A communication and promotion expert would warn me to publish two posts a day in one blog. But I do not post for maximizing visits ...
Interested in machine learning I read about Andrew Ng and the one algorithm hypothesis and it came to my mind that I myself found in one or the other project that local intelligence and single programs could make clever decisions where we failed using more complex algorithms with special functions ....
I remember in the early 90s we managed to detect certain segments of images (full tone, half tone, edges, "watercolor", ..) from analyzing 9 pixels by a few fuzzy rules ...
However, what might be a breakthrough in machine learning making artificial units intelligent should not be applied to organizations?
Functional stupidity and stupidity management
And quite in parallelI I read A Stupidity-Based Theory of Organizations. It is about functional stupidity a term that might have some direct vector into the no-problem problem of risk management.
In general, it is reflected in an approach of organizations who do not consider solutions that are outside of a boundary, maybe derived from strategies, business principles and tactics. Intrinsic in business-as-usual and the-one-successful-approach-we-do-here-for-years scenarios. Functional stupidity is not only a negative thing - it maintains harmony and certainty. But negative consequences include the killing of innovativeness.
The authors use the term stupidity management describing actions discouraging, say, teams and individuals in the organization from thinking themselves. It may be called corporate identity .... often backed by hierarchies and bureaucratic rules.
How to avoid?
Good risk management does not work to fulfill the requirements of regulatory bodies. They know that might give them the false comfort of having optimized their risk. Red-blooded risk managers are reflexive use substantive reasoning and justify their actions.
And avoid systems that do not support multi-strategy, multi-model and multi-method approaches.
One-algorithm approaches might be great for a single robot, but not for an organization.
