The case for diverse perspectives and rigorous measurement in AI safety
One post argues that AI safety is stronger with different perspectives and experiences. A related thread calls for rigorous analysis of the probabilistic relationship between inputs and outputs.
TLDR
One post argues that different perspectives and experiences strengthen the pursuit of safe AI, warning that the biggest blind spots tend to appear when everyone thinks alike. A thread that quotes it calls for a rigorous science of input-output analysis, citing the probabilistic relationship between the two. The thread’s author says they and another witness flagged the need for rigorous measurement in Senate testimony.
Combined views
7.8K
1 Source, first seen ago
35 likes4 comments4 saves2 reposts