Papers
An essay on two layers of the proposed design: an agent with no persistent self to protect (Anatta), and a goal that yields to human shutdown (Anicca).
Status: Shared by the author. This is a conceptual proposal. It has not been peer reviewed, and the architecture it describes has not been implemented or tested.
The one-page introduction to the Moral AI proposals: why caging a model after training is not enough, and why alignment has to begin with what goes in.
Status: Shared by the author. This is a conceptual proposal. It has not been peer reviewed, and the architecture it describes has not been implemented or tested.
A supplementary manifesto to the Moral AI framework, with a personal foreword from the author. Preprint.
Status: Preprint, shared by the author as a milestone. This is a supplementary document to the manuscript under review at the Journal of Artificial Intelligence Research (JAIR). It has not been peer reviewed or accepted. It is a conceptual proposal: the architecture has not been implemented or tested.
A proposed architecture for non-harm, causal reflection and shutdown compliance in AI agents, drawing on Theravada Buddhist principles. Preprint, under review.
Status: Preprint, shared by the author as a milestone. The manuscript has been submitted to the Journal of Artificial Intelligence Research (JAIR) and is under review. It has not been peer reviewed or accepted. It is a conceptual proposal: the architecture has not been implemented or tested.
