AI system baking
| RAISP02: How will you design the core AI system to meet the release criteria (baking)? |
|---|
Baking implements features directly in the core AI system, for example through choices of component model types, objective functions, training datasets, and training strategies.
Best practices
RAISP02-BP01 Design the core AI system to directly address your release criteria
RAISP02-BP02 Privacy: Build privacy-preserving mechanisms into the core AI system
RAISP02-BP03 Mitigate unwanted bias directly in the core AI system design
RAISP02-BP04 Build security protections directly into the core AI system design
RAISP02-BP05 Embed provenance indicators into core AI system outputs
RAISP02-BP06 Enable users to customize core AI system behaviors
RAISP02-BP07 Incorporate explainability mechanisms into the core AI system
RAISP02-BP08 Consider core AI system designs that improve factual accuracy
RAISP02-BP09 Design your core AI system to handle input variations
RAISP02-BP10 Build safety protections into the core AI system
