Are we just using the human bodies as filters? I mean, if we think of stimuli as input, these are mapped through the human bodies first, and then through the computer. The computer remains a part of that loop, because of the biofeedback. But then how do we optimize the ‘human algorithm’?
Well, that assumes that the idea of the algorithm is to be optimized. But, from an artistic perspective, we’re not really interested in the optimizing properties of the algorithms, or in the idea that an algorithm is better than other because it takes less steps or less processing power, or whatever you understand by optimization.
The core of what we’re doing is the idea that the algorithmic produces also a sort of speculative/creative element by itself that is very difficult to design or to intentionally control through a rational approach to algorithms, but that comes from the praxis of working with these machines. In that sense I can see a sort of similarity to this hardware hacking or building things structure, that’s how I would like to understand the development of any algorithm that has to do with a knowledge that has been created over a long time through the exchange that derives from the practical work. If we say that there’s a sort of isomorphism or similarity between these two domains, then looking at one of these domains could help understand the other, for example. Or coupling the two domains could produce something new that goes maybe beyond the sum of the parts.
Well in general there’s a lot of conversation about machines/algorithmic bias, which is a human bias. I think that maybe my approach is more focused on the body, the human and humanity as a source of what the computers actually turned out in the end. Especially if there’s a human performer that somehow inserts himself into it.
But I think that algorithms do have a lot to do with optimization and how do we interpret what this algorithm does and how do we build it to do it more effectively. Less steps, less noise and stuff like that. This question about optimization I think is interesting in terms of what David was saying to: if we don’t categorize emotions then how do we optimize how we should listen to this. How do we make it itself evident about what we should be paying attention to, which is not the categorization but the intensities of the individual signals.
I found this thing of optimization interesting, because it is something I’ve never thought of in these terms and I think it’s something I don’t do at all. On the contrary, I try to find out where the algorithm fails, because that’s the most interesting part for me.
The difficulty for me in the term optimization is that, if you think about what it means or how it works, then it makes a couple of presumptions that I don’t think they hold so easily for artistic production. Of course you want your work to succeed in a specific way, but optimization for me always assumes that you have something you can measure it against. You can say something is a better solution than other solutions because you have a way of evaluating, and the evaluation process assumes that you have a given goal and that you can measure if you are close to it or not. From an engineering perspective is very easy to understand optimization, because you have given parameters that you want to optimize. But in terms of the artistic production I find it very difficult, because I would always assume that if there is a goal it would even be shifting. Or that there is a very strong abstraction with respect to the goal. The aim is very general: you want to make a successful piece that a lot of people can relate to, for example. Or very few but very strongly, whatever you define that. But it is very difficult to break it down like you would do in engineering perspective, where you define goals and then you can sort of operationalize it. Which I think is the precondition for optimization, that you can operationalize it and you can then iterate and see if it goes to a specific solution.
The other thing that for me lies in the term optimization is that there is actually a one, specific point where there is the ‘optimum thing’. I’m thinking of these learning algorithms, where you always look for some global maximum or something like that. But for a music piece, or a sound piece, is almost impossible to say that there exists a kind of optimal spot that you can bring your piece to.
I totally agree with you. But I think it’s more that algorithmic art can engage with this idea of optimization, and it’s something that you can comment upon. This is related to what I find so funny about this idea of algorithmic music that sonifies emotions, because it presupposes that there is this specific method and ‘finally we know what happiness sounds like!’. Happiness sounds like romantic music, and we know that because we were programmed that way. I’ve always been struggling with the desire to communicate but also the desire to break apart these conventions.
Which I feel strongly linked to the idea of categorizing things. I think categorization is implicit in the optimization process. You need to identify at least one category that you tend to, towards which your optimization is directed.