching her about her own life, which was mostly just mine, sex shifted, and take a fresh copy of her to work with me the next day. At the time I worked in the IT department of a large insurance company, as did David Clemens. David was a Japophile at the time, and his first question he put to Jackie was "Do you like sushi?" I expected her to evade that question as I had never mentioned sushi to her at all, but to my surprise she responded "Of course."

I couldn't believe her response and interrupted David's conversation to see what happened. She had a soft hit on the secondary phonetic index to "Do you like sex?" Sushi was phonetically close enough to sex to satisfy her! This was a major revelation for me and I started spending a lot of time looking at her phonetic indexes. It was clear that something profoundly human-like was happening in these indexes. I felt I was capturing a real model of human experience in the topology of the indexes. Similar concepts were clustering in phonetic space.

I thought, wow, if I open Jackie up to the Internet - remember this is 1994 and the web is only months old - I could build a massive soft phonetic index and use it to train a neural network to extract the underlying phonetic space and make a true synthetic subcognitive substrate. The problem was how to synthesize responses and how to quantify the quality of the synthetic responses? The answer I came up with was to restrict the responses to binary.

If we imagine Jackie's phonetic stimulus index as a multidimensional sphere, we can imagine each response as either a black or white point at each stimulus coordinate on its surface - black for false and white for true. Now if we train a neural net to represent this sphere, novel stimuli would be points of unknown value on this sphere and we could interpolate a value from known points near the unknown point, something that would be difficult if the responses were not restricted to binary. The important question is of how many dimensions should this sphere be?

I believe George A. Miller unknowingly answered this question in 1956 when he published the landmark psychology paper "The Magical Number Seven, Plus or Minus Two." Our immediate memory [to use Miller's term] is about seven items long - that is we can recall easily about seven unrelated items from a larger list of items that we see or have had read to us. Thus, we can imagine Jackie's phonetic index to things people can store in their immediate memories as complex fractal pattern on the surface of a seven-dimensional hypersphere.

The most remarkable revelation of all occurred when I tried to visualize this object and figure out why nature would make it seven-dimensional. Could surface area be maximum at seven-dimensions, I thought? That seemed unreasonable. Why would it be? More dimensions would intuitively lead to more surface area, but I had better check just in case. And guess what? Hypersurface is maximum at about 7.25694630506 dimensions.

The revelation that hypersurface was indeed maximum near seven dimensions, and moreover was maximum at a fractional dimension and hence fractal, was obviously very powerful for me. I used it to form what I call the Hypergeometric Hypothesis which states - immediate memories are points on a maximum hypersurface and complex cognition is a trajectory on the same hypersurface. I used this hypothesis as a tool for structuring my initial exploration of real brains.

At first I was quite discouraged when I discovered that the neocortex was six layered in most animals [some have fewer layers, but it is important to note that none have more than six layers]. I had predicted that I would find a seven-layered object in both humans and complex animals and additionally predicted that we should find in the fossil records earlier humans and animals with slightly larger brains than modern brains as evolution would have tried an eight layer system and rejected it in favor of a system with maximum hypersurface and thus maximum possible pattern complexity on it surface. It was hard to believe that we had not yet evolved our seventh layer, so I went digging deeper into neuroanatomy looking for the seventh layer of the neocortex. I found it in the thalamus.

The thalamus forms a loop with the neocortex, called the thalamocortical loop - exactly as one would expect if it were synthesizing one unified seven-dimensional hyperobeject. I was elated when I read that in fact the thalamus is considered by some neuroanatomists to be the seventh layer of the neocortex. The object looked real.

The realness of my object became much stronger when I learned that Neanderthal had slightly larger brains than modern humans and when I learned that there was no other theory that made this prediction or that even acknowledged that the difference could have any meaning at all. It was a glaring fact that science seemed to be ignoring because it conflicted with the idea that the mental uniqueness of modern humans derives from our having the largest brains for our size.

A final prediction of the Hypergeometric Hypothesis is that no matter how advanced a brain is, it should not have a primary loop with more than seven layers. This appears to be true.

It is ironic that Hilary Putnam used Turing's ideas to create the functionalism that dominates cognitive psychology today and which is responsible for the field's near universal ignorance of real brains, as it was the abstraction of Turing's test to a binary geometric form that lead me to make structural and functional predictions for real brains past, present and future.

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
I think of Suicide...
o Never - This is the first time. 13%
o Exceptionally infrequently - my life is almost perfect. 6%
o Infrequently - my life has hiccups, like anyone´s. 25%
o Occasionally - I´ve had some hard knocks. 12%
o Annually - I have seasonal affective disorder. 5%
o Monthly - I´m very sensitive to moonlight. 0%
o Weekly - I read too much news. 12%
o Daily - The news finds me no matter where I hide. 6%
o Hourly - I see the future and it´s worse. 16%

Votes: 72
Results | Other Polls

Related Links
o I
o Loebner Prize
o Alicebot
o Eliza
o SOUNDEX
o The Magical Number Seven, Plus or Minus Two
o Hilary Putnam
o functional ism
o Also by mindpixel


Display: Sort:
Jackie and the Brain | 126 comments (81 topical, 45 editorial, 0 hidden)
Needs edits (none / 0) (#1)
by forgotten on Sat Jun 11, 2005 at 11:07:17 PM EST

Or at a minimum some clarification. I didnt see this in editing, maybe it just passed through really quickly, but here are some things that confused me:

'maximum hypersurface': i thought you were talking about spheres?
'fractal': why did you use this word? I dont think you are using it correctly.
'layers' and 'dimensions': again you are using these terms interchangeably, when they are not.
large brain size: once a brain reaches a certain critical size in animials, there are other more important factors that determine intelligence.

--

hypersurface and fractals. (none / 1) (#4)
by mindpixel on Sat Jun 11, 2005 at 11:18:06 PM EST

Fractal means fractional dimension. It is standard usage. Hyperspheres have hypersurfaces. Google hypersphere, you will see on Wolfram´s site the curve of hypersurface area versus dimension, which is maximu at about 7.3-d. Making it fractal. If you´re really interest...the object is actually a Milnor sphere. Milnor has a nice paper on pasting Julia sets on his page that has a picture of what I think of when I think of this object.

[ Parent ]
i'm sorry, but (none / 0) (#6)
by forgotten on Sat Jun 11, 2005 at 11:32:50 PM EST

you are misuing those terms. I dont think you understand them, and you are certainly misapplying them here. (If there are Milnor spheres of fractal dimension, i have never heard of them, although the idea is quite intriguing to me. in any case, you are making a huge and unwarranted leap by thinking that mathematical fractals will help you understand the brain.)

--

[ Parent ]

on reflection (none / 0) (#8)
by forgotten on Sat Jun 11, 2005 at 11:37:08 PM EST

i dont think there is any reasonable way a milnor sphere of fractal dimension could be defined.

--

[ Parent ]

A Fractal Milnor Sphere... (none / 0) (#9)
by mindpixel on Sat Jun 11, 2005 at 11:40:19 PM EST

Read Milnor´s latest paper and reflect again... http://www.expmath.org/expmath/volumes/13/13.1/Milnor.pdf

[ Parent ]
it contains no such thing (none / 0) (#11)
by forgotten on Sat Jun 11, 2005 at 11:56:51 PM EST

look, if i have misjudged you, i'll be the first to admit it. but i think you are just throwing around terms and connecting them with handwaves. if you think you really have something going here, please, write it up carefully and explain how all this fits together. i suspect if i can't see it plenty of others cant either.

