Autochamber: interview Henry Warwick with Christina McPhee

INTERVIEW:

HENRY WARWICK

Christina McPhee and Henry Warwick

On semiotics of sound from data

Henry Warwick is an interdisciplinary artist who specialises in digital

media. Equally at home in digital video, electronic music and digital

imaging, his work examines contemporary society as dominated by the impacts,

both inspiring and catastrophic, of technology, geopolitical economy,

demographics, and culture on human society. His works are held in private

collections around the world. He has an MFA inInterdisciplinary Art from

Goddard College. He lives in San Francisco, California.

CM: Essentially it remains a problem of topology. How do you 'show' or

create a 'listening' environment which condenses diurnal data fields into

condensed topologies that are metaphors of the decline of photosynthesisis

from August to December?

The part I can’t quite see is how to preserve the autochamber sequence of

meanings stretched over a live data dynamic.

HW: You discuss a question of topology re: mapping / moulding the data

provided in such a way that is humanly readable and understandable. Whether

this uses Max or Ableton is another question, quite apart from the

topological issue.

CM: I mentioned "live data".

HW: There is no live data in autochamber.

Live, i.e., synchronic data is characterised by its fluctuation in a

continuous manner between a data generator and a data receiver, such that

the data is only an experience of 1:1 synchrony between the generator and

receiver. There might be a short delay (such as the speed of sound, or

light, or electricity over wire, etc.) or even a long delay (where data is

buffered and then sent) but the point is, there is no replication possible

as it is experiential.

singer -> (sound) -> listener

is live, synchronous.

singer -> (sound) -> (recording device) -> (media) -> (sound) -> listener

is not live. It is diachronous. The media can provide identical (or nearly

identical) data every time, like this:

singer -> (sound) -> (recording device) -> (media) -> (sound) -> listener

(media) -> (sound) -> listener

There are mixes between the two, such as:

composer -> (media, sheet music) -> (ensemble) -> (sound) -> listener

the media is diachronous, but the performance of the ensemble in performance

is synchronous to the listener / receiver, but diachronous to the composer

as it is a mediated relationship via the sheet music.

CM: Autochamber has a similar structure. It is not a live structure - it is

not synchronous at all?

HW: Think of it this way:

OBJECT (Autochamber)

GENERATOR (Satellite and ground machines record natural data / autochamber

machines record data of CO2 etc. of prairie)

MEDIA (Excel spread sheets / Excel spread sheets)

MEDIA READER (software that converts data to sound directly / my eyes and

hands entering data into processor controllers in software)

MEDIA (a CD or DVD)

LISTENER

I spent a lot of time typing stuff and moving GUI components in a program

that does a certain number of things to process sound. We are using natural

data sources, i.e., data sources that are not precisely predeterminable by

any known mathematical algorithm, even if their over all value can be

predicted with little accuracy but generally correct in aggregate. There are

some that can model approximate the behaviour, but these models are

EXTREMELY complex and require supercomputers, and even still, they only come

up with approximations

Autochamber's natural data source is a controlled experiment that produced

data of a specific character. If you ran the same experiment next year, you

would likely get very similar data, as CO2 production always decreases from

Summer to Winter (hence the prediction of general data values in aggregate),

as do all the different life processes of vegetation in the prairie.

Elements of Cage

Cage used a similar process for many of his works. The (idealised) structure

of Cage's typical process is like this:

composer, Cage -> (random generator, I CHING) -> (media, sheet music) ->

(ensemble) -> (sound) -> listener

or, to quote Cage:

the sounDs

of the bUgle

were out of my Control

tHough without

my hAving

Made the effort

the wouldn't have been Produced.

(Cage, J. "M" page 30)

This is actually a bit of a conceit, because any given score generated by a

randomising system (in his case, an elaborate coin toss) is subject to

approval or disapproval. This conceit is a necessary one. If Cage tossed a

coin, and (by some random fluke) it produced the theme to Deutschland Uber

Alles, I doubt that he would have kept that set of coin tosses.

We are using data from outside ourselves, created by natural phenomena

(Cage: coin toss, myself: autochamber data) in order to generate or process

an audio environment for an installation or performance, and we have no

predetermined control over the data itself: we simply control how the data

is used (Cage

arranging it for notation to control performers, myself using it to adjust

parameters of audio software processors).

