The plants in the Gardens by the Bay evoke a sense of flourishing-by-design; photo by Victor from Unsplash. In this post, I’ll propose an approach for addressing three distinct concerns facing the actors responsible for generative AI systems: the computational costs facing AI operators, the concern that Generative AI systems may erode their own foundations by reducing traffic to platforms where training data is created, and the lack of credit given to individual users and communities for their data contributions. Our proposal is to give people credits, akin to the Bing search engine’s “Rewards”, for their past and future data contributions, which can be used to query expensive AI systems. All users would receive some replenishing supply of credits, in accordance with the collective nature of the vast data underlying AI, with additional credit given to notable data contributions. This approach can simultaneously address the incentives faced by AI-operating firms, online platforms, and individual users. We discuss the potential for a relatively lightweight system – that is quite similar to existing features of Bing, Dall·E 2, and cloud computing – to enable inclusive governance of AI systems.
Bing Rewards for the AI Age
Bing Rewards for the AI Age
Bing Rewards for the AI Age
The plants in the Gardens by the Bay evoke a sense of flourishing-by-design; photo by Victor from Unsplash. In this post, I’ll propose an approach for addressing three distinct concerns facing the actors responsible for generative AI systems: the computational costs facing AI operators, the concern that Generative AI systems may erode their own foundations by reducing traffic to platforms where training data is created, and the lack of credit given to individual users and communities for their data contributions. Our proposal is to give people credits, akin to the Bing search engine’s “Rewards”, for their past and future data contributions, which can be used to query expensive AI systems. All users would receive some replenishing supply of credits, in accordance with the collective nature of the vast data underlying AI, with additional credit given to notable data contributions. This approach can simultaneously address the incentives faced by AI-operating firms, online platforms, and individual users. We discuss the potential for a relatively lightweight system – that is quite similar to existing features of Bing, Dall·E 2, and cloud computing – to enable inclusive governance of AI systems.