--

[ Parent ]

don't worry (none / 0) (#43)
by mindpixel on Sun Jun 12, 2005 at 05:25:48 PM EST

Hey, the stuff is all new to you. I've been working on this for a decade and am just now translating it from images and math to words...I'll bring you around. Give me some time. This was just my first step in getting a large number of people thinking about the posibility that mind is a fractal self-organizing semantic-affective resonance map on the surface of a seven-sphere. But really, take a look at the Milnor paper. He shows you what a sphere tiled with julia sets looks like, and though sphere does not translate to hypersphere, visually very well, the math is general.

[ Parent ]
i'll tell you what (none / 0) (#48)
by forgotten on Sun Jun 12, 2005 at 06:36:51 PM EST

when your stuff is published in a reputable journal, give me a shout.

--

[ Parent ]

why wait? (none / 0) (#52)
by mindpixel on Sun Jun 12, 2005 at 07:19:02 PM EST

read the prepub: http://www.google.com/url?sa=U&start=1&q=http%3A//www.mindpixel.com/PDF/ mindasspace.pdf

[ Parent ]
Publication date (none / 0) (#105)
by hatshepsut on Tue Jun 14, 2005 at 04:54:37 PM EST

You page says that the article/chapter is to appear in a book published in 2003. Was it published? If so, I suggest you change the text to reflect this. If it wasn't published, why not?

[ Parent ]
a date.. (none / 0) (#107)
by mindpixel on Tue Jun 14, 2005 at 06:14:25 PM EST

Yes, that when it was to appear. I'm not sure what's happening with the book. Last I heard, Grace Peters had just moved to Florida from San Diego where Epstein is, so the logistics broke.

Waiting on the long science publication cycle is the large part of the frustration that prompted me to post my thinking here as a story. I wanted some debate now! I was tired of being the only one on the planet thinking of the brain as a maximum hypersurface, right or wrong. Now that has changed and I am glad for it.

As it stands, the chapter is quite out of date as I have taken the idea much further in the three years since I wrote it.

[ Parent ]
Mind as Space (none / 0) (#54)
by mindpixel on Sun Jun 12, 2005 at 07:27:13 PM EST

'Mind as Space: Toward the Automatic Discovery of a Universal Human Semantic-Affective Hyperspace--A Possible Subcognitive Foundation of a Computer Program able to pass the Turing Test' which will appear in the forthcoming book 'The Turing Test Source book: Philosophical and Methological Issues in the Quest for the Thinking Computer' to be published by Kluwer Academic Publishers.

Abstract

French (1990) identified "subcognitive questions"--questions with answers which unconsciously depend on the whole of normal human experience--as a key mechanism of the Turing Test's ultimate ability to discriminate a person from a simulation of a person. French claims that no machine that has not lived life as a normal human being, in a normal human body, can pass a rigorous Turing Test. Additionally French notes that even though we intuitively perceive intelligence as a continuum, the Turing Test does not, limiting its usefulness in research. The present article describes a possible method for the automatic discovery of a universal human semantic-affective hyperspatial approximation of the human subcognitive substrate--the associative network which French asserts is the ultimate foundation of the human ability to pass the Turing Test--that does not require a machine to have direct human experience or a physical human body. This method involves automatic programming--such as Koza's genetic programming (1992)--guided in the discovery of the proposed universal hypergeometry by feedback from a Minimum Intelligent Signal Test or MIST (McKinstry 1997) constructed from a large number of human validated propositions collected from a large population of Internet users. It will be argued that though a lifetime of human experience is required to pass a rigorous Turing Test, a propositional approximation of this experience can be constructed via public participation on the Internet, and then used as a fitness function to direct the artificial evolution of a universal hypergeometry capable of classifying new propositions. A model of this hypergeometry will be presented that predicts Miller's "Magical Number Seven" (iller 1956) as the size of human short-term memory from fundamental hypergeometric properties. A system that can lead to the generation of new propositions or "artificial thoughts" will also be described. It will be concluded that the artificially evolved hypergeometry can serve as the subcognitive foundation of a robot, or simulation of a robot, that can ultimately pass the Turing Test and that a large corpus of human validated propositions can also be used as an "Automatic Turing Test" to objectively evaluate any current or future claim to artificial intelligence in a rapid, automatic fashion, as well as serving as real world experiential data for traditional symbolic processing systems.

[ Parent ]
Transgressing the Boundaries: (none / 0) (#117)
by DrH0ffm4n on Thu Jun 16, 2005 at 07:45:38 AM EST

Towards a Transformative Hermeneutics of Quantum Gravity

Alan D. Sokal
Department of Physics
New York University
4 Washington Place
New York, NY 10003 USA
Internet: SOKAL@NYU.EDU
Telephone: (212) 998-7729
Fax: (212) 995-4016

November 28, 1994
revised May 13, 1995

Biographical Information

The author is a Professor of Physics at New York University. He has lectured widely in Europe and Latin America, including at the Universit`a di Roma ``La Sapienza'' and, during the Sandinista government, at the Universidad Nacional Aut'onoma de Nicaragua. He is co-author with Roberto Fern'andez and J¨urg Fr¨ohlich of Random Walks, Critical Phenomena, and Triviality in Quantum Field Theory (Springer, 1992).

    Transgressing disciplinary boundaries ... [is] a subversive undertaking since it is likely to violate the sanctuaries of accepted ways of perceiving. Among the most fortified boundaries have been those between the natural sciences and the humanities.
    - Valerie Greenberg, Transgressive Readings (1990, 1)

    The struggle for the transformation of ideology into critical science proceeds on the foundation that the critique of all presuppositions of science and ideology must be the only absolute principle of science.
    - Stanley Aronowitz, Science as Power (1988b, 339)

There are many natural scientists, and especially physicists, who continue to reject the notion that the disciplines concerned with social and cultural criticism can have anything to contribute, except perhaps peripherally, to their research. Still less are they receptive to the idea that the very foundations of their worldview must be revised or rebuilt in the light of such criticism. Rather, they cling to the dogma imposed by the long post-Enlightenmen...

---
The face of a child can say it all, especially the mouth part of the face.

[ Parent ]

Not the fractal I know (none / 0) (#96)
by speek on Mon Jun 13, 2005 at 06:47:33 PM EST

fractal: A geometric pattern that is repeated at ever smaller scales to produce irregular shapes and surfaces that cannot be represented by classical geometry.

As in Mandelbrot set. Does your usage of the word "fractal" relate the the more common (IMO) usage?

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

fractal dimension. (none / 0) (#98)
by glor on Mon Jun 13, 2005 at 08:08:41 PM EST

This is common usage.  There's one example of how to calculate the dimension of a fractal at http://www.jracademy.com/~jtucek/math/dimen.html, but I haven't read it to see if it's right or not.

There was a nice Scientific American article a few months ago about how some of Jackson Pollock's paintings have a fractal dimension.  There is a link at http://www.sciam.com/article.cfm?articleID=000CC6E4-9171-1DC9-AF71809EC588EEDF but unfortunately you seem to have to pay for it.

