Sundial reposted this
New piece is out — "The Value of Correlations: Validation and Anomaly detection" (https://lnkd.in/g_rSeTCq) The first part ( https://lnkd.in/gRWG2cEV) of the three part series explored how orthogonal context gives AI the independent dimensions it needs to build a specific interpretation of a business and not a generic one. This piece explores the other side: how correlated metrics give AI the internal consistency checks to validate that interpretation and surface anomalies. Together, they point to the same truth: the quality of AI analysis depends on the type of context it has access to. The last part of the series which will be published in two weeks discussed how we can build a system to improve the quality of analysis.