Installation
2006

Synopsis

Most of the time, the Babel Machine does nothing. It stands quiet in a corner; a dot matrix printer from the eighties parked comfortably on a pedestal, the surrounding floor covered with tractor feed printer paper. Then, suddenly, the beast within awakens in an aggressive burst of output. It prints about a page length and returns to its contemplative silence.

The fresh print out seems at first incomprehensible, but in a second glance it becomes much more intricate. It's almost like a peek into the mind of a paranoid schizophrenic. In the third reading it starts to almost make sense, it is about something... but what? I recognize this - there was something about this in the paper this morning!

Explained

The Babel Machine consists of a dot matrix printer, tractor feed paper and a computer with internet connection running custom software. The software uses various Web 2.0 technologies to scan for current news reports and blog based discussions regarding these news items. It harvests these discussions and stores them in a database. If there are enough number of responses to reach a minimum threshold, the next step is activated. The software will then use a Markov chaining algorithm from the harvested discussions, format the results and output them on the printer.

Markov Chaining

Without going into detail, Markov chaining is a relatively simple but effective method to use a flexible amount of input data to generate a flow of output. The interesting property of Markov chaining is that the output flow is statistically correct, relative the flow of the input data sets.

Conclusion

In this case, the discussions on news reports are the input data, meaning that what is generated for output is based entirely upon the sequences of words used in the harvested and analysed discussions. Though, when using text as input data, the accuracy of the chaining does not increase with the amount of input data. On the contrary, there is an increasing entropy taking place, meaning the more information gets fed into the database, the more chaotic is the result of the output. Amusingly with retained statistical accuracy.