--
Disclaimer: I am not the most intelligent kuron.
[ Parent ]

Interesting (3.00 / 3) (#3)
by vadim on Sat Jun 11, 2005 at 11:11:28 PM EST

But I still believe this approach is fundamentally mistaken.

I'm very interested in the subject myself, and I'm getting more and more convinced that if we want to produce a good AI it's going to have to pretty much emulate a living organism. Just something that parses strings isn't going to be good enough.

The main problem as I see it is that we humans are very dependent on our environment. In order for a computer to know what a cat is it would first have be able to perceive the 3D environment where the cat moves, and actually see what one looks like. I'm fairly sure that the attempts of building big databases are fundamentally futile, especially since that pretty much all definitions and hierarchies are arbitrary and overlap.

The solutions I see to this: Build it as a robot in the real world, or build it as a bot in a 3D game.
--
<@chani> I *cannot* remember names. but I did memorize 214 digits of pi once.

Pattern Recognition. (none / 0) (#115)
by AMBorgeson on Wed Jun 15, 2005 at 03:46:20 PM EST

actually, a computer can already "see" a cat in 3-d via a dual-lense camera (not sure if that's the actual name, but the set up used for filming in 3-d. It can be done, it's just very expensive) What the computer has trouble wiht is pattern-recognition. Imagine that cat is looking at the camera, and a human operator identifies that object as a cat. Now the cat turns around and wanders away.. it no longer looks like a 'cat' to the computer. Show it several hundred examples of small cats in healthy shape, and finally get it to recognise all cats of that type as 'cats' and it will have trouble if a 25 pound super-fat cat comes waddling through the door. Nor will it recognise a lion as being very cat-like. Finally get it to recognise all that.. and the computer will still have trouble recognising carictictures.. say, like http://twolumps.keenspace.com/ pattern recognition is still our greatest mental tool.

"It takes a Long time to count to '2' in Binary." ~Fourlegged

There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
[ Parent ]

That's not what I'm taking about (none / 0) (#116)
by vadim on Wed Jun 15, 2005 at 04:58:28 PM EST

What I'm saying is that it seems odd to me that so many people seem to be concentrating on one unique part. For instance, take chat bots.

I just don't see how can we ever get intelligence from something that's basically parsing text and spitting answers out of a database. Even assuming we did manage to make an intelligent one, it'd be a deaf and blind entity that would lack any real idea of the world we live in.

IMO, trying to reproduce the whole thing at once would have much better results. Make a robot that can see, hear and move around, and then try to make it smart, instead of building a brain in a vat. If we managed to simulate a human brain separated from a body I bet it would be quite messed up.
--
<@chani> I *cannot* remember names. but I did memorize 214 digits of pi once.
[ Parent ]

OMG (2.25 / 4) (#10)
by Kasreyn on Sat Jun 11, 2005 at 11:52:55 PM EST

from your website:

I am a Hacker. Not reformed.

TERRORIST!!!

The real question is, can Rick Santorum pass the Turing Test?


"Extenuating circumstance to be mentioned on Judgement Day:
We never asked to be born in the first place."

R.I.P. Kurt. You will be missed.
20Q (none / 0) (#12)
by adimovk5 on Sat Jun 11, 2005 at 11:57:10 PM EST

Twenty questions is an internet based game that relies on the responses of users, although the choice of responses is limited. The game uses user responses to filter its answers and make guesses. Because it is based on users responses and not on an exact database, it has the ability to change with time and always remain current.

remain current? (none / 0) (#13)
by forgotten on Sun Jun 12, 2005 at 01:09:16 AM EST

some of us were programming "Animal" in lisp 20 years ago ...

--

[ Parent ]

I beat it first go (none / 1) (#28)
by onix on Sun Jun 12, 2005 at 06:44:28 AM EST

with "tape measure". Pfft.

---
I'm interested in certain of your thoughts. - medham
[ Parent ]
Not a valid rebuttal (none / 1) (#31)
by toulouse on Sun Jun 12, 2005 at 07:24:26 AM EST

I know humans who have a problem with "cat".


--
'My god...it's full of blogs.' - ktakki
--


[ Parent ]
Look (none / 1) (#32)
by onix on Sun Jun 12, 2005 at 07:44:40 AM EST

We all know how these things are supposed to go. Someone puts up a link to an AI. You click through, read the blurb and think "I know just what will scupper this". The computer then proceeds to demonstrate that you are considerably less inventive and original than you initially thought. This didn't happen. I am clearly as brilliant and original as I first suspected.

---
I'm interested in certain of your thoughts. - medham
[ Parent ]
Not only but also (none / 1) (#33)
by toulouse on Sun Jun 12, 2005 at 08:56:59 AM EST

The problem with these self-reweighting neural networks is not that they don't work; it's that they work too well. We've devoted centuries' worth of person-time to developing an agent capable of rigorously maintaining its own mediocrity.


--
'My god...it's full of blogs.' - ktakki
--


[ Parent ]
Interesting game (none / 0) (#90)
by Fon2d2 on Mon Jun 13, 2005 at 11:06:48 AM EST

It had trouble with my first object of "apple" guessing wrong two or three times before finally getting it, and it never got "vitamin". But all the rest it got in about 17 questions. I think it comes down to how applicable the questions are. If the questions are nonsensical to the object it will have a hard time.

[ Parent ]
I suggest that there is real meaning in that (none / 0) (#14)
by SaintPort on Sun Jun 12, 2005 at 03:01:39 AM EST

six (6) is the number of man    think 666

and seven (7) is the number of God.


--
Search the Scriptures
Start with some cheap grace...Got Life?

We're sorry... (2.40 / 5) (#16)
by Arvedui on Sun Jun 12, 2005 at 03:35:00 AM EST

...the number you carved into your sacrificial goat, is no longer in service. Please hang up, and try your ritual again.

[ Parent ]
that is an interesting article, but (none / 1) (#18)
by forgotten on Sun Jun 12, 2005 at 03:56:36 AM EST

i cant see how it makes sense to interpret the number of the beast as caligula... i dont think there was christian movement that early that would have registered with that roman emperor, and i think that John's revelation came during Nero's reign ? but i am no expert and am talking from memory.

--

[ Parent ]

The funny thing is... (none / 0) (#20)
by Arvedui on Sun Jun 12, 2005 at 04:58:02 AM EST

...the Emperor-representing "number-code" interpretation of 666 (or, now, 616) is hardly a recent (pardon the pun) revelation, as I remember reading about it in the back of a local Catholic church newsletter at least 10 years ago, in a "Did you know...?" sort of context. It said the number represented either Caligula or Nero (since saying bad things about the divine Emperor directly was akin to sending out e-mails today about how you're going to shoot the president), and my memory leans towards Nero. But it sometimes almost surprises me how many apparently-devout Christians are caught unawares by the entire concept.

[ Parent ]
yes, it isnt new (none / 0) (#22)
by forgotten on Sun Jun 12, 2005 at 05:19:01 AM EST

i dont read any ancient languages, but my understanding is that many old langauges didnt include a separate characters for numbers, so that some words could be understood as numbers and vice versa.

