Abstract
AI solutions require model training. The data provided for the model training are usually deployed on centralized servers. However, there are some issues that originate with centralized systems such as data centralization, lack of privacy (data can be publicly revealed), lack of control over data (illegal or third-party data sharing), data contributors do not benefit from the data usage (they are not paid for their data). Decentralized Artificial Intelligence is a paradigm that shows the potential to achieve complete data privacy while training the Artificial Intelligence model. The system trains the AI model on data contributors’ data stored at their end. The system does not require to send any data to any centralized server for training, and thus protect data contributors’ data privacy. Smart contracts on Ethereum blockchain provide specifications required for model creation. Training is done in a sequential manner where each data contributor can train and update the model only once. Data contributor is rewarded with Ether coins for their contribution to train the model with their data. Moreover, the trained model will be accessed by the administrator to help make intelligent decisions, and thus provide better services to their customers.