Reaction
The ethics of working at OpenAI and the limits of embedded AI evaluators
Stella Biderman says she has told people for over a year that she doesn’t think working at OpenAI is moral.
TLDR
In a linked essay, Stella Biderman argues that embedded third-party evaluators can identify problems at AI companies but cannot compel change without government backing. She raises concerns about auditors’ limited access and commercial conflicts, and alleges OpenAI has repeatedly ignored warnings about cybersecurity and model monitoring.
Combined views
1.3K
1 Source, first seen ago