In general, I am more impressed with people not from elite private universities. It means they didn’t have networks and prestige handed to them; they had to be scrappy.
— Natalie Jackson (@nataliemj10) June 28, 2020
Sure, you have to be good (or wealthy/legacy) to get in, but value of degree shouldn’t be advantaged. https://t.co/fbxpkPTO8M
Epilepsy causes seizures: the brain gets so overexcited that it malfunctions. One way to “cure” extreme epilepsy is to sever the connection between the hemispheres, making the brain less overconnected. Social media is the overconnection that causes epilepsy in society.
— Ilya Sutskever (@ilyasut) June 27, 2020
Some musings on NLP/ML self-study. tldr:
— Sasha Rush (@srush_nlp) June 26, 2020
* Reading a paper is really hard, give yourself a lot of time and slack.
* Pick depth over breadth. If you do depth well, you get breadth for free.
* Be actively skeptical of all claims. Read your own drafts with 2x that skepticism. https://t.co/f1taVV74tE
It's rare for speculative research ideas to end up being "correct" (at least in their initially-proposed form), so the value of sharing them lies primarily in the perspectives they open, not in their correctness.
— François Chollet (@fchollet) June 18, 2020
Be wrong in illuminating ways.
It's been a year since we last used GANs.
— Jason Antic (@citnaj) June 16, 2020
You may have been lead to believe that GANs are -the- way to get realistic results but believe me, there's actually better ways IMHO. I can't tell you exactly what we're doing now but I can tell you this:
2/
Decent people getting complacent about the status quo is often a bigger obstacle to progress than reactionary activism
— François Chollet (@fchollet) June 13, 2020
1/ With DeOldify being a real business now we invested in 4 Quadro RTX 8000 48GB video cards back in March. Expensive- definitely. Worth it? Totally. I'm now having the revelation of "oh...bigger models and bigger batches do help!" rather than endlessly chasing clever hacks.
— Jason Antic (@citnaj) June 13, 2020
"Have you run PCA on it?" is the data scientist's equivalent of "Have you switched it off and on again?"
— Neil Lawrence (@lawrennd) June 12, 2020
The scariest data is the missing data, whether it's child abuse reports that are not filed, unreported rapes, or missing police disciplinary records. We are seeing movement towards filling in some of these blind spots.
— Cathy O'Neil (@mathbabedotorg) June 11, 2020
My newest Bloomberg column:https://t.co/3OYFpwJIy4
Saw people talking about this topic again, so I'll re-share! In addition to the info in the quoted tweet, I also didn't even start this Twitter account or my podcast until I was in my 30s.
— Data Science Renee (@BecomingDataSci) May 10, 2020
"You're a failure if you haven't 'made it' by age X" is a fallacy, anyway. https://t.co/Syiz1jpfAx
1. Surveys are low-risk with a good expected payoff. But I believe most labs are hoping to find the next breakthrough, so they invest in ideas with high potential payoff. This is similar to startups. Labs play the role of the VC and they want to hire the next unicorn.
— Denny Britz (@dennybritz) May 8, 2020
We need better default plots for regression. https://t.co/P81NQrMXaR
— Andrew Gelman (@StatModeling) May 7, 2020
