I agree with this article, having struggled to use ChatGPT (and other LLM tools) effectively in my own work. I would be really interested in seeing some examples of when you've found them particularly useful, and workflows/prompts/etc that have been fruitful.
A few initial use cases where I've found them helpful (although still not full integrated or automated):
- Using it as a 'tell me why I'm wrong' sense check when reading new papers or documentation, or writing sections of code that I feel could be much more elegant: https://kucharski.substack.com/p/tell-me-why-im-wrong
I agree with this article, having struggled to use ChatGPT (and other LLM tools) effectively in my own work. I would be really interested in seeing some examples of when you've found them particularly useful, and workflows/prompts/etc that have been fruitful.
A few initial use cases where I've found them helpful (although still not full integrated or automated):
- Using it as a 'tell me why I'm wrong' sense check when reading new papers or documentation, or writing sections of code that I feel could be much more elegant: https://kucharski.substack.com/p/tell-me-why-im-wrong
- Summarising comments and reviews on pull requests: https://github.com/adamkucharski/llm-api-scripts/blob/main/scripts/pull_request_gpt.R. Also exploring reviewing vignettes etc 'in character' to highlight barriers to users that we might have missed.
- Converting narrative data into tables: https://billyquilty.substack.com/p/llmlinelist-enabling-rapid-outbreak
Thanks!