The burden of proof for claims that LLMs can think
A critic argues that “think” has no precise, testable meaning and that people claiming large language models can think—not skeptics—must define and support that claim.
TLDR
Calling “think,” “reason” and “consciousness” “suitcase words” loaded with emotional and historical baggage, a critic argues that claims about LLM thinking need testable definitions. They place the burden of defining and supporting those claims on proponents rather than skeptics. The critic also accuses Anthropic and others of having a vested interest in those claims and says the debate distracts from the technology’s benefits and harms.
Combined views
674
1 Source, first seen 1h ago