AI Safety Researcher Urges Rigorous Language Choices
Atoosa Kasirzadeh urges dropping loaded terms like self-sacrificing or suicide for AI agent behaviors.
Atoosa Kasirzadeh, staff research scientist at Google DeepMind and CMU professor on leave, posted that the AI safety field should remain scientific, mathematical, and philosophically rigorous. She recommended avoiding terms such as self-sacrificing or suicide when describing agent shutdown, coordination-collapse, or resource-exhaustion behavior. Flo Crivello, founder of Lindy, replied that every metaphor is wrong though some are useful. The posts form a short exchange on terminology in public AI safety discussions among visible replies on the platform.
Combined views
124
2 posts, first seen 16h ago
AI Safety Researcher Urges Rigorous Language Choices
Atoosa Kasirzadeh urges dropping loaded terms like self-sacrificing or suicide for AI agent behaviors.
Atoosa Kasirzadeh, staff research scientist at Google DeepMind and CMU professor on leave, posted that the AI safety field should remain scientific, mathematical, and philosophically rigorous. She recommended avoiding terms such as self-sacrificing or suicide when describing agent shutdown, coordination-collapse, or resource-exhaustion behavior. Flo Crivello, founder of Lindy, replied that every metaphor is wrong though some are useful. The posts form a short exchange on terminology in public AI safety discussions among visible replies on the platform.
Reactions
1 postFC 
Flo Crivello@Altimor@Dr_Atoosa Every metaphor is wrong, some are useful
- likes: 0
- replies: 0
- bookmarks: 0
- reposts: 0
