Abstract
The deployment of AI in consumer products is currently focused on the use ofso-called foundation models, large neural networks pre-trained on massivecorpora of digital records. This emphasis on scaling up datasets andpre-training computation raises the risk of further consolidating the industry,and enabling monopolistic (or oligopolistic) behavior. Judges and regulatorsseeking to improve market competition may employ various remedies. This paperexplores dissolution -- the breaking up of a monopolistic entity into smallerfirms -- as one such remedy, focusing in particular on the technical challengesand opportunities involved in the breaking up of large models and datasets. Weshow how the framework of Conscious Data Contribution can enable user autonomyduring under dissolution. Through a simulation study, we explore howfine-tuning and the phenomenon of "catastrophic forgetting" could actuallyprove beneficial as a type of machine unlearning that allows users to specifywhich data they want used for what purposes.