Guidelight AI launches v1.0 capability testing standard focusing strictly on an AI system's capacity to cause harm
The framework outlines seven principles for evaluating catastrophic risk scenarios
#1203Steven Adler@SJGADLERStrong testing is the bedrock so that many other AI safety practices depend on
We've just published @GuidelightAI 's v1.0 standard for it; let us know what you think!
Tegan McCaslin@teganmccaslin21/ New from @GuidelightAI : standards for AI capability testing, built around 7 principles for measuring the abilities of an AI system relevant to catastrophic risk.
Read our full Capability Testing standard here: https://www.guidelight.ai/capability-testing
Steven Adler@sjgadlerStrong testing is the bedrock so that many other AI safety practices depend on We've just published @GuidelightAI 's v1.0 standard for it; let us know what you think!
#1203Steven Adler@SJGADLERStrong testing is the bedrock that so many other AI safety practices depend on
We've just published @GuidelightAI's v1.0 standard for it; let us know what you think!
Tegan McCaslin@teganmccaslin21/ New from @GuidelightAI : standards for AI capability testing, built around 7 principles for measuring the abilities of an AI system relevant to catastrophic risk.
Read our full Capability Testing standard here: http://guidelight.ai/capability-testing
Steven Adler@sjgadlerStrong testing is the bedrock that so many other AI safety practices depend on We've just published @GuidelightAI's v1.0 standard for it; let us know what you think!