Preview
[aimc.txt v1.1]
NOTE CHANGES: NEW COMPO DEADLINE, DETAILS FOR ENTRY. LOOK FOR
[*] TO MARK CHANGES!!!
+--------------------------------------------------+
|**************************************************|
|* THE AIMC: ARTIFICIAL INTELLIGENCE MUSIC COMPO *|
|**************************************************|
+--------------------------------------------------+
YES THAT'S RIGHT! I, your humble servant and lame-ass composer Rimbo,
am holding a competition all on my own.
"Gee, why?"
Well, ya see, erhm...I have a day job. I am actually a Computer
Science student at the University of Texas at Austin, and right now
I'm doing a project dealing with AI and Music. However, in order to
pull this off, I need a -BUTTLOAD- of musical examples, specifically
formatted for the project. I'm talking about 100 or 200 examples,
here.
"Great. Why should I help?"
Because this project will further human understanding!
"Sure it will."
Because it's for a good scientific cause?
"Keep trying."
Because I'm distributing $50 out of my OWN POCKET two winners--$35 to
the best overall MOD, and $15 to the best "simple song." [to be
explained later]
"Now you're talkin'. What do I do now?"
Read on.
*****************
MAIN CONTEST RULES:
- All songs must use the same sample set (see below for details).
- All songs must be 4-channel Protracker-compatible MODs.
- Enter as often as you like.
- Upload entries to ftp://ftp.cs.utexas.edu/tmp/rimbo
-[*] Deadline is 11:59pm CST, Saturday, December 2, 1995.
- $50 will be distributed among the two winners (see below).
MINOR details:
- All songs must use the same sample set.
In order to simplify things, all entries must use the following sample
set.
WARNING: THE FOLLOWING SAMPLE SET WAS DE-
SIGNED FOR SCIENTIFIC PURPOSES ONLY. ANY
OTHER USE OF SUCH A "GENERAL MOD"-STYLE
SAMPLE SET WILL RESULT IN GENERAL PUBLIC
HUMILIATION BY THE MOD-WRITING COMMUNITY
AT LARGE, BEING BOMBED BY OVER 13,000 UNOPENED
PACKAGES OF WINDOWS NT, AND A HAIRCUT BY
BILL GATES' HAIRSTYLIST.
00 Piano -16va below Middle C 10 Synth lead: "Siner"
01 Piano "Middle C" 11 Synth lead: Square wave
02 Piano 16va above Middle C 12 Synth lead: "Rezonatix"
03 String chord: Major(c e g) 13 Synth lead: Sawtooth
04 " " : Minor(c d# g)14 Organ
05 " " : Sus2 (c d g) 15 Strings (solo)
06 " " : Sus4 (c f g) 16 Saxophone
07 " " : V7 (b f g) 17 Flute
08 Bass (Kick) Drum 18 Synth Bell
09 Snare Drum 19 Xylophone
0A HiHat Open 1A Orchestra Hit
0B HiHat Closed 1B Bongo Drums
0C Crash cymbal 1C Ac. Guitar
0D Toms 1D El. Bass
0E Cowbell 1E Clavinet
0F Rimshot 1F Brass Section
A copy of a 4-channel MOD with these instruments should be at
ftp.cdrom.com in the /pub/demos/music/contests/aimc directory
by the end of this week.
IF you want to use your own samples, or do not like the samples given,
you may sample your own versions--however, you must put them in the
same slot. For example, if you don't like my snare drum and prefer a
tighter snare, you may sample your own, but you must still put it in
location 09. Also, tuned instruments (INCLUDING tuned percussion)
must be tuned to the same pitches; if you want to use your own
Orchestra Hit sample, it must be sampled at the same note and
samplerate as mine. WHEN IN DOUBT, CHECK WITH MY EXAMPLE FIRST.
***
- All songs must be 4-channel Protracker-compatible MODs.
No S3Ms, XMs, MTMs, ULTs, FARs, MIDs, DMFs, etc etc etc. No
exceptions. I'm sorry, I don't have the time to code a converter for
every format out there.
***
- Enter as often as you like.
Shoot, if you can re-orchestrate something you've already written to
fit these requirements, even if it has already been released it may be
entered. For this project I need a tremendous quantity of inputs, and
I'm not particularly concerned about the quality of them. However, if
you are concerned with winning, you may want to concern yourself with
quality.
Also, I shouldn't have to add this, but given recent events: All
entries must be your original work, and not someone else's. If you
re-orchestrate someone else's work to fit the requirements, the
original author will receive credit for the piece. If you claim
someone else's work as your own, you will be mail-bombed by twenty
thousand angry elves.
