Collaboration Monsters for AI
Brewster Kahle, Primavera De Filippi
The time has perhaps come when everyone does not just want openness, but rather fairness, especially when there is a fear that corporations will suck up everything that is open. Today, fairness has become perhaps more important than openness. What role might the Internet Archive (and other similar institutions) play, in this coming age of artificial intelligence, when access to training data is perhaps the most important factor in its development? How might fairness offer an alternative narrative to the dichotomy of open vs. closed?
The Copyfair License is a particular type of open license that aims to promote fair and equitable access to copyrighted material, including artistic and literary works, software code, datasets and model weights insofar as they are protected by the European sui-generis rights on databases. The goal of the Copyfair License is to ensure that, while access to these materials remain freely available to everyone, commercial and non-commercial uses can be subject to the principles of fairness and reciprocity. The license is inspired by the Creative Commons licenses and the OpenRAIL licenses, as regards the additional constraints introduced in order to ensure value-alignment with the expectations of the licensor
The time has perhaps come when everyone does not just want openness, but rather fairness, especially when there is a fear that corporations will suck up everything that is open. Today, fairness has become perhaps more important than openness. What role might the Internet Archive (and other similar institutions) play, in this coming age of artificial intelligence, when access to training data is perhaps the most important factor in its development? How might fairness offer an alternative narrative to the dichotomy of open vs. closed?
The Copyfair License is a particular type of open license that aims to promote fair and equitable access to copyrighted material, including artistic and literary works, software code, datasets and model weights insofar as they are protected by the European sui-generis rights on databases. The goal of the Copyfair License is to ensure that, while access to these materials remain freely available to everyone, commercial and non-commercial uses can be subject to the principles of fairness and reciprocity. The license is inspired by the Creative Commons licenses and the OpenRAIL licenses, as regards the additional constraints introduced in order to ensure value-alignment with the expectations of the licensor
The Collaboration Monster is a novel institutional arrangement that aims to create a synergetic ecosystem where value-aligned entities unite forces to drive the development of AI products and services in a collaborative manner. It relies on the innovative use of IP law (namely trademark and copyright law) to promote collaboration. By aggregating multiple actors and creating entanglement at various levels amongst these actors, the project aims to shift the payoff structure towards cooperation, fostering a more collaborative and innovative AI ecosystem.
The main goal of the Collaboration Monster is to align incentives across the network participants, with a view to fostering a cooperative environment where participants are invested in each other's success. In addition, the Collaboration Monster includes mechanisms for third-parties to share critical resources such as compute power or data only to those who abide by the charter - thereby promoting open innovation while ensuring fair distribution of benefits.