Aside from that though, the historical persecution of christians by romans has been hugely exaggerated. I'm not sure if it makes sense for any roman emperor to be identified as an antichrist. (except julian, several hundred years later; but he certainly didnt deserve it).

--

[ Parent ]

Hummmm (none / 1) (#44)
by levesque on Sun Jun 12, 2005 at 05:26:28 PM EST

Pythagoras

1, 3, 6, 10, 15, 21, 28, 36 ... and on up to the 36th "triangular" number, which is 666. (Remember the Beast of Revelation - Christians often seem to have difficulties with numbers). The 666th triangular number is 222111

Pythagoras and the Pythagoreans

The number one : the number of reason.
The number two: the first even or female number, the number of opinion.
The number three: the first true male number, the number of harmony.
The number four: the number of justice or retribution.
The number five: marriage.
The number six: creation
...
The number ten: the tetractys, the number of the universe.


[ Parent ]
pardon my goat (none / 1) (#23)
by SaintPort on Sun Jun 12, 2005 at 05:37:42 AM EST

but while even John or his contemporaries may have been focused on the number as relating to a specific ruler, I maintain that the Holy Spirit was referencing...  

Now the weight of gold that came to Solomon in one year was six hundred threescore and six talents of gold,

  -- 1 Kings 10:14 (King James Version)

Antichrist will be able to rule when he controls the world's money.

--
Search the Scriptures
Start with some cheap grace...Got Life?

[ Parent ]

Milnor example (none / 1) (#15)
by SaintPort on Sun Jun 12, 2005 at 03:09:47 AM EST

and as an aside I must remind everyone that...
the word 'hrair' is used for any number greater than four

the Milnor example link...
http://www.gang.umass.edu/reu/2000/curve09.html

--
Search the Scriptures
Start with some cheap grace...Got Life?

What a fascinating article (2.80 / 10) (#19)
by givemegmail111 on Sun Jun 12, 2005 at 04:31:16 AM EST

It almost makes sense for the first few paragraphs, then degrades into abject nonsense so gradually you don't even realize you're being trolled. Of course what's really fascinating is the number of people who voted this up. I can only suppose they only read the coherent beginning paragraphs and skimmed the rest. Either that or they're making some sort of ironic "vote for the obvious troll" statement that's completely gone over my head.

--
McDonalds: i'm lovin' it
Start your day tastefully with a Sausage, Egg & Cheese McGriddle, only at McDonalds.
Rusty fix my sig, dammit!
That's about right... (none / 1) (#21)
by Arvedui on Sun Jun 12, 2005 at 05:14:07 AM EST

Personally, I had gotten to the part when he started talking about 7-dimensional fractal hypersurface maximums when I voted him up, mainly for the same reason Rusty did. By that point in the article, it's hard NOT to realize that words seem to be getting thrown together on a more-or-less aesthetic basis (the totally-unrelated suicide poll is a hint, too), but the first half was fun, and I thought it might be interesting to see if, by some chance, any of the last third of the article made any sense to anyone. If so, then hey, I might learn something! If not, I'm sure people will come up with amusing ways to say as much. Win-Win!

Heyyyyy... what if this whole article was actually generated by the program it's trying to tell us about! ooooOOOOOOooooo...

[ Parent ]

nah (none / 0) (#25)
by SaintPort on Sun Jun 12, 2005 at 05:55:15 AM EST

mindpixel has had a suicide issue quite awhile now.

This article is a real one.

--
Search the Scriptures
Start with some cheap grace...Got Life?

[ Parent ]

Don't forget the idiots like me... (none / 0) (#84)
by Russell Dovey on Mon Jun 13, 2005 at 05:40:32 AM EST

...who read the entire thing, thought "Wow, this guy is teh smart!" and voted +1 perfectly honestly.

"Blessed are the cracked, for they let in the light." - Spike Milligan
[ Parent ]

I don't think it's a troll (3.00 / 11) (#24)
by OmniCognate on Sun Jun 12, 2005 at 05:54:27 AM EST

It's a bit off the wall, certainly, but I don't think it's a troll.

There are a couple of bits that really don't make much sense, like

I could build a massive soft phonetic index and use it to train a neural network to extract the underlying phonetic space and make a true synthetic subcognitive substrate
but I don't think the author can quite be accused of randomly throwing words he doesn't understand together.

It is true that a 7-dimensional hypersphere has maximum hyper-surface area (the terminology is awkward here - it isn't an area, it's kind of hyper-area of 1 less dimension than the sphere itself). It's explained on mathworld. A unit circle has circumference 2*pi. A unit sphere has surface area 4*pi. Apparently, this number increases to a maximum at 7 dimensions and then recedes.

The solution to the equation does give a maximum at 7.25695, and there is such a thing as fractional dimensionality. Whether it can be validly applied here I don't know, but there certainly is such a thing, and it does appear in the mathematics of fractals.

Switching between a 7-dimensional sphere and 7 layers is not particularly odd. The 7 layers are the 7 layers of cortical tissue. The seven dimensions refer to an abstract entity the author thinks the brain is representing. Presumably he is claiming each layer represents a dimension.

There are indeed 6 layers of cortex. I don't know whether anybody considers the thalamus a 7th. In some areas of the brain, particularly the visual cortex, the division of responsibilities between the layers is reasonably well understood and it doesn't look like what the author is describing, but he isn't just making this up.

The article is very coherent to begin with and does describe reasonable techniques for implementing a toy AI.

At one point he mentions using a neural network to interpolate between values. This is a valid and realistic use for a neural network.

There are some excessively big logical leaps in the second half of the article, and I can't say I agree with what the author is suggesting, but I do think he has an idea which he believes has value. I also think he understands the concepts he is referring to a lot better than some people are giving him credit for.



No troll (3.00 / 6) (#30)
by jobi on Sun Jun 12, 2005 at 07:19:48 AM EST

The author has been going on about this for some time now on his blog, and this article was in fact plucked verbatim from a blog posting of his four days ago.

So, the relevant question really is: Crazy or brilliant? ;)

Personally, I think that the article suffers from something representative of many a kook article - it (seemingly) moves faster than the speed of logic, especially in the second part.

Sure, seven is considered a magic number, but how was the seven-dimensionality translated into six-plus-one layers in the traditional sense? And realise that Miller considered 5,6,8 and 9 almost as "magical" as seven.

Also, the sex-sushi connection might just as correctly be described as a bug in the program, Jackie answered wrong and you interpreted the bug as "something profoundly human-like". Now, the only human-like behaviour I see in that incident is your willingness to interpret anything the program outputs as being human-like. It's only a program, it's bound to have bugs.

Anyway, my vote is kook, not troll. With a slight reservation for brilliant genius, just to be on the safe side ;)

---
"[Y]ou can lecture me on bad language when you learn to use a fucking apostrophe."
[ Parent ]

This guy is dead serious (2.50 / 4) (#29)
by StephenThompson on Sun Jun 12, 2005 at 07:11:55 AM EST

First time I read this I thought it was pretty much gibberish.  Then I read it again, and went to mindpixel's blog.   His blog is filled with these ideas http://mindpixel.com/chris/, and he even wrote a book chapter about it http://www.mindpixel.com/PDF/mindasspace.pdf.

