Today David and me (Daniele) had a meeting in the Experimentalstudio and we also came to speak about our selection process for the algorithm we want to adopt in SD. While talking about this process we had the feeling that we, as a group, find ourself now in a sort of "impasse" with respect to our search for an algorithm that everybody can meaningfully integrate and fully embrace in his segment. In this sense, the act of taking a pre-existent algorithm feels a bit "inorganic" and "external" to our practice (here I'm speaking for myself, but also David shared this impression).

David proposed a solution to reverse the problem: instead of choosing a pre-defined algorithm that we then would inflect according to each own's preferences, rather let the algorithm itself emerge out of a process of collaborative algorithmic (re)writing. This ideally would happen in an iterative circle, where each one of us could add/delete/or modify a single step of the algorithm. Starting from a very basic algorithm ↗, we would then proceed with these "alterations", one at a time and in series. This process should be fast and spontaneous, f.e. a new alteration every two days. The process ends when there are no more additions/deletions/modifications. ↗ ↗ ↗
{author: poz, date: 191127, place: experimental studio, function: proposal, keywords: [cadavre exquis, rewriting, collaborative, iterative, algorithm]}

{dp, 191216}

Probably we didn't explain clearly what we intended to do.

As Daniele pointed out, our discussion started by noting how the

method we have (more or less implicitly) chosen in order to get to our

"starting point", the algorithmic text or score, as we also called it,

felt a bit clumsy and in some ways not right or even "alien" to

us. Even if in the process interesting discussion sparkled, it was not

clear how we should in fact in the end get to that shared "algorithm"

we would then all use as departure for our works. The method of

looking at what is out there, what kind of algorithms, which one could

be interesting, felt a bit over-imposed, not really coming out of our

engagement (aesthetic and other) with the space or the ideas of the

work itself.

On the other side, we fled that the choice of this starting point, its

definition should be, as in our initial thought, a shared one: a

common "text" from which the parallel and different works would

sparkle, but still is a common ground on which all of us could stand

(at least partially).

We though that a good strategy to get to this point would be a sort of

iterative process, where each of us would have an equal role to get to

or construct this starting point. The text for the "algorithm" would

then emerge out from our mutual interaction and not as a solution to

some (more or less arbitrary) choice problem.

We thought we might iteratively try to define the algorithm in

text. That is, we would define some rules by which each one of us, in

turn would re-write a part or a step of the algorithm, until it

reaches a point which is "acceptable" to everyone.  In this way our

interaction, the generative process that produces the starting point

would also be a sort of algorithm, as sequence of steps following some

shared rules. And this score would be and partly "descend" from

everyone's choices and inclinations. 

As an example, we came up with the following idea.

- Start with a very simple algorithm formulated in a similar way as

  usually algorithms are described in computer science papers (as for

  instance the PSO algorithm), as a sequence of steps.  

- The description should be only textual: that is no mathematical

  formulae are allowed

- In turn, each of us, would have then the possibility to modify this

  algorithm by either:

  + add step

  + change a step

  + delete a step

  + pass: do nothing

- Each step should be "as small as possible", meaning that the step

  should involve as little computational steps as possible

- Each should take care to use a language as less specific as

  possible: i.e., if possible terms proper to DSP language should be

  avoided (use energy instead of RMS): this would ensure 

  different interpretations and implementations.

- At all times, the whole state of the algorithm would be visible to

  all. Therefore, probably the title /cadavre exquis/ is not really

  fitting, as in this technique only the last line would be visible to

  the next writer.

- Cycles would be very short, i.e. the time between subsequent

  re-writings should be small, maximum 2 days.

Of course, many aspects (for instance what is a "small" size of a

step) are still unclear and would need (maybe) clarification, but

this is an idea. Many aspect would become clear by actually performing

it. 

And clearly, a crucial decision is which starting point is

chosen. Still I think we could find a common start here.

