I like using LLMs to explore long documents, engaging in a kind of conversational skimming. "Summarize the main points... What about X, etc." I find it easier to engage with the full document after this surface-level interaction. So, I created a custom GPT based on the dissenting opinion in Trump v US. Take it for a spin, and share it with your friends. https://chatgpt.com/g/g-m7RaN1h7T-dissenting-opinions-trump-v-us #scotus
@Colarusso Thank you for this.
Follow
@trabern @Colarusso Be careful. Once I had someone discussing my scientific article with me. I found their arguments somewhat strange. It took us a while to figure out that they used ChatGPT to read it and it had hallucinated content, so we were talking of different things.
@mapto @trabern fair warring. This is why I describe the process as skimming and as only the first step of engaging the text. One should always remember that all LLM output is a secondary source and should be treated as such. This of course is also why I have the bot start its answer with a link to the primary source.