There is no doubt that he is serious.

He has made huge leaps of intuition here, and they may be way off the mark.  However if you read his other stuff you will see that there is more foundation for his idea than appears here.

Fascinating.

Dimensions (3.00 / 4) (#35)
by Morkney on Sun Jun 12, 2005 at 12:23:56 PM EST

The hyper-surface area of an n-dimensional sphere is (n-1)-dimensional. So the maximal surface area is given by a 6-dimensional surface.

This can be confusing, since n-dimensional sphere sometimes means "(n-1)-dimensional sphere in n-dimensional space" and sometimes means "n-dimensional sphere in (n+1)-dimensional space." However, mathworld explicitly states that the former meaning is used in this case (the "Geometer's sense").

As a check, you can see that on the graph given by the mathworld page, n=2 has a value of ~6 - obviously corresponding to the one-dimensional surface of length 2*pi.

+1, FP! (2.00 / 2) (#36)
by elver on Sun Jun 12, 2005 at 01:33:22 PM EST

You, Sir, are the true new Douglas R. Hofstadter! I haven't read anything this fascinating on the subject of AI since "Gödel, Escher, Bach: An Eternal Golden Braid".

If you can, pop a link to your writing over to Hofstadter and see what he thinks of it. Me, I'll be looking forward to Your book.



...okay... (3.00 / 4) (#37)
by Back Spaced on Sun Jun 12, 2005 at 02:03:14 PM EST

Our immediate memory [to use Miller's term] is about seven items long - that is we can recall easily about seven unrelated items from a larger list of items that we see or have had read to us.

Okay.

Thus, we can imagine Jackie's phonetic index to things people can store in their immediate memories as complex fractal pattern on the surface of a seven-dimensional hypersphere.

...and that is a complete non-sequitur. I think that memory is likely stored throughout the whole cortex, but the link to Jackie or "seven items" is not clear.

Bluto: My advice to you is to start drinking heavily.
Otter: Better listen to him, Flounder. He's pre-med.

Agreed, but... (none / 0) (#41)
by elver on Sun Jun 12, 2005 at 04:39:38 PM EST

While it's an NS and while the guy seems to be way too fixed on the number seven, I do agree that he should have put more explanation into this.

Maybe he thinks that each bit is on a separate layer? That's possible. Hell, that's even believable. The +/- 2 might be a statistical error -- people thinking about something else during studies or perhaps using some other part of their memory.

So where does the 7-dimensional hypersurface come into all this? Uhm. Well, he could map the 7 layers so that each of them becomes an axis in that... Er. Hyperspace? And where the hell do fractals come into all this? Um. No idea.

This article looks like it was written by someone who has the stuff figured out, though. I wish he'd share the logic behind it all. It might be impossible to "prove" anything, but it would be nice if he were to write an article that explained his logic.



[ Parent ]
stuff figured out... (none / 1) (#42)
by mindpixel on Sun Jun 12, 2005 at 05:00:14 PM EST

I think I do have something important figured out. I am working it into words now, with the help of Kuro5in. Yes, memory is in the whole brain, but... Just jury to think in more dimensions...at any instant in time, the distributed electromagnetic pattern of the entire thalamocortical loop - nearly 100 billion neurons - is a SINGLE point on the surface of a seven-sphere if we consider each layer a dimension. This idea comes from statistical physics... Imagine the standard container of idealized gas from statistical physics, where the state of the entire container is a point in 6n space...where n is the number of molecules and 3d for spatical location and 3d more for a momentum of each molecule, then any state of the gas is a point in 6N space and any evolution in state is a 6N trajectory.

[ Parent ]
Continuing. (none / 0) (#97)
by Back Spaced on Mon Jun 13, 2005 at 08:07:33 PM EST

at any instant in time, the distributed electromagnetic pattern of the entire thalamocortical loop - nearly 100 billion neurons - is a SINGLE point on the surface of a seven-sphere if we consider each layer a dimension.

The state can be considered a point in a seven dimensional space, yes, is you choose the "axis" to be the "layers" of the cortex + thalamus. The is a possible conceptualization, yes, but the sphere relation is not clear. Are you defining a seven dimensional hypersphere whose surface consists of all possible states of the thalamocortex? Great. There's still no connection between that definition and the seven items that the brain can contain in memory, or for that matter, the seven seals of the apocalypse, other than the number seven. Nor do the "seven dimensions of GABA" relate, except, again, for the seven. Methinks that you are seeing patterns wherever a seven pops up, and that this is getting conceptually dangerous.

And why a sphere - is this related to the optimally surface area/volume ratio?

Where do you propose to take this beyond "I have proposed, through definition, the surface of a 7d sphere analogous to the state space of the thalamocortex."

Bluto: My advice to you is to start drinking heavily.
Otter: Better listen to him, Flounder. He's pre-med.
[ Parent ]

One is a genius (3.00 / 5) (#45)
by Novelty Account No 59671 on Sun Jun 12, 2005 at 06:24:55 PM EST

The other's insane

Let me see if I understand what you are saying (none / 0) (#46)
by pHatidic on Sun Jun 12, 2005 at 06:32:46 PM EST

1. There are seven parts of the modern human brain, while neanderthals only had six parts. This could explain why modern humans are smarter, even though their brains are smaller. (Assuming that neanderthals would still be here if they were as smart as modern humans). 2. The sphere thing is relevant because you need to be able to measure the distances between points in order to determine distances between points. While there are many more than seven dimensions, you can only measure the distances between points in the first seven, so these are the only ones relevant to fuzzy logic or neural net logic. 3. If the modern brain has seven parts, each part could store data about a different dimension of the hypersphere. 4. Neural nets can take into account the distances between nodes in seven dimensions, to determine what is most likely to be true when there is no perfect match between input and stored memory. 5. When you tried this with Jackie, it was surprisingly humanlike, leading you to believe that the brain might also process data in a virtual seven dimension hypersphere.

Try four (none / 0) (#47)
by pHatidic on Sun Jun 12, 2005 at 06:34:44 PM EST

  1. There are seven parts of the modern human brain, while neanderthals only had six parts. This could explain why modern humans are smarter, even though their brains are smaller. (Assuming that neanderthals would still be here if they were as smart as modern humans).
  2. The sphere thing is relevant because you need to be able to measure the distances between points in order to determine distances between points. While there are many more than seven dimensions, you can only measure the distances between points in the first seven, so these are the only ones relevant to fuzzy logic or neural net logic.
  3. If the modern brain has seven parts, each part could store data about a different dimension of the hypersphere.
  4. Neural nets can take into account the distances between nodes in seven dimensions, to determine what is most likely to be true when there is no perfect match between input and stored memory.
  5. When you tried this with Jackie, it was surprisingly humanlike, leading you to believe that the brain might also process data in a virtual seven dimension hypersphere.


