Human, All Too Human

Horse_eBooks, Art, and algorithmic anxiety

So, the jig is up. Horse eBooks, that hilarious, enigmatic twitter account has been revealed to be a sham. The men behind the curtain have surrendered, and have announced Performance Art.

The interesting thing is that, in being announced as human-controlled performance art, Horse_eBooks loses most of its power. Knowing that each post wasn’t algorithmically generated, but individually, manually created, tells us differently to what we thought we knew about the technologies that use this technique; less about data and scripts, and more about our need to emulate machines, so as to prove our own supposed authenticity.

As an imitation of a broken algorithm, needlessly publishing half-sentences scoped from corrupted neural networks, Horse_eBooks makes us aware of those hiccups in the technology that we have created. As a piece of art, it ultimately worked better as the evidence of a broken script, an elevated, chaotic algorithm that ceased to have any semblance of its original purpose. We’ve all seen the ways that bots try to deceive Twitter, generating random combinations of words to use as tweets so they could exist without being (algorithmically) flagged as spam. There’s something to be pulled from these malformed sentences that barely emulate human dialogue, something that was interesting with Horse_eBooks. It’s not only the deconstruction of language, but a presentation of a machine’s eye view, our conversations out of context, delivered back to us in jarringly constructed form.

As mentioned by Paul Graham Raven when I spoke to him about the reveal, Horse_eBooks isn’t particularly revelatory, even as a writing method, ‘Burroughs got there first, just on a cruder platform’. Burroughs, Bowie, Dylan, all used this cut-up technique to find new uses and structures for language. I’d be interested to find out how they constructed every tweet, and how they found the ‘tone’ of a machine, the linguistics of an algorithm. You could argue, if you really tried, that Horse eBooks was an exercise in our own authenticity, that as the supposed controllers of our own language we should learn how they are corrupted. By playing The Algorithm, Bakkila and Bender unwittingly reveal our own anxieties about bots, language, and algorithmic reappropriation.

We all knew that Horse_eBooks had a creator, we just didn’t realise how complicit they were in its output. The questions I’m asking are where artists continue from here, how to use the broken algorithm as a material without being too instrumental in the result, without being too human. Where, and how, we lose control, is another question altogether.