11 points | by breadchris 6 days ago ago
3 comments
Trivial parsers were trivial before LLMs, too.
Ironically, using the LLM to write trivial parsers loses out on the best part of the LLM: its ability to parse natural language into some sort of structure.
Using the LLM itself as the parser frontend seems more interesting all round...
"It's not just for _____, it's actually ____"
Cannot unsee this incredibly played-out LLM trope, especially in this case where it's used almost every other paragraph.
This itself was written by an LLM, no?
Trivial parsers were trivial before LLMs, too.
Ironically, using the LLM to write trivial parsers loses out on the best part of the LLM: its ability to parse natural language into some sort of structure.
Using the LLM itself as the parser frontend seems more interesting all round...
"It's not just for _____, it's actually ____"
Cannot unsee this incredibly played-out LLM trope, especially in this case where it's used almost every other paragraph.
This itself was written by an LLM, no?