home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Ian & Stuart's Australian Mac: Not for Sale
/
Another.not.for.sale (Australia).iso
/
hold me in your arms
/
Nexus
/
nexus.software
/
fuzzy.interface
< prev
next >
Wrap
Internet Message Format
|
1994-05-22
|
28KB
From: hiscdcj@lux.latrobe.edu.au (Dwayne)
Subject: Re: Massively Fuzzy
To: lamontg@mead1.u.washington.edu (Lamont Granquist)
Date: Thu, 19 May 1994 01:05:21 +1000 (EST)
Cc: leri@pyramid.com ( leri), nexus-gaia@netcom.com (meta-list nexus)
In-Reply-To: <9405172234.AA101450@mead1.u.washington.edu> from "Lamont
Granquist" at May 17, 94 03:34:04 pm
I'm going to cross-post this thread to nexus-gaia and leri
if people get massively pissed off about it, I'll stop.
You have been warned.
Lamont Granquist blew a bubble, which danced and sang:
:
: > Over on the nexus list we had a flurry of discussion a while back about
: > a sort of generic news/mail/thingy-server, which would function as a WWW
: > device, allowing you to read mail and news, etc. One of the ideas was
: > that people could 'vote' on posts they liked, which would then get sent
: > to the server, and the more votes it got, the higher priority the post
: > would recieve.
:
: wild, i sat down over last summer and did a halfway serious attempt at
: trying to figure out how to go about designing something like this...
Really? Can we have access to your stuff?
: i didn't have the idea to make it WWW-based, though -- i didn't really
: know what WWW was capable of back then...
Well, mainly the idea of imbedded hypertext links is the key, I think,
because then you could do fun things like designate an 'attractor' as
such, which is just a list of links, listing a bundle of
posts/docs/whathaveyou which orbit a particular attractor, as such.
: > It seems to me that the rulespace that would be produced by this sort of
: > mechanism would go a good way towards defining just what the memespace
: > leri floats in is made up of, or other lists....
:
: this might be interesting. i wonder if something along these lines could
: eliminate the need for heirarchial newsgroups.
This was the major attraction. That and that instead of moderating
things, or forwarding articles to an entire list, you just link it into
whatever attractor/s you feel it fits into, rate it fro thingy to
thingy+n (..10, A..Z, whatever) for *each attractor* (I'm going to use
this terminology until someone suggests something better) and leave it
at that.
Those interested in it will see, those not won't. Yet all will be part
of the same community.
There are already 'virtual newsgroups' about. There is a newsreader
which scans for all occurrences of a keywaord, and creates a nwesgroup
based on it, and the newsgroup rules can be distributed. Forget the name
of the reader though :-(
: there would just be one
: large info-pool, and then the nets would extract the information out of
: it. to 'change a newsgroup' you'd merely pick up one of the "group
: neural nets" and then you could either modify this on your own, or hard
: link to the group neural net and kind of let yourself be blown around
: based on what the group thought was interesting...
Yup. And if there was some way of boiling the list neural net into
something less abstract, then it would make a great intro to the group,
you'd just look at the memetic attractors the group/s orbit, and decide
if you'd like it.
Agents could prowl around and report back on similar groups whose neural
net spec are similar to yours. Global search for 'friends' as it were.
I found leri because of the review in High Weirdness By Email, the
phrase 'electric commune' rang clear to me, yet what if I never saw that
text?
: > You could also possibly rank people, because posts that most of the
: > group found interesting would get higher rankings (this could be biased
: > towards prolific posters, who knows) but I think that this might be a
: > *very* iffy thing to do.
:
: i don't think you should do this.
Neither do I, but it's a possibility.
Academia would love it in journals......
: each post should be judged on its
: merits -- although i could see that any individual algorithm would probably
: want to use the past performace of the author as a criteria. i don't think
: it would be good to make this "official" though and compile public
: statistics on it.
I think it would be bloody horrible, myself.
: i think that people should make their own minds up on
: who is worthy in their own opinion and that it should be somewhat obvious
: as to who is considered worthy in the groups opinion without resorting to
: compiling statistics.
Yup. Too true. And, of course, it would then affect the quality of
people's posts. If I knew that everyone thought that *everything* I
wrote was wonderful (they do already, but anyway) I'd be unsufferably
smug.