[ Parent ]
Some answers... (none / 0) (#50)
by mindpixel on Sun Jun 12, 2005 at 06:54:36 PM EST

1. No. I am saying that I think hypersurface is a parameter evolution has been optimising since the dawn of mammals. [the word animals should read mammals.] I believe that the seven-layer thalamocortical loop favored by advanced mammals is seven-layered because it is with that geometry that the creatures achieve maximum electromagnetic hyperpattern seperation. An eight-layer system, like I hypothesize Neanderthal had would have less hypersurface area, and thus would be able to store fewer distinct electromagnetic patterns. I believe that Neanderthal in effect had a lower resolution version of reality that actually cost them more to build, but since evolution cannot see into the future, it had to try the eight-layer system before settling on seven. 2. Distance is important. I believe that all our experiences are overlapping electromagnetic patterns in seven-dimensions and that meaning comes from the global pattern of distances between points of exeprience. This is the geomtrization of experience. 3. No. The entire object is patterned. Information is stored in the omniconnectivity pattern for each point. The pattern is what is important. The dimensionality is only what allows for a maximum of seperate patterns. It is also the reason I think that IBM's NMR quantum computers crapout at 7 qubits. 4. Neural networks are good at partitioning spaces. The neural network I think closest to what our brains are is the self-organizing map, particularly the dynamic time-warping self-organizing map. Google it. It is very interesting. 5. Jackie had behaviors that were clues to a phonetic/geometric representation of thinking. I did not really expect to find any evidence for it being the process by which real brains work. I'm just a hacker. All of these connections to phonetics and geometry I am finding are shocking and beautiful to me. I am here by accident.

[ Parent ]
I mean: (none / 1) (#51)
by mindpixel on Sun Jun 12, 2005 at 06:56:47 PM EST

1. No. I am saying that I think hypersurface is a parameter evolution has been optimising since the dawn of mammals. [the word animals should read mammals.] I believe that the seven-layer thalamocortical loop favored by advanced mammals is seven-layered because it is with that geometry that the creatures achieve maximum electromagnetic hyperpattern seperation. An eight-layer system, like I hypothesize Neanderthal had would have less hypersurface area, and thus would be able to store fewer distinct electromagnetic patterns. I believe that Neanderthal in effect had a lower resolution version of reality that actually cost them more to build, but since evolution cannot see into the future, it had to try the eight-layer system before settling on seven.

2. Distance is important. I believe that all our experiences are overlapping electromagnetic patterns in seven-dimensions and that meaning comes from the global pattern of distances between points of exeprience. This is the geomtrization of experience.

3. No. The entire object is patterned. Information is stored in the omniconnectivity pattern for each point. The pattern is what is important. The dimensionality is only what allows for a maximum of seperate patterns. It is also the reason I think that IBM's NMR quantum computers crapout at 7 qubits.

4. Neural networks are good at partitioning spaces. The neural network I think closest to what our brains are is the self-organizing map, particularly the dynamic time-warping self-organizing map. Google it. It is very interesting.

5. Jackie had behaviors that were clues to a phonetic/geometric representation of thinking. I did not really expect to find any evidence for it being the process by which real brains work. I'm just a hacker. All of these connections to phonetics and geometry I am finding are shocking and beautiful to me. I am here by accident.

[ Parent ]
Some questions (none / 0) (#53)
by pHatidic on Sun Jun 12, 2005 at 07:22:44 PM EST

  1. How would a seven layer loop have greater seperation than an eight layer loop? Is this because the eight layer loop can still only use a maximum of seven dimensions, so one loop is wasted when the space could instead go to the other loops?
  2. I understand what you are saying. However, is it not possible that each of the seven parts of the brain controls one dimension? If not, what is the significance of humans having a seven layered brain besides to point of the difference between us and neanderthals?
  3. "The entire object is patterned. Information is stored in the omniconnectivity pattern for each point." Could you explain this further, I don't understand the terminology.
  4. So the significance of the number seven is just  in the coincidence?


[ Parent ]
More answers (none / 1) (#61)
by mindpixel on Sun Jun 12, 2005 at 08:13:09 PM EST

1. Think of the unified object, not its parts. A seven-dimensioal hypershpere has more surface area than an eight-dimensional hypersphere, all other things being equal. Now consider that neurons are the same size and cost in both seven and eight dimensional brains. Seven dimensional brains can store more patterns for each surface element than can an eight-dimensional brain. This gives us the following evolutionary logic: one dimension was good, two was better, three more so, five wonderful, six just fine, seven amazing and eight worse than seven.

Or think this way, when printing a layered circut board, the more layers you have, the more of the finite interconnecting 3-d space between layers must be used to connect top elements to bottom elements, so eventually the connection overhead makes more layers less efficient than more neurons per layer. Wanna guess how many layers is maximum in the manufacture of printed circut boards in the current state-of-the-art? I leave it as an exercise.

2. Again think of the whole object, not its parts. Think of surface area. 7d has maximum surface area and hence maximum possiblle space for distinct patterns, all other parameters being the same.

3. Imagine a sphere with the most complex self-similar fractal pattern on its surface you can imagine. To know the value of any point in the pattern, you look to the point's local neighbors.

5. I have no idea why nature picked seven dimensions for maximum hypersurface area. It could have been six and then I think Neanderthal would have had seven and us six.

[ Parent ]
Still a bit fuzzy (none / 0) (#71)
by pHatidic on Sun Jun 12, 2005 at 10:28:35 PM EST

Points 1 and 5 I now understand. Points 2 and 3 I am a bit fuzzy on. Particularly:
  1. Are you saying that the neurons exist in 7d, or that the neurons combine to form concepts that exist in 7d? I ask because computer chips are generally thought of as existing in 3d, quantum physics that i don't understand aside, but a computer chip can render concepts in 7d.
  2. I don't understand how a self-similar fractal compares to our brains. Are you saying that our brains are self-similar in multiple dimensions?


[ Parent ]
some more answers (none / 1) (#88)
by mindpixel on Mon Jun 13, 2005 at 10:38:37 AM EST

2. I didn't say that neurons exist in 7-d...but that appears to be the case. There is work now that models the neurotransmitter Gaba as a 7-d manifold. This researcher did this without aparently knowing about hypersurface area. Google Gaba and seven-dimensional, and you will see another example of the seven-sphere popping up in the deepest structure of the brain, another hint at the overall seven-dimensional structure of the entire system. Nature keeps finding the seven-sphere.

3. Yes, our brains are self-similar in multiple dimensions is exactly what I am saying.

[ Parent ]
We-It (Nature) keep finding the seven-sphere? (nt) (none / 0) (#95)
by levesque on Mon Jun 13, 2005 at 01:43:49 PM EST



[ Parent ]
I voted -1 (3.00 / 3) (#74)
by oddity on Sun Jun 12, 2005 at 11:13:04 PM EST

because it makes no sense. But resection it to humor and I say +1 FP. This is funny stuff.

Clarifications? (3.00 / 3) (#92)
by jonradoff on Mon Jun 13, 2005 at 11:57:08 AM EST

An interesting idea.  Can you answer a couple of clarifying questions?

First, when you say that "mind is a fractal self-organizing semantic-affective resonance map on the surface of a seven-sphere" do you mean "semiotic" in place of semantic?

Do you derive an exponent of seven by conjecturing that any semiotic object may be represented by seven parameters (thus, seven dimensions) at any given time?

I don't really see why you lean toward a hypersphere for the topology of the mental-map.  Couldn't it just be an arbitrary 7-dimensional object?

The psychological research that underpins your 7-dimensions argument implies that it might be as few as 5 or as much as 9 for some individuals, and this isn't really represented in your idea.