There is an inverse correlative to this: ultra-strong encryption systems

using natural systems (radioactivity) to generate random number sequences.

Part 2: Differences

The difference between Cage and [autochamber] is that Cage's fundamental

device may be indeterminate, but he uses the indeterminacy as a tool, like a

brush, to fill in the music. Autochamber uses data derived from natural

phenomena, and the chaos and patterns within, as controlling devices central

to the object itself.

This is especially underlined in Autochamber, where the data is used to

manipulate processors that are acting on a field recording of the data

generators themselves, making the piece self-referential and

self-identified.

The form of the work is derived from recordings of the autochambers in situ,

that are processed by the data generated from the autochambers themselves.

How the data was used becomes the interesting point, as it reflects the

sensibility of the art and the artist. I use the datasets to inform my

toolsets to process the audio in order to make my creative statement. In

this way, I am not tied to a given toolset, or a given dataset. I'm using

data from Excel spreadsheets as interpretable datapoints to control audio

processing in an application.

Initial Conclusions

1. autochamber, and Cage work from similar places: using unpredictable data

to make diachronous datasets that control performances.

2. Cage uses the data in order to fill works, while autochamber use these

data itself as a part of the content.

video.

5. Making something "human scaled" is fairly simple by re-assigning time

values. The diachronous nature of the source data permits this.

A mimetic strategy

CM: Tell me again what time frame you were working with, is it the full 3

months fall into winter (hence yielding the frenzied activity early on and

then quiet at the end save for the lovely creak of the robot in the wind?)

HW: The first 4 minutes (the loud fidgety part) is derived from the first 4

days of the data.

The second "wind" part is from the last 4 days.

The data supplied re: CO2 has a 2hr interval. So each sample sets a 5 second

interval in the music, which makes one minute = one day.

Using this method, each month = approx. 1/2 hour, so a 3 month season = 90

minutes or so of audio. A 4 month season = 2 hours.

It's an intriguing problem: since I haven't the time to build a custom app,

I used something off the shelf (Ableton Live and various VST plug ins I've

DL'd from the net). The thing is, these apps are not friendly to raw data -

they're made for musicians and sound designers for their intuitive use as

instruments, so the nomenclature and valuation and organisation of the

program reflects musical interests, not data analysis. So, I have several

hundred data points: how does something like the CO2 data get integrated

into, say, the bass response?

Well, I saw that the EQ has a 15 dB range, ±.

So, I roped off certain dB to correspond with values 1 - 10. So, values 0

through 1 = -15 through -12, and so on. So I would sit with a little "cheat

sheet" and look up the value and apply it.

Example:

BASS RESPONSE, EQ:

seconds: 5 10 15 20 25 30 35 40

value: 5.534 6.533 6.011 6.993 7.052 7.543 6.232 5.322

dB: 0 +3 +2 +4 +5 +6 +2 +1

The interesting part is, audio is logarithmic - every rise of 6 dB is an

apparent "doubling" of loudness. So the extreme ends of the scale are quite

extreme. This is good, as it provides a lot of contrast, and the contrast

enhances the ability to distinguish parts in it.

The "art" of this exercise has been selecting the data ranges from the

columns that are most useful. Sometimes the CO2 data isn't as useful as the

data from something else, such as the average temperature, or the radon

levels.

In that case, I would use the data from that column, but I would always use

the data from the proper *time range* to maintain some integrity of the

data. This lends an element of mimesis and portraiture: odd traditional art

notions that one would rarely associate with electronic music, but deepen

and enrich the experience of listening to the music.

I do consider this music, but given I've been listening to John Cage and

Karlheinz Stockhausen since I was 15, and that's 31 years ago, my musical

sense is rather broad...

In fact my approach to dealing with this data / music, I am specifically

referencing techniques used in John Cage and Lejaren Hiller's HPSCHD.

Several Months ago, I arranged that piece to be performed by my computer,

using Ableton Live. However - it was a performance as expected by the

composers from a home stereo unit, in fact Hiller is quoted as saying: "the

home listener's hi-fi set is integral to the composition".

