Astra user reports unrelated criticisms during a fact-check
The user's example contrasts a specific claim about data centers not adding water pollution in one place with a general warning that water pollution can still be a problem.
TLDR
After asking Astra to fact-check something they'd written, a user complained that it produced "kind of fake unrelated criticisms." They illustrated the mismatch with feedback about water pollution being a problem in general, rather than addressing whether data centers had added pollution in a particular place. The user described this as a problem with large language models broadly, not just Astra.
Combined views
45K
3 Sources, first seen ago