If I thought that everyone hated everything I wrote, well, instant
lurker, at best.
: > Hey, if we had a leri-net, then people would just feed their personal
: > rule-list into the general rule-list, and rules which are duplicated
: > would be reinforced.
: > Hmmm, this sounds better the more I think of it.
:
: this does start to sound better and better. it also seems to be more true
: to the model of "memetics" than newsgroups are...
Newsgroups are a function of technology, exactly the same as books.
They in no way mimic the way people operate. It's about time we had
something *did* mimic the way people operate.
Wish I could program well enough to create this...
Dwayne.
if we do not hold on to hedonism, we will lose our souls. -- mordwyn
.sig under construction internet---> hiscdcj@lux.latrobe.edu.au
irc--------> ddraig on #leri in undernet
From: hiscdcj@lux.latrobe.edu.au (Dwayne)
Subject: Re: Massively Fuzzy
To: lamontg@mead2.u.washington.edu (Lamont Granquist)
Date: Fri, 20 May 1994 02:58:47 +1000 (EST)
Cc: leri@pyramid.com ( leri), nexus-gaia@netcom.com (meta-list nexus)
In-Reply-To: <9405190219.AA92314@mead2.u.washington.edu> from "Lamont
Granquist" at May 18, 94 07:19:27 pm
Lamont Granquist blew a bubble, which danced and sang:
:
: i'm not exactly sure what this has to do with nexus...
nexus is about building communities. This would be a useful tool.
also, it was first mentioned there (well, I first saw it there) and not
all of the people on the list are on leri, so I thought cross-posting
might be useful. But, like I said, if it shits people, I'll stop.
: > Really? Can we have access to your stuff?
:
: It was all written down and its all trashed now. I was trying to figure
: out a good method to collect the votes and some stuff like what flexibility
: the server should have and what kind of data structures you'd need. I
: wasn't going for something really professional using RSA fingerprinting
: to authorize the votes and stuff, just something which could easily be
: hacked into nn or rn by users...
I do know that lynx allows input, well, I presume HTML allows input, so
you could vote using that. I *presume* that you can reply to stuff
through lynx/mosaic, so therefore reading news through WWW-ware is
possible, so then *voting on* things on a WWW-site is possible.
I'm talking *completely through my hat* here, I haven't played with
HTML, lynx or mosaic as anything more than a user, although I'll get my
own account as soon as I fill out the form, so I'll be playing with
HTML next week...
: I also never got very far...
Neither have we so far.
: My idea was to have a "complete newgroup server." The server would have
: the capability to archive posts and provide easy info-server access to
: it (gopher/WWW/ftp/etc),
I presume that this could be done with email as well...
: it would also collect "votes" and store them
: individually. Then people could download a list of message-IDs based on
: votes, download the articles based on votes, or sub to a mailing-list
: which sent message-IDs or articles based on votes.
Or grab the top x messages by whatever criteria you liked, votes being
one of them.
Most number of votes, highest tally, most 10/10 votes, etc.
: The criteria used for
: votes would be fully customizable by the infoconsumer and would allow one
: to (for example) recieve a mailing list of a newsgroups based on articles
: which recieve >50 "popular" votes or which are voted on by a list of
: individually selected "moderators."
This was pretty much the original idea we had on nexus. Hmmm. Parallel
evolution.
: It could also allow for the easy
: creation of "virtual moderated newsgroups" which would simply be
: suggested moderators for infoconsumers to use as their criteria, or as
: starting points for their criteria.
Can anyone explain how the extropian reputation market works?
I wonder if people could market their votes?
You'd get paid to recommend posts to people.
We have movie critics now, why not info-critics.
Wow. Hmmm.
: Of course searching via WWW/gopher/etc
: should ideally be capable of using voting criteria as one part of the
: search expression. I was also mulling over ways to use/incorporate
: NNTP/NOV into this to make it easy to use from NNTP capable newsreaders.
This means something, but not to me. I presume that this would then be
usable via rn/nn/tin/etc.....
Can't you just hack another header?
X-vote:
1/10::hiscdcj@lux.latrobe.edu.au+2/10::bignob@important.com+9/10::newbie@clue
less.com
etc.
damned vi wordwrapping.
