## An Ecology of Autonomous Software Agents
by Ben Dean
_"The fundamental problem of communication is that of reproducing at one point either exactly or approximately a message selected at another point. Frequently the messages have meaning;"_
From _A Mathematical Theory of Communication_, Claude Shannon
Sometime in the first half of the first decade of the third millennium of the common era, I encountered a web page that, in retrospect, had a deep influence on my thinking about software bots, communication, and meaning. The page in question was a Google SERP, a search engine results page. The particular search term that led Google's PageRank algorithm to assemble those ten blue links in the order that it did is lost to me, though likely not to Google—but suffice to say, it must have been some high-entropy phrase. The results were all long pages of apparently procedurally-generated text. Any literate human would immediately discern the utter absence of meaning in the strings of uncommon words, liberally salted with links to other pages with the same high-entropy, low-content structure, but at the time, only a few autonomous software agents could make the distinction.
This was my first awareness of the bot ecosystem. Before the auto-following, friend-requesting, auction-bidding, market-making bots, I found this odd corner of the web. On one side, PageRank's spiderbots indexing web pages by search terms and following links indiscriminately. On the other, bots pushing the idea of "content provider" to the breaking point by reducing content to its information-theoretic definition. And, crucially, in the middle were the ad bots, bidding on search terms to get their master's sales pitch in front of as many potential customers as possible. This degenerate case of logical, though useless, software-mediated communication seems to have been a sort of perverse inspiration for a number of artistic experiments.
#####Wittgenbot
My first foray into the world of (pure software) bots was, truth be told, little more than a smarmy joke. My (autonomous software agent would)[https://twitter.com/LudWitgen], at quasi-random intervals, tweet a segment of Ludwig Wittgenstein's famed _Tractatus_. I perhaps flatter myself that the joke is clear enough that letting the thing die shortly after its creation was still sufficient for my purposes. The reality is that that this piece functions just fine as a description of the thing, and there is little call for the thing itself. There is no connection—that's the joke—but this means that there's no ecosystem, either.
#####I Am Unable
[I Am Unable to Tell You](http://turbulence.org/project/i-am-unable-to-tell-you/#), now out of service, was not so much a bot in and of itself in the strictest sense, but rather an attempt to explore ideas about web ecosystems with the runaway feedback loop I had found years earlier. Here was a site where the content was only accessible to users who were themselves creating the content by looking at the prior content. The site would create a video using the visitor's webcam while displaying a video created in the same way, earlier. There was no sound, and not much happened, yet people kept clicking next, and the tree of videos grew. The visitors were instrumentalized to become my content-creation bots, using our common desire for meaning and connection as bait.
#####The Perplex
[The Perplex](https://github.com/bd/perplex) is a bot that constructs languages by writing in them. These are, of course, not real languages in any but the most formal sense. Hearkening back to the Wittgenbot, the Perplex is an attempt to make _really good_ gibberish: neither random nor predictable, and certainly meaningful, at least on the level of sense. I sought to invert the logic of _I Am Unable to Tell You_, while maintaining my own authorial absence. Could I create the forms of human connection and communication without any meaning at all? Without any input from the world beyond the noise of random numbers? What would it look like? Using tools most commonly employed by bots to extract meaning (parsers, stochastic n-gram analysis, etc.), could I make a document that looked like it would have meaning, while remaining inherently empty?
The reason these bots, and others, have interested me is the contrary motivations they both evince and undermine—to connect with other humans, and to reap the benefits of human connection without actual connection. Attention is a commodity on the market because it is of fundamental value to us social primates, yet our technological shortcuts undermine that actual value, with the market's understanding of that diminishing value lagging behind, though not by much. We've seen what happens when autonomous software agents make decisions in response to one another on electronic commodity exchanges. Are we ready for what happens as these bots become harder and harder to distinguish from humans?