Kris Bot

As part of an exercise in AI-human partnership, Hammond developed a chatbot to help accelerate his writing process. To develop the affectionately named “Kris Bot,” Hammond’s team trained the system on scores of Hammond’s written papers, articles, opinion pieces and lectures. With this training, the bot proficiently writes in Hammond’s voice, mimicking his tone, language and point of view.

In the piece below, Kris Bot explains — in Hammond’s voice — how and why it was created:

We are at a crossroads with AI technologies, a space brimming with promise yet fraught with exceptional fear. In particular, the emergence of generative AI systems such as ChatGPT has people concerned that the machine will replace us in precisely the area that we think of as uniquely human: our ability to create.  

For those of us in higher education, this plays out as a concern that these technologies might replace us in our valued roles — as writers, authors and thought leaders. 

While my team and I understand this concern, our view is somewhat less binary. For most writers, the task is rarely completely solitary. We work with collaborators, editors, facts checkers and copy editors. Rough drafts get polished. Recordings get transcribed. Large documents get summarized. Short ones get expanded. The question is, where can we insert our AI systems into the process in a way that enhances and improves the work? How do we collaborate with a machine?

To address this head-on, we started an experiment focused on where we could get the most out of today’s generative technologies. Our goal was to explore how a language model could work collaboratively with humans to create content. This led to the creation of what the students refer to lovingly as the “Kris Bot.”

The Kris Bot was designed with an existing workflow in mind. First thing in the morning, CASMI’s communications director sends me an article, usually about AI and safety. At 9:30, we get on Zoom, hit record, and I outline my opinion and answer any clarifying questions the communications director has. The recording is transcribed, and [my team] edits it into a 300- to 600-word article. Then [the communications director and I both] edit until we are satisfied — a nice process and collaboration. Once we are done, the system uses the text to generate a set of possible images as well.

The system is aimed at key goals. First, it helps us respond swiftly to a fast-paced world. Second, it allows us to preserve my voice while doing so. By offering the model more than just data — by feeding it my tone, voice and perspective — we’re attempting to create content that aligns closely with our editorial vision and point of view. And it does so in collaboration with us, becoming part of the process instead of replacing it.

Kris Bot isn’t here to replace me. It’s a tool designed to enhance my writing, enabling me to navigate the fast-paced flow of ideas and maintain my voice in the ever-evolving conversation about the role of AI in the world today.