Naaaah. that'd be too cumbersome. And a central voteserver would get
hosed if it tried to cope with all of the net.
Still, I guess seperate sites could keep their one vote lists. Hmmm.
: At some point, it would be nice to have it fully integrated into
: newsprograms... You'd fire up your newsreader, which would log-on to your
: local NNTP host, then you'd either pull down a list of votes from the
: local NNTP server, or else you'd query a vote server to get the list of
: votes. If an article was not found on the NNTP server due to it being
: expired, the NNTP server could attempt to query the newsgroup archives
: to get a copy if it hadn't been cancelled. Voting would be handelled
: automagically by the newsreader, etc... You'd have to find the right
: client/server and local-information/distributed-information balance, but
: it should work...
Yeah, the default would be 'zero out of whatever vote' so that you
didn't have to stuff around with voting against crappy articles, but if
you were really impressed with it then you would vote for it.
Something I have been mulling over for a while is that people write a
shitload of crap on usenet, and the occasional gem is read, passed on
some people mail you and say 'good post' but that's about it.
You could perhaps tie a vote in with digicash, and pay people in cents,
or tenths of cents, or something.
If you wrote an amazing article somewhere, then you'd get highly
recommended, AND you'd get a reward in a more material sense. This
appeals to crass commercialism, sure, but it also means that starving
artists who are brilliant writers would get some support from the world
community.
I have no problem with paying someone one cent if I read a great
article. If I read a lot, my net.connection might cost me an extra
dollar or two a month.
But if 10,000 of the people on the net like the article, then the author
would suddenly discover that they were a hundred dollars better off or
something.
Make use of that Arts degree, guys :-)
: I shudder to think of the CPU load and disk hogging this would cause if it
: wasn't written efficiently... =)
Hmmm. Yeah, I guess so. Implementation side of it is beyond me. I can
crap on about these ideas because I have no idea if it is possible or
not, I just know what I would like to see happen.
: > : i didn't have the idea to make it WWW-based, though -- i didn't really
: > : know what WWW was capable of back then...
: >
: > Well, mainly the idea of imbedded hypertext links is the key, I think,
: > because then you could do fun things like designate an 'attractor' as
: > such, which is just a list of links, listing a bundle of
: > posts/docs/whathaveyou which orbit a particular attractor, as such.
:
: this sounds good -- so each attractor would be a hypertext page to the
: links?
Ummm. Yeah. So, on leri, we have threads, which are mainly subject
titles, I guess. But, instead, we would have a page of links, and the
links could be dynaically presented, so that the highest rating posts
appeared on the top of the list (which would probably make the high
ratins self-supporting. hmmm.), if you chose to view it that way.
I'd make it so that as each new attractor was entered (the page was
displayed by your browser for the first time) it asked you how you anted
to see it, highest-rated first, lowest first, by age, etc, much as elm
allows me to sort my mailbox by poster, length, subject, mailbox order,
etc.
: > This was the major attraction. That and that instead of moderating
: > things, or forwarding articles to an entire list, you just link it into
: > whatever attractor/s you feel it fits into, rate it fro thingy to
: > thingy+n (..10, A..Z, whatever) for *each attractor* (I'm going to use
: > this terminology until someone suggests something better) and leave it
: > at that.
: > Those interested in it will see, those not won't. Yet all will be part
: > of the same community.
:
: well, i'm not sure that you'd need to submit it to any particular
: attractor... ideally you just post it -- anything with MDMA in it would
: likely get picked up by the Ecstasy attractor(s), etc...
Oh, yeah, but I mean, instead of forwarding articles....
shit, look, I'll show you.
I'm about to cross-post an article here. It's on a big pagan thingy in
the US. Now, everyone who subs to the list will see it. It is of
interest to the pagans, to those interested in
alternative-community-type gatherings, performance art, etc.
Now, instead of forwarding it to the list, thus creating a massive waste
of bandwidth as 100-odd copies go winging about the world, I just link
it's URL into the pagan-leri-attractor, the
performance-art-leri-attractor, the alternative-leri-attractor, etc.
Or the performance-art-attractor:leri/nexus-gaia or something.
I've just fwd-ed it to two seperate lists, but all I've done is created
a pair of links. Very little bandwidth is used, and people who aren't
interested in it won't see it.
