Report
A dead-salmon analogy for overinterpreting LLMs' internal calculations
A post invokes a dead-salmon scanner anecdote to criticize how people interpret LLMs' internal calculations.
TLDR
A post quotes an account of a dead salmon whose scan reportedly showed brain activity while it was shown pictures of people and asked to guess their emotions. The author uses the anecdote to criticize what they see as widespread overinterpretation of large language models' internal calculations.
Combined views
—
2 Sources, first seen ago
— likes— comments— saves— reposts
