“With today’s science we usually can’t show with high confidence that dangerous behavior isn’t there”
What makes a good AI safety test? Experts explain why even the best techniques may not be powerful enough
Headlines