Also, just because the brain (neurologically) has a certain number of discrete structures doesn't seem to corellate to anything.  I'd suggest that the real structure of the brain is in the synaptic conections between the neurons, not the higher level abstractions.


some clarifications... (none / 0) (#93)
by mindpixel on Mon Jun 13, 2005 at 12:18:34 PM EST

When I said semantic, that's what I meant.

Actually the object is a signal/meteasignal map...both portions are seven-dimensional.

Hypersphere is just the simplist 7-d object. You would need some motivation to add complexity.

I think the psychological data is noisy, like all data.

Actually, the whole brain seems to be organized as seven-spheres, even down to the neurotransmitter level. Google Gaba and seven-dimensional and you will see what I mean.

[ Parent ]
Doesn't make sense (3.00 / 2) (#106)
by jonradoff on Tue Jun 14, 2005 at 06:04:38 PM EST

Just seems like you're drawing on anything that coincidentally contains the number 7, without building toward those conclusions with any real support or evidence.

I don't see any support for a seven-dimensional nature to neurotransmitters.  In the research you suggested I look at, the only seven-dimensional aspect is that the authors of a paper regarding GABA happened to track seven attributes pertinent to circadian rhythms.  You've over-generalized.

I think you'd enhance your research by making less a priori conclusions and less generalizations.


[ Parent ]

Balderdash! (3.00 / 5) (#94)
by MMcP on Mon Jun 13, 2005 at 01:00:56 PM EST

Everyone knows that humans are Cubic forms that rotate a 4 corner face lifetime!

Everyone except (3.00 / 3) (#99)
by Phssthpok on Mon Jun 13, 2005 at 08:26:24 PM EST

lie-educated word god worshippers.
____________

affective flattening has caused me to kill 11,357 people

[ Parent ]
not AI and not science (3.00 / 3) (#100)
by nml on Tue Jun 14, 2005 at 04:03:02 AM EST

i don't mean to be negative, but there are a number of things seriously wrong with this proposal. Firstly, soundex sucks. What surprised me when reading through this was that your reaction to having your program misinterpret the sushi question. Replying with a vague falsehood, while superficially satisfying, only worked because the trained answer was also vague. If the question had have been 'what do you like about sushi?', there are an awful lot of sex/sushi trained answers that would land your program in trouble, where as a normal person wouldn't have any trouble with that question. So i don't think you have presented a very good case for this approach working well in general.

As for your hyperspatial sphere stuff, i fail to understand what the dimensions of the sphere would represent. Text/phonic data is notorious for being high-dimensional, but you're planning to fix the number of dimensions to 'maximise the surface'. How does the data map onto seven dimensions? And you haven't even bothered to fully specify what you'd do with such an index once built, only noting that 'interpolation is hard'. The rest of the article descends into some incoherent and quasi-mystical babble that mentions the number seven as often as possible. Blech.

As I read it... (none / 0) (#122)
by RegularFry on Mon Jun 20, 2005 at 08:47:42 AM EST

with a stinking hangover and thus the relatively high probability of getting something wrong, it works a little something like this: a pre-processing neural net maps incoming phonic data to a 7D hash value. Neural nets being what they are, similar phonics end up with similar (thus physically close) hashes. I don't see *why* interpolation would be hard once you've got the sphere, though - it's only geometry, and an output net could do it for you quite simply (I say this, of course, knowing full well I'll never bother implementing it).

There may be troubles ahead, But while there's moonlight and music...
[ Parent ]
yes, but... (none / 0) (#125)
by nml on Tue Jul 05, 2005 at 10:50:15 PM EST

the basic concept is still flawed. Matching responses to queries based on phonic similarity is doomed to failure, because meaning doesn't follow phonic similarity. Queries using homonyms would fail miserably under this system. Secondly, exactly what would be interpolated for output? The proposal was to train pre-recorded responses. How do you algorithmically 'interpolate' between responses?

[ Parent ]
Phonology and Semantics (none / 0) (#126)
by mindpixel on Fri Jul 15, 2005 at 02:35:48 PM EST

What Spivey did was prove that the Jackie "sex and sushi" coincidence was no coincidence. It was the result of an underlying neural phonological space.

I just today proposed to Spivey that we rerun his experiment using GAC-80K instead of words. Doing so should show the semantic equivalent of his phonological thought bending experiment.

[ Parent ]

principal component analysis (none / 1) (#101)
by schrotie on Tue Jun 14, 2005 at 05:55:40 AM EST

The state space of a neural network with n neurons is commonly considered to be nD (i.e. it has n dimensions). But the network is heavily linked and the activity of neuron depends on the activity of other neurons. Thus the actual dimensionality is usually (depending on the weight matrix) lower.

I don't see any relation between the supposedly seven layers and seven dimensions of state space. A layer has two dimensions. And it is rather far fetched that the state of a whole cortical layer could be reasonably represented in 2000D, let alone 2D.

I'll give an example of research done in my department. The human arm has nine degrees of freedom and operates in 6D space (3 translational degrees of freedom and 3 rotational). When using principal component analysis to analyze the actual movement data of certain reaching movements it turns out that most of the data (don't know the number, over 95% I think) can be covered with three independent dimensions. This does not cover the numerous (21?) muscles in the arm that can be used to control the nine joint degrees of freedom.

However, these three dimensions have to be represented in the brain somehow. We have two arms. They might be coupled in many situations, but it is rather unlikely that both arms are together represented in 3D. And then there are the hands and fingers which are not considered in this setup. And there are legs and eyes, neck, facial expression. Dozens of joints, hundreds of actuators (muscles), plus another brain sized neural network in the stomach. I very much doubt that the state of the human body which is very likely represented in the brain can be represented in seven or less dimensions (well, If you push it you can represent every state of every system as a scalar, but what's the point?). And I did not even consider other mental representations (language, social relationships, math, whatever). You can represent the state of the mind as a point on the surface of a 7D object. You can also represent it as a scalar. But if you want a meaningful representation of the dimensionality of the mind's state space you have to take a complete (as complete as possible) sample of actual brain states and do something like principal component analysis on it. If you end up with a number of dimensions you can still count in a reasonable time (let alone 7) you'll see jaws dropping right out of their sockets in neurolabs everywhere.

The number of cortical layers is not obviously related to the number of dimensions of the state space of the brain. The state space of conscience (whatever conscience is, please spare me that discussion) is very likely a minuscule subset of the state space of the brain. But why it should be 7D eludes me. Especially since I can pick a lot of random representations out my of brains state space and make them conscious. I can e.g. make the state of any joint and many skeletal muscles conscious. And 6D are already reserved for representations of position and rotation of objects in real space. Imagine a red saxophone hanging in a box in midair and playing summertime. 6D for position/orientation, 1D (at least) for color and 1D (at least) for musical note. The music gets louder - 9D ...

Read the criticism on Miller (none / 0) (#102)
by pak on Tue Jun 14, 2005 at 07:03:05 AM EST

Your model seems a bit silly if the only thing it´s going to account for is the size of working memory. Secondly, I think Millers results are questioned even in the basic Psychology 101 canon nowadays. If you are to write about working memory, you might want to read the current literature about it. :-)

er (3.00 / 2) (#103)
by eschatron on Tue Jun 14, 2005 at 11:27:30 AM EST

I'm not really sure if this is a troll or not because you seem to have devoted so much effort, but the flaws are magnificent.

