David Widder

Postdoctoral Fellow, Cornell Tech

David Gray Widder, a postdoctoral fellow at Cornell Tech, examines ethical, political and economic aspects of ‘open’ AI systems – with a recent study focusing on big tech and concentrated power.

Widder says:

"I support the FTC's goal to reduce anti-competitive practices in AI. However, the FTC's recent statement ignores the underlying concentration of resources that prevent a competitive AI landscape, thus over-promising what open weight models can deliver towards this goal. 

"Most importantly, it does not follow that open-weight models are key to competition, when critical resources like computational power needed to build, scrutinize and use AI remain concentrated among a few companies.

"The FTC's emphasis on fine-tuning AI access overlooks how the handful of firms able to build foundational models thus wield significant editorial and market power to build dependency on their foundation model offerings. 

"The FTC's assertion that open models enable portability between hosting providers ignores that hosting providers often intentionally impose barriers to portability.

"Finally, the statement begins with a proclamation that history shows that open-source drives competition. However, the story is much more mixed, and there are many examples of open source being used as part of companies' larger strategies to kneecap competition. Android is open source, but Google is currently facing multiple investigations for using it to build an anti-competitive smartphone empire. 

"On balance, I do believe we must push for openness in AI. However, we must do so in a clear-eyed way, by linking this issue to the more significant concentration of resources (compute, data, labor) which underlie concentration of power in AI, so as to not over-promise what 'open' AI can give us."