Thread regarding Morgan Stanley layoffs

This report is misleading. I worked for one of the banks mentioned for fifteen years, the last five in the data science group that developed predictive models for the wealth management division. All models were subject to review by the Model Risk Management (MRM) group, which was independent and rigorous. On top of that, models were subject to the regulations and review by federal regulators like the OCC. The internal and external oversight was conducted by extremely competent data scientists who regularly forced changes to our models. I have not been there for a few years, but during my time, the only model type ever approved for use was a decision tree, despite other model types frequently performing better. Significant importance was placed on understanding why a model made a prediction. With a decision tree, you can follow the exact reasons for the prediction, while with other model types understanding the prediction can range from difficult to literally impossible (e.g. neural networks). The oversite I described was applicable to any algorithm that provided output used in a decision, whether it was a set of business rules in Excel or a machine learning algorithm. Having witnessed the commitment to model risk management, I find it hard to believe that all of a sudden, now that we are calling models 'AI', that either banks or regulators have relaxed. Of course, there will still be mistakes and potentially bad actors, but in my experience, the bank I worked for took this very seriously.

Fake News