“Normal accidents” and how well we understand LLMs
A commenter contrasts failures in other technologies that don’t necessarily signal design flaws with what they see as an especially shallow understanding of how LLMs work.
TLDR
Discussing AI agents, a commenter says other technologies can have “normal accidents”—failures that aren’t necessarily evidence of a product design flaw. They argue that our understanding of the mechanisms behind large language models is especially shallow.
Combined views
444
1 Source, first seen 1h ago
likes