Report
A formula could flag undesirable AI outputs before they happen
Live Science says the formula tracks an AI's internal “compass needle” to flag undesirable outputs before they happen.
TLDR
Live Science reports that physicists say simple math can predict when AI is about to “go rogue.” It describes a formula that tracks an AI's internal “compass needle” to flag undesirable outputs before they happen.
Combined views
649
1 Source, first seen ago