I, personally, would probably subscribe to ALL of the leri-attractors,
because I'm an info-junky, and all of the nexus-attractors, but you
wouldn't have to. You would get EXACTLY as much or as little info as
you could deal with.
And, you could perhaps make it possible to have people link *other*
people into other attractors, so that if you thought someone was
interested in something, you could just pass them the link.
And you could still mailbomb (attractor-bomb) deadshits :-)
: *wiiiiild* thought just occured...
:
: what you're going to wind up doing is something like taking the Ecstasy
: attractor and linking in the "me" attractor, which will be a special
: post-filter which will promote articles which are replies to things you've
: written, or questions which you'd like to reply to....
Yup.
So you'd get anything on ecstacy, *plsu* anythign people thought was
related to ecstacy.
You could screen out votes by people who's knowledge you thought was
bunk, or who recommended crap in the past.
Rather than *actively recommending things to people* by forwarding,
mentioning, etc, you'd be PASSIVELY recommending things to people.
If you thought highly of something on drugs, for instance, I might well
read it, because I respect the knowledge you have shown in the past on
drugs. But, assuming I thought your stance on politics sucked (no idea,
it's the first example I thought of) I'd discount anything based on
that.
Or your knowledge of history, or basketweaving, or whatver.
This is how people work. We don't have killfiles (well, sort of) where
we completely discount *everything* someone says, but also, just because
someone is an expert in mediaveal courtly romance doesn't mean you
should take their pronouncements on packet radio as gospel.
Similarly, you could link to *attractors* and killfile certain people
*on that subject only*........
: damn, this is getting cooler the more i think about it...
Tell me about it. It ebbed and died on nexus. I hope some of the
people there pick up the thread. Wish I could program.
: now i just gotta figure out how to program neural nets and how to do them
: efficiently on von neumann machines...
Hmmmm. Yeah. What about parallel processing? I have no idea what I'm
talking about, so I'll shut right up when it comes to implementation.
I'll just rattle on about how I think it should work, and let people who
can do this implement it. Wish I had a job doing this, this is fun.
: > There are already 'virtual newsgroups' about. There is a newsreader
: > which scans for all occurrences of a keywaord, and creates a nwesgroup
: > based on it, and the newsgroup rules can be distributed. Forget the name
: > of the reader though :-(
:
: 'nn' does this. its the newsreader i use, and this is more or less the
: kind of thing i had in mind to begin with...
Aha. Yes, I've heard of this with nn. They don't have nn on lux (grrr)
and I have nn on my linux box, but it's not connected to the net.
*sigh*.
: > Yup. And if there was some way of boiling the list neural net into
: > something less abstract, then it would make a great intro to the group,
: > you'd just look at the memetic attractors the group/s orbit, and decide
: > if you'd like it.
:
: yeah, if you could precipitate some kind of description out of the neural
: net code automatically... that could be placed as the first item in the
: hypertext document for the attractor.
Yes, this is exactly what I had in mind. Can this be done? Can you say
to the neural net "give me the rule-set upon which you sort posts" and
then feed it in to another neural net?
Rather than a list, you'd have a sea of posts, circling each other, with
no real structure, but simply the *relative* structure which your
neural net assigns to it.
Everyone would see a different list, as every post would be rated
differently, as we would assign more or less value to each person
voting, and so each vote would be weighted against what you thought of
that person, and what you thought of their past judgements....
Hmmmm.
You could get your neural net to take note of how you vote on a post,
and it could compare *your* votes with *every other person's vote* on
that post, and thus decide how closely they think *as regards that post*
After a while the neural net woud start to realise that you think along
the same lines as person x, and perhaps say to x's neural net "psssst...
pass me along anything x rates highly that I haven't seen"...... you
could get your neural net to search for similarly-minded individuals,
and then you could get in touch with them. Wow, social filters in
reverse, rather than keeping people *out*, you'd meet people you could
possibly never ever meet otherwise. Hmmmmm.
Is this possible with current technology?
Why do I study history? Gaaaaah.
: > Agents could prowl around and report back on similar groups whose neural
: > net spec are similar to yours. Global search for 'friends' as it were.
: > I found leri because of the review in High Weirdness By Email, the
: > phrase 'electric commune' rang clear to me, yet what if I never saw that
: > text?
