Q: OK. I’m still reeling a bit from what you just said about Walter Benjamin and machines. For one thing, I had to go and take a look on your website at a lot of your art again. I saw entirely new things in your painting now than before. I had no idea the non-pattern thing was so important. I’m thinking about your work entirely differently than before.
This leads me to a question I am not sure that I know how to ask. How do you know for sure? I mean how do you know a machine could not do what you do?
ReeL: A machine may still be able to mimic my work superficially, but I suspect there will still be a way to tell the difference. No matter what, machines are still limited to algorithms. That is a very specific and interesting limitation.
Q: OK, I have to admit I do not know much about algorithms, or even what they are, really. What is important here?
ReeL: Fair. We now know there are subtleties, and unexpected nuance. I’ll try to illustrate with a simpler problem: The problem of generating a random sequence. Technically, an algorithm cannot generate a random sequence.
Q: Why not?
ReeL: Almost by definition. If it did, the sequence would not be truly random. It would have a pattern, the pattern imposed by the algorithm. So as far as algorithms go, patterns are key. This is why my previous statements emphasized patterns and my requirement to paint free of algorithmic reproduction.
Q: So if no algorithm, no machine?
ReeL: No, not quite. The case of machines is more complex than the problem of agorithmic generation of random sequences. A machine can get a random sequence in other ways, for example, by detecting and recording a natural process that is known to be random, like the decay of an appropriate radio-isotope. You see, machines can detect and record, as well as calculate. This is part of their power to mimic. Machine mimicry goes beyond calculation, but generation is still limited to the realm of algorithms.
This is in part why Turing did not see the whole implications of artificial intelligence as we know it today. Turing was primarily interested in algorithms, calculation, and the kind of problems a very fast calculator could tackle, and the problems associated with constructing instruction sets, or what we today call “code” to direct those calculations. Turing did seem to have seen the possibility of the code itself being generated by a machine. So in that sense, he saw a crucial piece of the current situation.
Q: So the machine, since it can do things other than calculate, adds more complexity to the problem and Artificial Intelligence uses this full complexity and greater potential of the machine.
ReeL; Yes, essentially. The trans-algorithmic realm of machines adds new wrinkles.
Q: Are there other artists dealing with these issues?
ReeL: On one level, yes, all the time. For example, as far as the limitations of algorithms, this is an area that a lot of today’s musicians have more awareness of in terms of the arts. They may not think of it in the way I talk about it, but they experience its consequences all the time. For example, we still have music machines that use live recordings of real musical instruments as a basis for their synthesized sounds: it still is often better to digitize real phenomena, that is, detect and record, than to generate, that is, calculate from scratch algorithmically.
Ask a good music engineer about generating really good sounding synthetic percussion tracks. A lot of times they’ll manually alter the track just off the beat or the strict algorithm to make it sound more natural. There’s a whole art to it. In this case, people go out of their way to preserve the human mind behind the machine.
Q: Apart from all this, didn’t someone named Boole cover a lot of this much earlier? Hasn’t some of this stuff been around a long time? Since the nineteenth century?
ReeL: Yes, he basically invented Boolean algebra, or what he called the “mathematics of thought”. And Boole was aware of machines, such as Jaquard looms, that could work with instruction sets. Ironically, Boolean algebra, turned out in one sense to be far more important mathematically than he dreamed–it’s fundamental to understanding what is called Universal Algebra, that is, most other algebras at a certain abstract level–while as far as machines and electronics go, it turns out it is not quite as all-encompassing as he thought. Originally, it was thought that Boolean algebra could at least represent all electronic circuits. It turns out that is not so. There are relatively simple recursive circuits that cannot be represented by Boolean algebra, yet are relatively easy to construct physically. This was not discovered until much later, long after Boole. And recursivity is a big part of modern programming and Artificial Intelligence.