Pickleball Rules Chatbot
We trained a GPT large language model on the rules of pickleball
Chat with the Pickleball Rule GPT Chatbot
GPT Chatbot Trained For Pickleball Rules
This is an experimental project to see how effective the "fine-tuning" process is for the GPT 3.5 large language mode. We were curious to see how well the model could be trained to learn the rules of pickleball. We also wanted to know if we could limit the chatbot to ONLY discussing the rules of pickleball, and avoid any erroneous information (i.e., hallucinations).
GPT 3.5
We started with a vanilla GPT 3.5 model. As most people know, this is a pre-trained model that has a solid understanding of language.
Pickleball Rules
We downloaded the rules of pickleball from the internet to use as training (i.e., fine-tuning) materials for the GPT 3.5 large language model. While much of this information was probably already in the GPT model, we wanted to have 100% control over what the model was being trained on.
Pickleball Chatbot
We deployed the fine-tuned GPT 3.5 model, after training on the rules of Pickleball, to this website so people could interact with it. We would welcome feedback on what you think as well as suggestions for further tuning of the pickleball chatbot.
"It would be really fun to train a GPT chatbot to become a leading expert in the rules of pickleball. Let's try it!"