{author: DP, function: proposal, keywords: [cadavre exquis, rewriting, collaborative, iterative, algorithm]}

cadavre exquis

{kind: title}

Experimentalstudio, 191127

David, Daniele

Entry point

1. take any input signal

2. amplify or attenuate it

3. sned on any output channel
{kind: pseudocode}

iterative rewriting of an algorithm

we could try an experiment similar to the practice of cadavre exquis, where we work on a shared algorithm, iteratively altering its single steps, one at a time. Single instruction can be added, deleted or altered. We iterate until we have a sort of "Frankenstein" that then becomes the abstract basis to implement our segments.

rules or "an algorithm to make an algorithm"

• Each person can alter only one single instruction per iteration

• DSP terms are banished

• Maybe introduce a limited number of words
  

• Each single step or instruction shall be formulated "as abstract as possible". For example, if one wants to add a step that "calculates the derivative of a value", this should be reformulated as something like "take the variation of a value", in order to let the instruction open to different interpretations

{DP}

******** 

 
In my opinion, this entire work (as well as ALMAT's) is a "critique"
(in the sense of a detailed analysis, or a "bringing to light" of
inherent or implicit assumption/characteristics) of what we call the
"algorithm". The departure of this work here is exactly the fact that
given ONE score, MULTIPLE parallel interpretation may diverge from it,
all containing or implementing the same algorithm. That is, what we
call the algorithm is NOT what is contained simply in its "score" or
description. As you say, the boundaries (or also its ontology) of an
algorithm is not captured by classical formulations. The performance
of the programmer/artist is essential as well as their reciprocal
interaction and the interaction of the computational process with the
data it is applied on or it generates. This aspect being even more
underlined here in this process, in which this score or formulation by
choice and explicitly, is GENERATED through interaction.

So, I think, from this perspective, on the contrary, this approach (the
/cadavre exquis/) is exactly in line with our perspective.

******** 

We did not want to criticize or change what you call the disposition:
this is probably a misunderstanding. We were searching for a process
leading us to the starting point for the disposition to work with. The
segmentary situation is exactly what we wanted to enable, or to
highlight. Choosing a common starting point and then leave
algorithmic agency unfold in our parallel works.

"I often begin 'a work' with an algorithmic idea that I want to
explore, and it can well mean to start with existing pieces of code
and "well known" algorithms." I also experienced the contrary in my
own work and I usually avoid re-using preexisting pieces of code. I
find that implementing an algorithm from its (often different and
diverging) formulations is a performative act of me interacting with
the "algorithm" (probably nearer to its ontological essence, or at
least I would wish) I don't want to miss. It is this process that
generates most of the inspirations.

********

Again, the idea here is to explicitly produce a situation which shows,
even performs this incommensurability between the actual agency of the
algorithm on this steps. In trying to "reduce" or take smallest
possible steps, we try a sort of atomization of a process description,
we try to make the essential excess that agency has even more crucial.

How is the PSO a basic algorithm? Maybe we have a different
understanding of 'simple' .I read 3 steps: the first two composed of
multiple sub-steps: the second is even recursive. In our /cadavre
exquis/ idea, the PSO is already a very complex aggregate of
steps. But, you are right, the concept of "size of step" is blurry and
needs further clarification.

You are right, if you read the first text as a completed text. We
failed here to note that these where more notes, to read one after the
other. That is:
- we could begin wit 'take input signal, amplify, output to output
channel', but we would rather try to use no DSP terms.
- The example 'calculate the derivative of a value' wanted to point
out how that sentence brings with itself the language of calculus
(not linear algebra) and that one should avoid such specific
definitions in favour of more general.
The idea here is exactly to point out this difficulty (possibly
impossibility): NOT to refer to a specific context.

Personally, I actually like the idea of a 'Frankenstein' score (not
yet a program), an (maybe) improbable aggregate of different bits and
pieces, from different backgrounds, bringing with themselves different
understandings of the situation, of algorithmicity of
computation... And then, take that 'Frankenstein' very seriously.

