In 2016, the Demystification Committee has been investigating the slippery world of offshore finance and the secrecy jurisdictions through which it flows.
The initial body of work that resulted from the investigation is collected under the name Offshore A—Z. Prior to the Panama Papers leaks, this investigation took the form of data gathering and analysis. The snippets of information that could be found online about companies potentially involved in oblique financial practices were scraped using, as a source, the online company registers that tax havens keep as an empty gesture towards transparency. This information, comprising records on half a million companies, and naturally full of gaps, was compiled into a single database in an attempt to identify patterns and links.
On this database, we trained and ran a number of algorithms aimed at filling in the gaps, speculating – through technology – on the forms of some of the companies. A web scraper looks for keywords and attempts to find images to give the companies a visual representation. A neural network, trained on the objectives we could find for the offshore entities provides a company mission statement for each nondescript entity. A Twitter Bot releases one speculative company description every 15 minutes, linking to the centralised database to provide a permanent record of the possible activities of these entities, the Offshore A–Z.
Today, we release an interface of the neural network we trained: offshoreaz.company/objective/generator/
When queried with any word the neural network will return a coherent, speculative offshore company mission-statement. The statement’s length and variation (complexity) can be controlled with two sliders.