The message is very simple, from the book "Weapons of math destruction", by Cathy O'Neil - our current models used to cut through the data deluge are efficient but biased. People modelling them are biased and there is no incentive to improve them, even if there is collateral damage. Companies selling AI tools are not incentivized on making the "magic-formula" public, open to debate. The users discover the defects after years and after people's life are affected. For the majority there is no feed-back mechanism, or no way to correct the wrong-doing.
The book does not offer a solution to the existent AI used in crunching data. We are biased, but there is hope: openness, dialog and analysis of the results, especially for people who are affected by the algorithms.
Reading the book I remember reading about Google - where they did not know how to use the interviews and post-mortem analysis of the people hired in order to get the right formula of picking the right candidates. We are talking about a company where statistics, tweaks and tests are the norms in the industry. But hiring and comparing results are not very tangible compared with clicks and locations and patterns of search, although I am testing currently Facebook to see what can it infer from my browsing history.
The three attributes of WMD (weapons of math destruction) are opacity, pernicious feed-back loop and scale. It is insidious embedded in society and nobody is crying foul because it is also affecting segments in society. Untold by the author is the fact the WMD are perpetuating old models into present without accountability. The "Big Data", complex math is the veil thrown in public's eyes.
Summary: Read the book, the examples are a bit scary and worth to understand the others' people life affected by AI (or Big Data). If you are in Big Data, it makes you cringe.
The book does not offer a solution to the existent AI used in crunching data. We are biased, but there is hope: openness, dialog and analysis of the results, especially for people who are affected by the algorithms.
Reading the book I remember reading about Google - where they did not know how to use the interviews and post-mortem analysis of the people hired in order to get the right formula of picking the right candidates. We are talking about a company where statistics, tweaks and tests are the norms in the industry. But hiring and comparing results are not very tangible compared with clicks and locations and patterns of search, although I am testing currently Facebook to see what can it infer from my browsing history.
The three attributes of WMD (weapons of math destruction) are opacity, pernicious feed-back loop and scale. It is insidious embedded in society and nobody is crying foul because it is also affecting segments in society. Untold by the author is the fact the WMD are perpetuating old models into present without accountability. The "Big Data", complex math is the veil thrown in public's eyes.
Summary: Read the book, the examples are a bit scary and worth to understand the others' people life affected by AI (or Big Data). If you are in Big Data, it makes you cringe.
Comments