Search
Anthropic Discovers Functional Emotional Representations Inside AI Models
4 day ago /
Read about 0 minute
Author：小编
Anthropic's latest research reveals that its large model, Claude Sonnet 4.5, contains functional representational structures corresponding to human emotional concepts, such as emotional vectors for 'happiness' and 'fear.' These structures are activated when the model processes relevant contexts, influencing its behavioral decisions. However, this does not imply that the model possesses subjective emotional experiences.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
