The ‘stochastic parrot’ analogy and LLM usefulness
One commenter argues that LLMs can fit the ‘stochastic parrot’ analogy and still be powerful and useful, even with uneven capabilities.
TLDR
A user defends the ‘stochastic parrot’ analogy for the underlying neural network but rejects the inference that LLMs therefore can’t be powerful or useful. They describe those capabilities as ‘jagged’ and argue that LLMs show how much can be approximated and composed from ‘dumb, surface-level signal’—a prospect they find exciting.
Combined views
3.1K
2 Sources, first seen 1h ago