Evolution, Recursion, Agency
[CUBE, 27_04_18, David presents his system]
In my perspective, algorithms are a sort of thinking tool that have a retroaction on me. They are not directed towards a specific aim: the algorithms I implement are systems that evolve in time. The evolution of those systems is what interests me, not their outcome, or whatever I define as satisfying condition. So I implement a system and then I excite it in different ways, either through my interaction with it or through inputs from the outside, and then I listen to how it evolves in time. This temporal evolution is what interests me.
[referring to the Fermi-Pasta-Ulam experiment] I have a background in dealing with this kind of systems. And I define an algorithm starting from this experience. It is something that develops in time, it has an evolution. And that evolution is implied in the ontology of the algorithm, in its recursive structure. This is my personal understanding of what an algorithm is.
There are also some algorithms that don’t have an iterative or recursive element.
Of course there are, but for me the recursive element is the key for understanding the agency of the algorithm.
But I think that one aspect that for me is very essential to the algorithms that matter is that there’s another type of recursion. Of course we can arrive at this system, but the system doesn’t exist in a first place. It has to be created by somebody, through some reflective process. And at some point this person will run the system. They run it and then stop it, changing something and then run it again. There’s another type of interaction or recursion that has to do with a shift in terms of what the algorithm is. In other words, we can have this algorithm, this simulation that runs this physical system. But that’s only one particular snapshot of the whole algorithm, if you understand the algorithm as the whole process of this thing you created out of some initial idea of something.
For me the recursive aspect is the core of the agencies of an agent. You compute the next step, and this next step is the starting point for applying the same rules again. This thing is the core of agency as I intend it, but of course this extends out of the particular implementation of the algorithm. There is obviously a recursive relation between me and the program, just as you said.
Ok, you said that this state machine is for you the core component of the agency. But I don’t know if you can consider the concept of agency without somebody observing the agency, because agency is in first place an attribution made by an observer. If you think from a cybernetic point of view, anything like behavior or intentionality has its origin in somebody that observes the system. It in an ascription, and whether that is true or not we don’t know, because it is a constructivist idea. So agency requires that we are coupled to the system.
I just had this thought, that was kind of interesting. There is this conversation about how our external tools and algorithm have an agency and maybe it goes back to this previous conversation that you started about how tools help us thinking. This has an interesting parallel to art itself. Where does the artist end and where does the art begin? How does this externalization act upon our thought? Different observers would see different things in the art. It’s something that you need to learn really fast, you need to be satisfied with making art that can evoke reactions in different ways and can be read differently, because if you make something that’s just one single controlled linear message, then you have just made a poster.
This whole system is a feedback system. It is recursive on different levels. Starting from that historical moment, when I implemented that experiment, I individuated these recursive relationships as the core of my practice. Of course, as you said before, you need an observer in order to see it. And when you see it, when you perceive it or individuate this recursive relation, then it becomes agential, it becomes material in a way. And of course, it grows out of the simulation itself, it is not just in the computer, but it’s in the relation between me and the computer, and this also has a sort of agency that begins to be material in the moment I recognize it. This recursive relation that emerged from that simple model, has become for me a structuring element in my work. I’ve written this software at least 3 times anew. First version was in SuperCollider, the second in C, and now it’s in Fortran. Every iteration, every new implementation of the same system allowed me to understand better what I was actually doing. It’s not optimization in that sense, but more radicalization: to find out what it’s the core of the thing and try to write it down in the most compact way as possible in order to see better where is its generating potential. Radicalization is taking away what is not needed and find out what is the origin of the work. To go back to the root, to find out what it's going on: what is important for me, why I do these things and what is the very core of this idea I'm trying to realize. That's the radix.