Remarks on "AI"

In the Carnegie Mellon Textiles Lab, we do not use "Generative AI" to make any lab creative output (e.g., code, writing, images).

We do this because, among other reasons:

• We give people credit when we use their work. "Generative AI" systems mix the work of many people together in a way that makes it impossible to give proper credit. (We do not view fractional, estimated credit given to hundreds as proper credit.)
• We do research both to discover new ideas and to improve our own understanding and skills. Offloading the creative engineering work of research diminishes our opportunity to build our own abilities and discover new problems in the minutia of solving existing ones.

This is not an exhaustive list of reasons. Particularly, these are merely pragmatic justifications based on our function as an academic research lab, and do not begin to reach larger societal concerns. Nonetheless, either of these reasons, alone, is sufficient to prevent us from engaging with "Generative AI" when producing creative outputs.

Briefly Noted

As a lab, we respect dissent, discussion, and nuance. We are a group that will always discuss our choices, and do our best to reason through ambiguous cases together. Being right is a continuous process, not a static position. We seek to act under our best understanding of the world, while acknowledging that this understanding is continuously evolving. We will revisit and update these remarks as discourse in the lab changes.

There is room for subversion and experimentation in our lab. Some lab members choose to use "Generative AI" -- thoughtfully and critically -- in other ways; including the production of intermediate artifacts which do not appear in lab publications, or as a component of systems they develop.

We use "Generative AI" above as a catch-all term for systems that use statistical techniques to approximate human-produced creative output using internet-scale data. You would be correct to observe that such techniques are not the only ones that run afoul of our principles; and, indeed, as a lab, we seek to avoid all such techniques.