Are humans really blind to the gorilla on the basketball court? - Teppo Felin | Aeon Essays Scientific experiments don't generally attract widespread attention. But the 'Gorillas in Our Midst' (1999) experiment…aeon.co The takeaway is that the questions necessarily come before the answers. The idea that we can just let the data tell us the “right answers” simply works to confirm the bias that we have the right questions to begin with — or, in the lingo that I had used before, it brings us to the realm of the “right answer bias.” The human thing is to ask if we are asking the right questions, how wrong the right answer is with regards to both the original question and other questions, etc. This is something we are increasingly less able to do.

Most critiques of atheists and atheism is written from a believer’s perspective and focus on God. This is problematic. If the existence of the Divine were so obvious, there wouldn’t be any atheists. If God’s existence were so obvious, there wouldn’t be any believers. God exists in a realm where…

Long ago, I had wondered whether humans could actually pass the Turing test, given how they are “trained.” Now, there are excited talks about how Google Duplex allegedly might have passed the Turing Test, on the basis of some haircut appointment. …

This post is a reflection on this article from the Economist. One quote that baffles many who read this is the following: Indeed, these biases may also inform views about bias. Women are far more likely than their male colleagues to say that gender gaps are rooted in inequities in…

Suppose you believe that the mean of the “truth” is X, then you see a large set of observations whose mean is Y !=X. Do you still believe X? The answer depends on the variance of the distribution. …

The recent accident involving the self-driving car in Arizona made me wonder a bit about the distributional consequences of machine learning and the usual “right answer” bias that often accompanies such endeavors. The simple minded version of the problem is that the reality is messy: improbable events are not necessarily…

I don’t think I ever realized what N. N. Taleb might have meant by “My field is error avoidance” until I had a chance to reflect on Branko Milanovic’s review of his work. The implication of error “avoidance” is that errors of the model (i.e. the uncertainty) always exist and…

One of the things that I’ve been noticing time and again is that “data science” is not really “science,” but, in some sense, exactly the opposite thereof, but not necessarily in a bad way. …

The introduction to the so-called “regularization” problem always starts by invoking the bias-variance tradeoff. …