Grand Theft Autocomplete: 'AI' enclosures and their threat to data sovereignty

Abstract

The way we distribute research data and software today relies heavily on the open web: largely standardised formats, decentralised data hubs, public availability, affordable compute, free learning resources, and findability afforded by a few ad-supported web indexes. This openness, however, is increasingly under threat. The companies that currently attempt to monetise large language models also make it much harder to keep websites online and position their chat interfaces as intermediaries between knowledge and users. The ‘AI’-enclosed web empowers Big Tech to further interfere in learning, education, and research. This talk draws a parallel to the historical appropriation of common land in 18th- and 19th-century England. With this historical revolution of the rich against the poor to learn from, it outlines the specific dangers to the knowledge commons in the next few years. In conclusion, the talk attempts to identify what can be done to preserve and and empower individual, institutional, and transnational data sovereignty in the face of this not entirely new threat: critique the neoreactionary ideology behind the technology, empower ‘boring’ technology, identify affordable machine-learning tech, and embrace local and peer-to-peer computing.