Earlier this year I and Iohanna Nicenboim did some research into using Large Language Models to create chatbots to act as More-Than-Human Design Research assitants.
Personally, my desire to do this work came from considering a big issue in the More-Than-Human sphere. That is, while the motivation for exploring/doing More-Than-Human Design is very compelling and simple (e.g., paying more attention to the broader consequences of our actions), entry into actually doing More-Than-Human Design seems quite difficult and confusing.
Rubber Ducking and Oscilliation
During the project I became attached to the idea of likening how using generative AI in this way is somewhat akin to Rubber Ducking, hence the title and image associated with this post.
Side note: the Productive Oscillation paper I worked on unpacks this a little, and argues that in the context of doing Design Reseach, one might just cherry-pick the parts of the various theories that seem useful and generative without having to be overly concerned with the finer nuances.
The Bots
The two bots that we created - named MoTH and Oblique - tried out different approaches to building a bot, they have different aims, and different strengths and weaknesses - all this is discussed in the paper.
What stimulated this post was a chance conversation with Jiwei Zhou - I discovered that Jiwei had found the bots in the OpenAI GPT store and has been using them. !his was a (pleasant) surprise to me as I had forgotten that we published them that way!
The purpose of this post is to link to the various versions of the bots to invite experimentation with them. They are prototypes and come with no guarantees, but if you’re interested then please do have a go with them and see if they’re useful. Any feedback is welcome!
There are many ways to access these bots which may be or more or less relevance for you depending on your workflow and what tools you use, are familiar with, and maybe pay for.
MoTH
MoTH is a fully interactive chatbot. It uses this Chain of Reasoning approach. You can upload files to increase MoTH’s specialist knowledge on a particular idea or aspect of More-Than-Human theory.
You can access it directly here:
- Access MoTH via the OpenAI GPT Store (which usually requires a subscription).
- Access MoTH via the Poe platform, this time using the Claude 3.5 Sonnet model as the foundation (once again, this usually requires a subscription).
- The original version of the underlying prompt we used to create MoTH is available over here - enter it into the Large Language Model of your choice and adapt as you see fit.
Oblique
Oblique attempts to combine some of the idea-generating ethos of Oblique Strategies with More-Than-Human thinking. It is not intended to be a chatbot but rather a brainstorming tool that takes basic concepts and generates provocative routes forward.
You can access it directly here:
- Access Oblique via the OpenAI GPT Store (which usually requires a subscription).
- The original version of the underlying prompt we used to create MoTH is available over here - enter it into the Large Language Model of your choice and adapt as you see fit.
