Humanities & A.I. Markov Convo

Welcome to the Humanities and AI Conversation Generator.

You can click "Thought" or "Reply" to generate text cards for an ongoing conversation.

The "Thought" version of P.D. is compiled from my one-to-all posts about the readings and exercises in my AI and the Humanities class. The source text is very often about ideas and my process notes from other projects.

The "Reply" version of P.D. is compiled from my one-to-one reply comments on my classmates' work, their projects, and some of the ideas they raised over the course of the semester. Their first names may appear in the text as a result!

---

Process note:

The images on this site were generated with ChatGPT's Pixel Art+ integration. I uploaded a few images of myself and asked it to generate two profile pictures based on the vibes of the images, with a short description of my general appearance; then I asked it to generate an image of the two of me sitting across from one another in a cozy cafe for the header.

The code that runs this website, both HTML and Markov.js, were produced with the help of ChatGPT 4o with Canvas. I prompted the structure of the website and linked in the images and text that feeds the two separate Markov chains that make up the two versions of myself, because they are two versions of me:

One reason why I frame these two versions of myself in conversation is because they wouldn't have actually ever spoken in class: "Thought" P.D. is speaking at large, to an audience of the professor and his peers, while the "Reply" P.D. is excitedly trying to make connections in a sort of "Group work" mentality. I think there are different affects in the work, as a result.

Other Thoughts

Working on this project helped me identify how I would go back and develop more adroit Markov projects, especially with the project (linked in the footer item "Humanities and AI Portfolio") that takes from all of my previous poetic work. I worked a lot with ChatGPT to adjust the Javascript so that sentences began with common startwords, not just with 'any' word, so that strings would be capitalized as regular sentences would.

I'm also rather shocked with how easily it was to specify exactly how I wanted this site to look. Most of the HTML was written on the first go, and it only took me two extra prompts to specify the aesthetics of this page. The 'thought/reply' structure with the cards and images were nailed on the first go! I'd love to use this structure to put poets and writers in "conversation" with one another (open source ones, of course ;).

Another application of this structure could be in building a renga, a traditional form of collaborative poem created by several poets casually over the course of an evening, or a cento, a collection of lines and found work from other poems and texts. I'd like to test the limits of this form to see how many different "bots" (to use that characterization extremely loosely) I could run at one time.

Closing Remarks

This semester was so transformative for me. I still feel like I have a lot of reading to do, but I feel like I've been brought on-board for these discussions in a way that is really helpful. This is AI and the Humanities, not "computer science and natural language processing," but I feel really excited to do more exploration in this field and I feel that I've been equipped with both the helpful tool that these models can be for coded projects and the language I need to prompt well in Python, Javascript, and HTML.

I will continue to use these models to help me code tools that I can use on non-generative AI research projects, but I also have a series of AI-oriented research questions I want to investigate from here on out:

What can other models do in the way of poetry? How can prompting creatively help me gain a better understanding of how I and AI process text to generate poetry?

Can I program an "erasure" poetry "generator?" Could I actually fine-tune a bot to speak in the kind of poetry that reminds me of myself and my own work?

What biases still stick out of the models when we ask them to perform certain tasks (like asking them to imitate or appropriate images and text), and how can prompting well help us pick up on the sorts of barriers that still exist between AI development and ethical use?

and I know I can do it, as well as approach a variety of creative projects, using the resources from this class! Thanks, Dr. Salter!