flâneur

GPT and Pouring Language Through Shape

blog.humphd.org · 1,068 words · saved by 1 readers

I've spent the past few months building programs with GPT-3.  It's amazing and it's forcing me to rethink so much about how I write code or approach problems.  I was first inspired to try using it after reading Simon's great post about how to get started. Since then I have been reading as much as I can, most of it Twitter oral-tradition that's being passed down in realtime.  For example, the @goodside Twitter thread is particularly inspiring, and I've learned a lot from tricks I see Riley using. A lot of people are excited about ChatGPT and its human back-and-forth response style.  However, I've been more interested in seeing how far I can push GPT as a source of data.  I've spent a lot of time trying to force it to give me JSON data in a particular format.  It can be frustrating.  My own work is largely based on consuming or creating web APIs of one kind or another, and I've been exploring how one might leverage GPT to imagine new data sources and backends. Ethan Mollick has written t

I've spent the past few months building programs with GPT-3 . It's amazing and it's forcing me to rethink so much about how I write code or approach problems. I was first inspired to try using it after reading Simon's great post about how to get started . Since then I have been reading as much as I can, most of it Twitter oral-tradition that's being passed down in realtime. For example, the @goodside Twitter thread is particularly inspiring, and I've learned a lot from tricks I see Riley using. A lot of people are excited about ChatGPT and its human back-and-forth response style. However, I've

saved by

related reading