:
: right, but it would be easy to have your attractor look for posts on
: similar attractors... those items would grap a lot of attention and
: then could be dealt with by the group...
Yes. Links from my neural net to yours, and vice versa. If your votes
on, say, food consistently matched mine, then an article on a
completely seperate meta-attractor (newsgroup/mailist/whatever) on food
which you liked could be passed on to me.
So a newsgroup would not so much be a central address where people
posted to, or a central *thing*, but it would be a nebulous grouping of
like-minded people and their ideas in what they do and don't like. Hmmm.
: > : i don't think you should do this.
: >
: > Neither do I, but it's a possibility.
: > Academia would love it in journals......
Well, publish or perish and all that. Peer-reviewed journals, you write
it, they tell you it's crap and rfuse to publish it, or they think it's
great and they publish it and people review it. Much the same, but more
immediate, more personalised, and much more sense.
: > : it would be good to make this "official" though and compile public
: > : statistics on it.
: >
: > I think it would be bloody horrible, myself.
:
: "having a chilling effect on public discourse" is a phrase that crosses
: my mind....
Yerp. It was just an idea. I'd do something horrible to anyone who
did it.
Much like going to a party and telling everyone what a terrible fuck
that person, yes, that one there, in the blue top, is.
I've seen this done. I did something horrible. :-)
: > If I thought that everyone hated everything I wrote, well, instant
: > lurker, at best.
:
: right, but this happens today without any kind of neural net filters
: and such...
Oh, sure, but only in one direction, really. It'd be nice to be able to
read a heap of stuff which I will love.
I get stuff forwarded to me all the time, by all sorts of people. It's
great, most of it I'd never see otherwise. This way, it's automatic,
and rated, so you'd see the stuff you will *really* like (assuming that
the neural net is capable of sussing you out) down to the stuff you kind
of like and have the time for today.
I do that now.
I get a shitload of mail. I do a scan for 'dwayne' then I do a scan for
'nexus' then I start reading the other stuff (yeah, yeah, I have
procmail, i'm reading the docs to know how to use it)....but it is still
sorted by mailing list, rather than what I like.
This is a function of *technology*
Books are sequential not because we can absorb information most
efficiently that way, but because that's about the only way to impart
info cheaply.
Same with the net. This is why hypertext is such a good idea, it matches
the way our (well, mine, anyway) brains work. We jump from idea to idea,
as it catches our fancy and our interest.
: > Newsgroups are a function of technology, exactly the same as books.
: >
: > They in no way mimic the way people operate. It's about time we had
: > something *did* mimic the way people operate.
: >
: > Wish I could program well enough to create this...
:
:-)
Wish I didn't repeat myself so often :-)
Does anyone else have anything to add to this?
Can it be done?
*should* it be done?
Dwayne.
if we do not hold on to hedonism, we will lose our souls. -- mordwyn
.sig under construction internet---> hiscdcj@lux.latrobe.edu.au
irc--------> ddraig on #leri in undernet
From: Mark Grant <Mark.Grant@isltd.insignia.com>
Date: Thu, 19 May 94 18:51:03 BST
Subject: Re: Massively Fuzzy
-----BEGIN PGP SIGNED MESSAGE-----
hiscdcj@lux.latrobe.edu.au (Dwayne) said :
> Can anyone explain how the extropian reputation market works?
At the moment it doesn't, I think the person who ran it left the list,
and it was sold to someone else (who also left the list), so it's been
shut down for the time being. The basic idea was that people could
register 'reputations' of things (people, technologies, groups, etc) on
the list for which they received a number of shares (and 'money' maybe
? I forget). Then people would buy and sell shares in the reputations
depending on how 'reputable' they were thought to be.
It worked reasonably well, but had the problem that a few people spent
much of the time trying to make money through market fluctuations, so
that while on a long-term basis the reputations matched reality, on a
short-term basis they could go all over the place. Also, not enough
people used it regularly.
I still have the documentation that it sent out in response to 'help'
requests around somewhere if you want me to send you a copy.
> You'd get paid to recommend posts to people.
> We have movie critics now, why not info-critics.
Yep, looks like being a big business in the 21st century. I think it
might be better to vote on threads rather than individual articles.