********

"I would also avoid 'aesthetic preferences' as the criterion for
selection." I think what we should have said here is just
'aesthetics' (without preferences). That is a criterion that is not
really avoidable. What I meant here is that we may explicitly say that
what is usually buried under the narrative of science: that we
(including scientists and technicians) take decisions (also?) on the
grounds of your aesthetics.

This was another misunderstanding. We actually did not think of
exactly the same process as in the "classical" /cadavre exquis/. In
our version all the participants would have access to the whole
algorithm at all times: everyone could read the whole thing always.

I understand your objection about the sequentiality, but still, I
think one could still act in 'parallel' to the others to a certain
extent. Everyone, could follow some 'agenda' one may or may not be
able to realize, accepting a compromise.

{author: DP, function: response, keywords: [cadavre exquis, rewriting, collaborative, iterative, algorithm, interpretation, performance, score, parallel, aesthetics, experimentalism]}

{hhr, 191128} A preliminary comment:

The departure of the text (cadavre exquis) is the assumption of finding 'the algorithm'. I think that in the previous sessions we had already retracted from this 'direct form', as we discussed the difference between 'algorithm', its context, computation models, and 'algorithmic configuration'. I remember back in our lecture performance at Transpositions Stockholm two years ago, there had already been a question by Michael about the ontology of 'algorithm', and even before in xCoAx 2017 there was also some audience uproar when I questioned the boundaries of 'algorithm'; so I think we probably agree that it is futile to remain with 'the algorithm' or 'an algorithm'. Other pages were just titled 'Rewriting' and 'Data and Process', so I think that whenever we go to 'the algorithmic' as the medium of computation, we are much happier navigating the space. Also, an algorithm does not need to be a computer program algorithm; an algorithm can be a ritual that one performs (for example). ↗

You say we come to an impasse ↗ as beginning with 'an algorithm' would be 'external' to our practice. I would agree with the sentiment—something does not feel right yet—however I would disagree that the reason lies in the disposition here. I often begin 'a work' with an algorithmic idea that I want to explore, and it can well mean to start with existing pieces of code and "well known" algorithms. I think we deliberately created an "artificial situation" here, an "experiment". The experiment includes doing something which we perhaps would not do "normally", I think that is a fine decision. It is a new configuration we are trying out here. Most of the time, we would directly begin doing "our work" irrespective of others—who may come into the process at a later point. In other cases, we iterate over very long time spans, think of Anemone Actiniaria, a "configuration" we have been working on four five years now. The disposition now was: A segmentary situation for four individuals, it is one with several "horizontal" equal or homogeneous parts which come together as a whole, while at the same time algorithmic agency manifests itself as material excess and speculative detour from the pretext in the process of experiment and implemenation. Thus "differential segmentation". ↗

Emergence yes, perhaps again emergence of "an algorithmic ensemble" instead of "an algorithm". I think this idea is right. Let us reduce to 'collective algorithmic (re)writing' for a second. Then let us go back to the question of segmentality.

I am not convinced of the 'step of the algorithm' or the exposition of a 'basic algorithm'. I think what we have gained so far in this project and elsewhere about algorithmic agency contradicts these again, in the very same sense that you felt 'inorganic' about the disposition. The PSO was a basic algorithm, if you like. Instead, the proposal 'take input signal, amplify, output to output channel' is a signal flow diagram. Reading the additional rule "no DSP terms", I would even say it violates that rule. It is almost pure signal processing language :) The same goes for the example sequence 'calculate the derivative of a value'; while this is not signal processing language, it is of course the language of linear algebra. This is not to say that this should not be allowed, but to amplify the fact that we are always already assuming very specific configurations, types of signals (a differentiable value). So in a way the impasse reappears, only now—cadavre exquis—we end up with a Frankenstein "program" that throws together the 'aesthetic preferences' of each of us. ↗

