Mikilvæg hugtök í heimi gervigreindar
Er þetta dreifni eða bjagi?
“In relation to the test of significance, the experimenter is often not so much testing a hypothesis as interrogating Nature, asking her, as it were, a question; and randomization is the device which ensures that Nature’s answer shall be a valid one.”
— Ronald A. Fisher, The Design of Experiments (1935)
The algorithm finds that criminals have shorter distances between the inner corners of the eyes, smaller angles θ between the nose and the corners of the mouth, and higher curvature ρ to the upper lip.
Bergstrom & West í bókinni Calling Bullshit
A machine learning algorithm is only as good as its training data, and these training data are fundamentally flawed. […] The algorithm in this paper is not picking up some underlying physical structures associated with criminality.
Bergstrom & West í bókinni Calling Bullshit
Behind every causal conclusion there must lie some causal assumption that is not testable in observational studies.
Judea Pearl
This issue arose in part because the researchers relied on existing data rather than new data collected specifically for the task. Generally, collecting data is expensive and time-consuming. So, some AI developers claim that existing data is enough to make good decisions.
Narayanan & Kapoor - AI Snake Oil