Stanford Seminar - Collaborating with GPT-4: Ken Kahn, Oxford

May 10, 2023 Ken Kahn, Oxford Large Language Model Generative Artificial Intelligence Systems such as chatGPT and GPT-4 continue to evolve explosively. Ken Kahn has been collaborating with ChatGPT-4 to create an app for generating stories. Previously, I created an app that uses large language models (LLM) to generate, criticize, update, and illustrate stories - https://docs.google.com/document/d/1qZgT8GleQ2keOFxPNYXAFsZ8RPL3ob0M_FnVtQ7jWZ4/edit?usp=sharing This was implemented using blocks I had developed in Snap!, a block-based programming language. It runs by generating prompts for a selected LLM. I created over 20 illustrated stories using it. The release of GPT-4 with its very strong programming capabilities led me to wonder if I could recreate my app in JavaScript without entering any lines of code. I found I could. After 111 dialogue exchanges with ChatGPT-4 I had over 500 lines of JavaScript, over 150 lines of CSS, and about 50 lines of HTML. With a few minor exceptions, my role was solely to prompt GPT-4 and copy and paste as instructed by ChatGPT-4. The entire conversation with ChatGPT-4 is available here: https://sharegpt.com/c/DiSGDvS Ken’s talk will be followed by an Open Forum where anyone will be welcome to ask questions and to share their experience collaborating with generative AI programs. About the speaker: Ken Kahn is the author of ToonTalk and an expert at computer based animated storytelling. Learn more: https://en.wikipedia.org/wiki/ToonTalk #gpt4