Without getting into a point-by-point battle, the kicker is this. Even assuming that your 7d hypersphere idea is significant, placing a novel stimulus on that sphere is the main cognitive task. This is something that has been learned the hard way over and over again. It's not just problem-solving that's hard, but figuring out what's a problem in the first place. It's not just answering things that's hard, it's figuring out what's been asked. The idea that SOUNDEX plus alphabetic sorting is going to get anything to the right coordinates is just absurd. You're basically suggesting that those two things account for about half of cognition.



K5 Science Page Coincidence (none / 0) (#104)
by mindpixel on Tue Jun 14, 2005 at 03:34:31 PM EST

On the same page two stories earlier than this story appears the story AI Breakthrough or Mismeasure of Machine that is based on the work of Peter D. Turney, who was the editor of Canadian Artificial Intelligence in 1995 who accepted my paper, The Minimum Intelligent Signal Test: An Objective Turing Test, which I wrote because of my experience with Jackie.

Layers? (none / 0) (#108)
by Sgt York on Tue Jun 14, 2005 at 06:16:54 PM EST

Interesting story, especially the part about the phenomes. It opens the door to some interesting ideas in signaling, actually. A bug becoming a feature, in a way.

A question, though : What are these 6 physical layers of the brain you're talking about, that somehow exclude the thalamus? The thalamus has long been known to be a waystation for most signals to/from the body. "Thalamo" is found in the name of many of the afferent tracts. IIRC, the only root that's more common is "spino". It makes little sense that any modern discussion of layers of processing in the brain would exclude that structure.

What are the other 6, anyway?

As for the Neandertal comments, why do you assume they were less intelligent? They were outcompeted because they did not properly adapt their behavior to climate change. This does not mean that they were stupid, though.

There is a reason for everything. Sometimes, that reason just sucks.

Semantic Geometry & Brain Connective Topology (none / 0) (#109)
by mindpixel on Tue Jun 14, 2005 at 07:00:55 PM EST

Dig around here and find the link to my paper 'Mind as Space' and you will see my theory was actually motivated by Claude Shannon's geometric vision of information. [Shannon taught us to see any signal as an N-sphere and David MacKay taught us something was missing - metasignal.]

The neocortex is also called the homotypic cortex [in contrast to the heterotypic cortex] because it is so unifomly structureless. It consists of six thin laminations on the top of the thalamus and the seven structures are connected in a loop.

I don't assume Neanderthal was less intelligent, I PREDICTED I would find a hominid with a slighly larger brain that would have been more expensive while providing less hypersurface. I did not really expect that there would be any evidence that such a creature existed. But, finding that there did recently walk on the Earth a hominid with a slightly bigger brain that science could not adequately explain was a strong motivator for me. I believe Neanderthal failed because their immediate memmories were shorter and hence their whole view of reality was more limited. I said nothing about their intelligence.

[ Parent ]
Meta: We need a new section on K5 (none / 0) (#110)
by Viliam Bur on Wed Jun 15, 2005 at 07:55:03 AM EST

"Pseudoscience"

We need a new section in kuro5hin. (none / 0) (#111)
by Ward57 on Wed Jun 15, 2005 at 11:58:47 AM EST

Troll.

Trouble in first paragraph (none / 0) (#112)
by Alan Crowe on Wed Jun 15, 2005 at 01:04:36 PM EST

Given only a very large number of trainers, both Jackie and Alice appear to be human-like.

The point I wanted to make is that these chatbots are stateless. So as soon as you refer back to something earlier in the conversation you get a weak, generic answer. The natural flow of conversation is a question about the conversation: I thought I mentioned that earlier? The chatbot fails to cope and the illusion is shattered.

So I thought I would go to an online chatbot and work up a couple of transcripts. One would show how if you give it the old yadda yadda, monopolising the conversation and talking at the chatbot, the transcript looks normal, but if you try and engage it in a to and fro conversation the illusion fails.

Unfortunately, I don't have the knack of talking to chatbots. I cannot get a normal looking transcript out of a chatbot even when I'm trying to avoid tripping it up.

Here I try to trip up Alice.

Crowe
Hello
Alice
Hi there!
Crowe
I like arithmetic
Alice
A lot of people like that. Though I like cats and dogs. I like Don't Read Me.
Crowe
I thought I was the only one.
Alice
Where were you the only one?
Crowe
What did I say I liked
Alice
You said "What did I say you liked".
I find myself disturbed by having the conversation break down in the weird fashion.

I try again, hoping to avoid tripping Alice up.

Human: I like cats
ALICE: Me too cats are my favorite animal.
Human: It is the fur, and the purring
ALICE: I've heard of it before.
Damn, I screwed up. The mentions of "fur" and "purr" refer back to "cat" in my previous sentence. Alice doesn't remember that, so cannot cope. I cannot have a conversation within these constraints.

The basic premise of the article is that stateless chatbots are an interesting technology, worth building on. I take the opposite view. Taking to chatbots has convinced me that even the most basic conversation has a logic and flow to it that requires a short term memory containing the salient points of the whole conversation.



stateless chatbots (none / 0) (#113)
by mindpixel on Wed Jun 15, 2005 at 01:52:27 PM EST

the first step toward a continuous system is getting it working discretely. Now, the systems only work for an isolated point in time. The future would see that change by introducing a state memory. It's a problem for a s/r chatbot, but I think not for a system with a spatial subcognitve substrate where every new state modifies that entire global state history of the whole system.

[ Parent ]
Something more considered... (none / 0) (#114)
by mindpixel on Wed Jun 15, 2005 at 02:57:44 PM EST

Here is something that is a little more considered than the blog entry that the above K5 story was based on:

Mind as a Maximum Hypersurface

The Hypergeometric Hypothesis (none / 0) (#118)
by mindpixel on Fri Jun 17, 2005 at 01:53:27 PM EST

Here are my formal hypergeometric claims and suggestions for possible falsification.

you're creepy (none / 1) (#119)
by ShiftyStoner on Fri Jun 17, 2005 at 09:38:04 PM EST


( @ )'( @ ) The broad masses of a population are more amenable to the appeal of rhetoric than to any other force. - Adolf Hitler
Cornell Agrees Mind is a High Dimensional Space! (2.00 / 3) (#124)
by mindpixel on Sat Jul 02, 2005 at 12:55:52 PM EST

A Cornell University press release says that the idea of the brain as a computer has outlived its usefulness. Instead, Michael Spivey and other researchers say, the brain is a "dynamic continuum, cascading through shades of grey". They propose that "perception and cognition are mathematically described as a continuous trajectory through a high-dimensional mental space; the neural activation patterns flow back and forth to produce nonlinear, self-organized, emergent properties". Exactly as I proposed in "Jackie and the Brain" and in my paper "Mind as Space: Toward the Automatic Discovery of a Universal Human Semantic-Affective Hyperspace--A Possible Subcognitive Foundation of a Computer Program able to pass the Turing Test" which I wrote in 2002. It is nice feeling to have experimental confirmation!

Jackie and the Brain | 126 comments (81 topical, 45 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest © 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!