> If you wrote an amazing article somewhere, then you'd get highly
> recommended, AND you'd get a reward in a more material sense. This
> appeals to crass commercialism, sure, but it also means that starving
> artists who are brilliant writers would get some support from the world
> community.
That sounds like a very good idea ! Roll on the digital banks....
> I've just fwd-ed it to two seperate lists, but all I've done is created
> a pair of links. Very little bandwidth is used, and people who aren't
> interested in it won't see it.
However, with Usenet or Leri, you can't tell who's read it easily. But
if, say, you're working for a company that discriminates against
Pagans, when the article is centralised like that it would be quite
possible for the company to set up a system to log everyone who
accessed that file. Worse, if it was considered 'politically
incorrect', it would be quite possible for the government to log it
too...
> Yes, this is exactly what I had in mind. Can this be done? Can you say
> to the neural net "give me the rule-set upon which you sort posts" and
> then feed it in to another neural net?
No, but you could dump all the internal variables in the network and
copy them into another network. One of the problems with neural
networks is precisely that you *can't* work out how they're judging
things.... and it's quite possible that if badly trained some inputs
would cause them to have a fit.
> After a while the neural net woud start to realise that you think along
> the same lines as person x, and perhaps say to x's neural net "psssst...
> pass me along anything x rates highly that I haven't seen"...... you
> could get your neural net to search for similarly-minded individuals,
> and then you could get in touch with them. Wow, social filters in
> reverse, rather than keeping people *out*, you'd meet people you could
> possibly never ever meet otherwise. Hmmmmm.
Actually, I suspect there may come a time when we'll just program the
machines to read and write the stuff for us as well and go party
instead... 8-) !
Mark
-----BEGIN PGP SIGNATURE-----
Version: 2.3a
iQCVAgUBLdum/mhZrcRdG1w1AQE8mwP/YWBZZd2weTpreR4jWtS7LMqkbc7N3UiV
eM5AoKn+frhZ9sp2yUyJoWOAU9oX4HFrzikQBKejJ86FDK+X0utoi1vli5NrQT1K
+A1vHU0xxgx/a209i+YREQp75hSa1Of8Y+WcyYXlkmV4kAp1zEg5Dfpht/rxwvqc
lNqTYW5GNEI=
=k+f9
-----END PGP SIGNATURE-----
From: "Torin/Darren/Who Ever..." <torin@netcom.com>
To: hiscdcj@lux.latrobe.edu.au (Dwayne)
cc: lamontg@mead1.u.washington.edu (Lamont Granquist),
leri@pyramid.com ( leri), nexus-gaia@netcom.com (meta-list nexus)
Subject: Re: Massively Fuzzy
In-reply-to: Message from hiscdcj@lux.latrobe.edu.au (Dwayne)
of "Thu, 19 May 1994 01:05:21 +1000."
Date: Sat, 21 May 1994 17:54:43 -0400
Dwayne spoke unto the world and said:
>There are already 'virtual newsgroups' about. There is a newsreader
>which scans for all occurrences of a keywaord, and creates a nwesgroup
>based on it, and the newsgroup rules can be distributed. Forget the name
>of the reader though :-(
It's strn. Oh, and strn also supports http connections...
Torin
From: "Torin/Darren/Who Ever..." <torin@netcom.com>
To: hiscdcj@lux.latrobe.edu.au (Dwayne)
cc: nexus-gaia@netcom.com (Nexus-Gaia List),
leri@pyramid.com (Leri Land -- Our Net Commune)
Subject: Re: Massively Fuzzy
In-reply-to: Message from hiscdcj@lux.latrobe.edu.au (Dwayne)
of "Fri, 20 May 1994 02:58:47 +1000."
Date: Sun, 22 May 1994 04:43:09 -0400
Dwayne spoke unto the world and said:
>I do that now.
>I get a shitload of mail. I do a scan for 'dwayne' then I do a scan for
>'nexus' then I start reading the other stuff (yeah, yeah, I have
>procmail, i'm reading the docs to know how to use it)....but it is still
>sorted by mailing list, rather than what I like.
BTW, the latest (minus 1) version of procmail implements scores. While
it is static and not self-teaching, it definitely uses scores. Say,
thanks, I might be able to implement something for leri and cypherpunks
so that when I see something that I like, I type in a score and perl
scripts manage the score files for me...Hm...
Torin