Reimagining Decentralized AI
Tomer Shadmy, Katrina Ligett · 2024
The paper uses aspirations mentioned in the initial research on machine learning decentralization as a lens for examining the current state-of-the-art and exposing opportunities for future innovations. We explore the potential and limitations of decentralized architectures in affording privacy and human agency for end users, competition, and collaboration for wider market and civic players. We then elaborate on the legal and technological developments necessary for decentralized machine learning systems to realize their liberating potential.