Editorial leadership and AI disclosure - Handbook of Bayesian Deep Learning

Editorial leadership¶

This textbook was created through a large international collaborative effort by the Bayes AI Consortium, bringing together 106 co-authors. The editorial team consisted of Theodore Papamarkou, David Dunson, Vincent Fortuin, Yarin Gal, Philipp Hennig, Konstantina Palla, Tim Rudner, Sara Wade, Andrew Wilson, and Cheng Zhang. Theodore Papamarkou conceived and initiated the project and led its development and coordination.

Use of AI tools¶

This book is first and foremost a collection of human effort: the work of many authors, reviewers, and editors. We did, however, also make use of AI tools—including GPT and Claude models—to help polish the content, and we believe readers are entitled to a transparent account of where and how.

The technical substance of the book is human. Every chapter was written by its human authors first, and the ideas, derivations, and judgements are theirs. AI assistance entered afterwards, and mostly in the editorial layers.

Cross-chapter review. Before the editorial pass, chapters were reviewed by authors of other chapters. Some reviewers used AI tools to assist their reviews, while every reviewer read the chapter they were assigned at least once. The cross-chapter review was therefore a mixture of AI-assisted and purely human review.

Editorial review. The editorial pass focused on style, cross-referencing, notational consistency, and similar matters of form rather than content. This stage made heavy use of AI tools. The editorial team first agreed, as humans, on the reviewing criteria and recorded them in a shared instruction file; AI tools were then used to generate editorial feedback against those criteria. Because the resulting feedback concerned styling and consistency rather than technical claims, authors were permitted to use AI tools when addressing it, and a subset of them did.

Reader’s guide. The reader’s guide in the front matter is largely AI-assisted: the editorial team supplied the prompts and the structure, and AI tools were used to expand them into the more detailed per-chapter descriptions. The editors reviewed the result.

Web version of the book. Theodore Papamarkou converted the book from its TeX source to this Jupyter Book web edition with the assistance of GPT.

Responsibility for the content of this book, including any errors that remain, rests with its human authors and editors.