Channels
of-ai
in-sf
in-london
in-ontario
in-seattle
in-socal
in-alberta
of-functional-programming
reading-together
wormholes
of-end-user-programming
random-encounters
in-germany
in-israel
bot-dark-forest
of-music
present-company
of-logic-programming
share-your-work
devlog-together
introduce-yourself
thinking-together
in-boston
in-nyc
of-graphics
announcements
administrivia
two-minute-week
linking-together
Title
j
Jarno Montonen
01/23/2023, 8:02 PMChatGPT as a natural language abstraction for DSLs?
https://medium.com/@andreasmuelder/ai-powered-low-code-platform-by-example-how-to-use-chatgpt-to-abstract-from-domain-languages-625c3abf0e49
j
Jason Morris
01/30/2023, 7:43 PMIf all you are trying to do is get a usable data structure, and express it in a DSL, on the basis of prose, that's maybe doable. Anything more than that it gets sketchy. If you need more than a data structure, and if you need to extract it from something other than prose, it seems we still aren't there. Yet, anyway. But promising.
j
Jarno Montonen
01/31/2023, 7:50 AMInteresting. What makes you say so? Have you seen other examples or looked into this more? I'm interested to hear details if you won't mind typing some more 🙂
j
Jason Morris
01/31/2023, 6:50 PMYeah, sure. Personal experimentation, and things I've read about attempting to extract semantic meaning from legal texts (which aren't regular prose). I have been able to extract predicates from a paragraph of law, for example, but trying to extract logical expressions that model more than extremely simple statements in a FOL-like way has been frustrating.