***
- Upload entries to anonymous ftp://ftp.cs.utexas.edu/tmp/rimbo/
If any of you have no Internet access or have difficulty reaching this
site, there are other ways around it. (1) Send as an attachment to
e-mail. (2) Send as uuencoded document in e-mail. (3) If all else
fails, post to alt.binaries.sounds.mods and send an e-mail to me with
the subject of the postings. My e-mail address is
rimbo@cs.utexas.edu.
[*] When you send it, it is important that:
o Your song have a title of some kind given to it
o You include your full name in the text file. Because
if I tell my professor that I received help from people
named "nolamerz, 31337stud, and Ghandi" I'll get a
very strange look.
o No need for any application for unique numbers or
anything like that, because I don't know most of
y'all anyway, and the other judges most certainly
don't know any of y'all.
o Information on how to contact you (be it phone number,
address, e-mail address is preferable) should you win.
Other than that, I don't care how you get it to me, just as long as you
get it to me!
***
-[*] Deadline is 11:59pm CST, December 2, 1995.
If I don't get your entry by then, there is no guarantee that I will
get it at all; the ftp site is a temporary site, and as I receive
entries I will be moving them to a permanent location. After this
date, there is no guarantee that I will be checking the location
before your entry is erased as part of the system's automatic
clean-up. So do enter by then.
Plus, and most importantly, this gives me a full week to test the program
on them before final examinations, so that I can get a grade for this
semester!!!
***
- Two winners will receive money!
The songs will be judged by the second reader, who is (provided I get
the second reader I want) a graduate student with a degree in music
composition, possibly my faculty adviser, and myself. One winner will
receive $35 (U.S.A. dollars). This winner will be judged on
songwriting quality alone--in other words, solely on how much we like
the song both as something to listen to and as a piece of music in
general. The SECOND song, the one receiving $15, will be a "nursery
rhyme song."
You see, the program needs to be able to LEARN to like music.
When you were a toddler, you didn't listen to pop music--you listened
to Sesame Street and nursery rhyme songs. You listened to very SIMPLE
music. In your teens, you listen to techno/pop/rock/country and other
music--much more complex than what you listened to as a child, but
still simple. As people age, they listen to music which is
technically more and more superior; and in general, most "classical"
music fans are very much older than pop-music fans. A few works (such
as Eine Kleine Nachtmusik by Mozart, or "The Reflex" by Duran Duran)
manage to bridge the age gap by passing off the illusion of simplicity
while actually being very complex--and as the listener ages, they pick
up on things they never really heard the first few times.
Well, in order for this program to work, it needs to be able
to "learn" music. It will start out only being able to understand
very simple songs, and as it listens to more music, it will be able to
understand more and more complicated music.
The practical upshot is this: We need songs that will be
simple. Not even chiptunes--but things which are short, have a catchy
melody, and very simple (if any at all) harmonies and chord
progressions, like "Mary Had a Little Lamb" or "Twinkle, Twinkle,
Little Star" or "Frere Jacques." One simple song will be chosen based
on its suitability for kids (maybe we'll play them for a room full of
3-year-olds and have them judge the songs), and the author of it will
receive the $15 reward.
***
If you have any more questions, feel free to e-mail me at
rimbo@cs.utexas.edu or at rimbo@ccwf.cc.utexas.edu.
**********
A few words about my project.
My favorite bit of AI research is machine learning. To me, that is
what makes something "smart" is its ability to draw in information and
make conclusions based on that information.
My goal in the future is to have a program which, given NOTHING, could
eventually be able to speak with human beings one-on-one. Given no
knowledge of the English language.
The benefit of using music (specifically MODs restricted as I have
done) is that, unlike music, a computer can read it easily, and
there's many more restrictions on the ideas built-in. Also, it's a
much less concrete medium. Most importantly of all, music lacks the
discrete meanings which are found in your typical English sentence.
What I'm hoping for this program to be able to do is to begin by
hearing simple music and slowly be able to progress to more advanced
forms. One thing which I believe is the prime source of our
subjective nature of music judgment is what the song can "teach" us.
There is a paper by Marvin Minsky on the subject which I didn't
necessarily agree with, but explains how a piece of music "teaches" a
theme, and I will upload it to the AIMC directory on ftp.cdrom.com if
you're interested.
But there is also an assertion made that says that we can't learn
anything until we almost know it. That's why kids don't listen to
classical music--it's too far above their knowledge of music in most
cases; it's also why adults stop listening to pop music; they've
learned everything such simple music teaches. It's like going back to
7th grade for them.
This is what I hope to mimic; I want the program to be able to
understand music the same way we do, and as time progresses, be able
to understand more music.
So, this is the project, and the contest with the project. Happy
Trackin'!
Jimmy Rimmer
rimbo@cs.utexas.edu
rimbo@ccwf.cc.utexas.edu
Rimbo of Lucid and ACiD
10/24/95
UPDATED 10/31/95