11:40 - 12:40 (UTC+01)
Talk (60 min)
Decision Making in The Absence of Ground Truth
Vanilla AI/ML works on a frequentist model of statistics, whereby you collect "ground truth" observations (say number of sales of ice cream versus daily temperature) and you use those observations to train a model to predict future sales based on the measured temperature. However, what do you do if you have no ground truth? Say you want to predict the risk of failure of certain strategically important assets, but you have no data on the assets? In this session, I will walk you through a solution I developed to enable a client to make sensible predictions of the risk they were carrying , despite having no relevant data.