Scott Alexander recently posted a twitter thread wondering if large language models (LLMs) like ChatGPT could be used to answer questions or to ask them.
The thread inspired this conversation with ChatGPT, exploring how LLMs could act as muses for inspiring human creativity, not oracles that provide answers.
The conversation covers different ways LLMs could be used to generate prompts for creative work, and how they could be used to challenge and develop ideas in a helpful way.
ChatGPT comes up with some challenging and interesting angles, and provides new references and ideas at each step, showing how it can serve as an on-demand thinking partner.
The conversation covers the ethical implications of using LLMs in this way, and whether there might be dangers of over-reliance on the technology.
Overall, Alexander finds the conversation useful as a muse, although ChatGPT’s responses are clearly from a machine and far from perfect.
Highlighted phrases were highlighted by the author in the original text.