I would also avoid 'aesthetic preferences' as the criterion for selection. I think it is what is responsible for the impasse. Now perhaps a thought on cadavre exquis. My understanding of it, as a surrealist technique, is that the text is hidden (except perhaps for the last line or tail), before being passed on to another person. This does not resonate with the "basic algorithm", unless I am missing something from your discussion. Although I like these kind of rituals, and I think we are onto something good here, I would question that it applies well to the situation of creating "an algorithm". If one goes back to the conceptual history of process (the Röttgers text), in the transition from alchemy to chemistry, Valentinus writes that who executes a process has to "especially pay attention to the sequence of the partial steps of the process", which I guess is akin to the classical/narrow idea of an algorithm as a pro-gram or recipe. In other words, while the cadavre exquis certainly produces "a text", it remains questionable if it actually produces "an algorithm" instead of a collection of transformations. The problem here is the mereotopology, the relation between the parts and the whole. The second objection I have is the sequentiality, it seems contradictory to the segmentary as a form of parallelism. ↗

What happens here, nevertheless, I think is already "our algorithmic rewriting" taking place. So we are at a particular iteration here, and I think we should allow ourselves to do a few more, until we are confident with the form. This I will try to sketch out in a separate sub-page. In it, I will try to return to the genesis of the piece and anchoring in "algorithmic segments". Perhaps the three elements I will take away now are, that we should forget about "the algorithm" for a moment, that what we are interested in is a rhythmic engagement of exchanging ideas that can take the form of a "ritual", and of the question of "blindness / opacity / partial knowledge" that animates the cadavre exquis.

{author: hhr, function: response, keywords: [cadavre exquis, rewriting, collaborative, iterative, algorithm, rhythm, ritual]}

{jyk, 191201}

I agree on the feeling that using an existing algorithm feels external, (though I am not against it at all and found PSO as well interesting to work with) as well as on the decision based on his/her own aesthetic preference.

I had a question in my mind how important the 'coherence' could be in the final work: there are 4 different individuals involved, and there will be a few common processes  - one way or the other.- and interactions in between.  Additionally we already have a strong common ground, 'the site(space),' the instrument.  If we don't let the aesthetic choices independent, then the meaning of our involvement could be weaker. Rather, diversity can be introduced, also considering the space that there are 4 stair ways in total (in a row) and they could have different identities, though they come from the same womb and have. in a certain degree, communications with each other. It feels more natural to me.

What I am questioning in most of the discussions (on making a choice on a algorithm or a certain step/process) is in regard to the experiences/listening processes. and feedback on those two. From a reading mode (data) to a listening mode(sound). (Perhaps I am missing some parts.) Maybe because of that, I am more attracted to a more open form for the starting point.

{author: JYK, function: response, keywords: [cadavre exquis, rewriting, collaborative, iterative, algorithm, diversity, listening]}

[poz 191202]

Maybe I didn't make myself clear on this point. With the term "impasse" I didn't refer to the overall disposition, which I still consider a valid experimental situation we are creating. I was more referring to the present state of things, where we are trying to find an algorithm that can "speak" to each one of us. As we were discussing in one of our first sessions, algorithms are not simply sequences of steps: an algorithm, in its description, embeds metaphors and ways of seeing the world that might resonate differently among each one of us. For example, the PSO algorithm comes from this "artifical life modeling" branch that is a bit distant from my personal way of approaching algorithms and computation. In this sense, an algorithm that resonates well with Hanns Holger view might be felt as "external" (extraneous is probably a better term) by me, or by David, or Ji. I think the impasse I was talking about comes from this friction: so far we discussed many potential algorithmic candidates, but every proposal was never fully validated unanimously because we have different ways of seeing things, and finding an algorithm that can relate to all of them it's not easy. One solution could also be that of simply embracing this "extranoeus sentiment" and even work with something that has little relation to our practice, but what we tried to sketch out with David during this meeting was an alternative strategy to derive an algorithm that could instead include all of our perspectives: an assemblage, a Frankenstein if you want, of all our different views and interests, in the form of an algorithm.

{author: POZ, function: response, keywords: [cadavre exquis, rewriting, collaborative, iterative, algorithm, diversity]}