HPSCHD as recorded for Nonesuch Records in 1969, is composed of 51 tapes and

three harpsichord players. The players (Antoinette Vischer, Neely Bruce, and

David Tudor) are playing excerpts from a source work "Introduction to the

Composition of Waltzes by Means of Dice", attributed to W.A. Mozart. The

sectioning and playing of these excerpts was conditioned and processed by a

computer. One section of solo allows the performer to play anything by

Mozart they so choose. The tapes are a collection of tones and sounds in an

equal tempered scale of 56 tones per octave. The patterns are overlaid and

continually change, resulting in a dense thicket of electronic and

harpsichordal sounds.

Then, perhaps the most useful part of the piece: the score that is

distributed with the record itself. A computer printout is supplied with the

record, and in data that looks like this:

. . . . VOLUME. . . .TREBLE. . . BASS

TIME. . .CH.1 CH.2 - CH.1 CH.2 - CH.1 CH.2

0.00. . .4 . . 2 . - 1 . . 2 . - 1 . . 4

0.05. . .4 . . 2 . - 1 . . 3 . - 4 . . 5

0.10. . .3 . . 1 . - 2 . . 3 . - 3 . . 2

etc.

Time moved in 5 second increments, and every five seconds, the listener had

to adjust the volume, bass, and treble of their stereo set to the above

noted values ranging from 0 to 4. I recorded the vinyl LP -scratches and all

- directly into my computer and then programmed the above noted values

supplied with the LP into Ableton Live, treating the program like a stereo

set. Only, I don't have the time pressure of performance: I was shifting the

emphasis from a synchronic art of performance to a diachronic art of

composition - all done in a computer, so that a computer would thusly

"perform" the composition of this music which was developed on a computer

some 35 years ago as a fulfilment of the performative compositional strategy

provided by the composers. I hope to release this recording as a CD soon.

In anycase, I saw that this kind of an approach would work well for

autchamber - instead of using the random data generated by Cage and Hiller

to process pre-recorded sounds as a live performance on a stereo system (or

as a dataset for computer software, in my case) I could use the data

generated by the autochambers to process the audio recorders of the

autochambers in software, as noted earlier. Since the data provided is all

diachronic excel spreadsheet data, and the audio is not a live onsite

broadcast, but merely compressed mp3 files of cassette recordings supplied

by the scientists, and is thusly not only diachronic but of a degraded and

processed texture to begin with, it only made sense to take this fashioned

thing and change it further, building on the strengths of the software at

hand, the patterns in the datasets, and the Cage/Hiller

performance-composition strategy in HPSCHD.

Autochamber starts where HPSCHD leaves off - volume, treble, and bass are

also manipulated by channel, but so are some other common processors (echo,

reverb) and some more esoteric processors (such as the downloadable freeware

ScrubbyVST which randomly mulches sounds according to certain time based

parameters).

This data could have been fed into a Max/MSP patch, but building the

necessary patches and testing them would have taken more time than the few

evenings I simply sat down and ground out the data entry into Ableton Live.

Also, by using "off the shelf" software and free downloadable plug-ins, a

questioning finger points directly at the technological nexus of HPSCHD and

autochamber itself: HPSCHD was developed on what was then, State-of-the-art

computers. But the final result was left to the random manipulations of a

comparatively low-tech stereo system and the vagaries of a human interacting

with it.

So too, I interacted with the vagaries of a simple to use piece of audio

composition software and its components. In this way, Autochamber "stands on

the shoulders" of HPSCHD, and looks to a very different horizon. Autochamber

is as much a work in conceptualism as HPSCHD, however, Autochamber 's

choices of processing parameters make for a more inviting and rewarding

listening experience as HPSCHD is exhausting and unrelenting in its

aleatory, dissonances, and density. A part of this is that the processes

acting upon the Harpsichords in HPSCHD are not derived from the Harpsichord

or the Harpsichord playing - they are randomly derived from a Chaos Engine -

the throw of die, the toss of coin.

Autochamber is taking the data from the machinery we are listening to, and

processing the sound of the machinery itself, making for a more coherent

aesthetic object and statement.

With each day lasting one minute, the piece would last an hour and a half

based on the data supplied (or 6 hours if one used an entire year's data)

one notices in changes of the audio the changes in the environment of the

autochamber itself, as it reported the changes in data. In this way, the

autochamber describes itself and its world in a more evocative and

understandable way.

Henry Warwick interviewed by Christina McPhee

San Francisco, California

November 10, 2004