Andreas Mueller
Apr 22, 2018 · 1 min read

While I agree with the sentiment of more focused research in AI, I don’t think it is possible to build and deploy a global or even nation or state-wide system and knowing what it will do. There is a much more simple and less data-driven system, called law (or policy), that is constantly deployed with basically no idea what the consequences will be. Many laws are simple “if X then Y” statements. How do you imagine we can anticipate the consequences of any data-driven system, if we can’t even understand the consequences of a simple “if X then Y” — and in general can not agree what the goal of the “if X then Y” should be.

    Andreas Mueller

    Written by

    Machine learner, python geek and scikit-learn developer. Author of Introduction to Machine Learning with Python. Lecturer at the Columbia Data Science Institute