When the “Open (For Business): Big Tech, Concentrated Power, and the Political Economy of Open AI,” a paper by David Gray Widder, Sarah West, and Meredith Whittaker, was published last August, we appreciated its thorough criticism of “open” in the context of AI.
To discuss the work presented in “Open (For Business)…” and explore possible solutions to the issues raised in the paper, we have invited its authors to join our monthly community calls, which explore the intersection of AI and the Commons.
As we stated in our response to the paper, similar to its authors, we also believe that “openness” alone is not a magic bullet, and “open washing,” in particular, contributes to obfuscating the injustices and power relations within the digital realm. During the call, participants agreed that openness no longer functions as a universal remedy for democratizing technologies, citing cryptocurrencies as a recent example fraught with environmental and power concentration issues. We also acknowledged that the technical facets and exact meaning of openness in AI remain unclear. Our team currently explores these issues with other stakeholders in the Open Source Initiative’s work to define OS AI; we have also been vocal about instances of “open-washing,” exemplified by Meta’s (Llama2) and Technology Innovation Institute from the United Arab Emirates (Falcon 180B) launching proprietary machine-learning models under the disguise of openness.
Our conversation underscored the need to incorporate the political economy into discussions on openness, emphasizing transparency as a precondition for addressing inequities. From where we stand, it is clear that the way forward is to strengthen the role of institutions that act in public rather than private interest. If we want to reduce our dependence on big tech, we — in the European context, that means the EU — must invest in public infrastructures. In the context of AI, this means, among others, establishing a robust and publicly accessible computational foundation for open source AI research and investing in trusted, commons-based data sets. Our response to the issues raised in the paper is guided by the concept of Digital Commons and a broader vision of technologies established and governed as public digital infrastructures. Specific recommendations on how to address the bottlenecks preventing the democratization of AI are outlined in our response to the paper.
We agree with the diagnosis of the current digital landscape presented by the authors of “Open (For Business)…”. However, although openness is not a silver bullet, it is worth preserving. The dual nature of openness in that it can both challenge and enable the concentration of power is what we call the Paradox of Open. Through our work, we come up with concrete measures to address the challenges that we also identify, based on concepts such as Digital Commons, Digital Public Space, and Digital Public Infrastructure infrastructure, without sacrificing openness.
This text is a commentary on the conversation with the authors of the “Open (For Business): Big Tech, Concentrated Power, and the Political Economy of Open AI” paper, which was part of our ongoing series of community calls called AI the Commons. If you’d like to join them, email firstname.lastname@example.org.