"Sometimes these people give the impression that their minds would be blown by a newspaper." 😂

it’s not necessarily a bad thing. It took me years to learn to recognize when it’s in my best interest to conserve my energy. easier to just do stuff but not always better

the great thing about academia is that we can all share the most

My point is just that nothing can be expected to be correct in expectation for a single finite sample

I kind of want one for "I fake it so real I am beyond fake", to go with my imposter syndrome

Nice post!

Note on this "more labeled data should translate into less variance and tighter confidence intervals for the corrected estimates, but they should be unbiased regardless"

Unbiased=asymptotically. Bias correction estimates can be noisy so you can def be biased for a fixed finite sample

Another classic

Visualization rhetoric never gets old!

So good

Our lab has produced multiple R1 tenure-track faculty 🎓, won lots of paper awards 🏆, & developed widely used techniques for uncertainty quantification & communication

Apply to Northwestern Phd in CS
www.mccormick.northwestern.edu/computer-sci...
or Technology & Social Behavior
tsb.northwestern.edu

The best is when you get to pass on these important ML concepts to your students, but minus a whole bunch of annoying assumptions