ChatGPT's Infrastructural Ambitions: AI, Commodification, and the Commons

Résumé analytique

OpenAI's ChatGPT arrives as the Government of Canada attempts to pass the Artificial Intelligence and Data Act (AIDA), and the authors urge policymakers to adopt the information commons as a key frame for understanding AI chatbots and the large language models behind them. ChatGPT could not exist without the collective production of commons resources, and its exploitation of those commons will only intensify as OpenAI and its competitors monetize chatbots. Commons-based approaches respond to demands for stronger collective rights in AIDA, which currently focuses on economic or psychological harm to individuals rather than the rights-based approach recommended by the Office of the Privacy Commissioner.

The authors take ChatGPT's 23 March 2023 rollout of third-party plug-ins as evidence that its owners aspire to make it infrastructural, a key platform for accessing the internet and completing everyday tasks. Behind this lies hidden labour, from data scraped from sources like Wikipedia to low-paid workers in Kenya hired to annotate offensive text. As the once-free tool commodifies through offerings like ChatGPT Plus at $20 USD per month, the authors identify two strategies drawn from Vincent Mosco: extrinsic commodification, harvesting historical common data such as Common Crawl under aggressive copyright interpretations, and intrinsic commodification, mining data from firms' everyday operations. Both undermine the reproduction of the commons, and the authors ask policymakers to take seriously this exploitation and the networked labour enabling it.