At one end of the spectrum, we have AI companies that do incredible work on billions of $.
— Chip Huyen (@chipro) August 9, 2019
At another end, we have teams like @spacy_io (2 pax) & @huggingface (3 pax) that also do incredible work on single-digit GPUs.
The ability to advance AI field is resource-independent.
Nope. Scientists need to remember that their public mission is to present the truth, the whole truth, and nothing but the truth. Or quit and become a lawyer.
— Chris Chambers (@chrisdc77) August 7, 2019
This advice, in particular, is esp. bad and is why we have publication bias, and why so much science doesn’t replicate https://t.co/WBZKPt4X0v pic.twitter.com/bwNxjyUXms
After working for a while I realize how hopelessly outdated the concept "years of experience" is. It's so easy to spend years on a job and learn absolutely nothing.
— Chip Huyen (@chipro) August 7, 2019
In my opinion, the real cost of stuff is not the price you pay at checkout, it's the mental, emotional and logistical cost of having stuff
— François Chollet (@fchollet) August 5, 2019
Reviewers need to be extra careful before assigning any paper a score of 1 out of 10 scale, regardless of label. Unless they violated codes of conduct, give “wrong” papers a 2 with constructive comments instead. Avoid isolating people from the field forever, like in this example: pic.twitter.com/9SupyscolQ
— hardmaru (@hardmaru) August 3, 2019
Academia rewards hyper-specialization. Scholars narrow their interests until only a handful of people in the world can understand them.
— Arvind Narayanan (@random_walker) August 1, 2019
That means you can make ground-breaking discoveries and contributions if you have even an undergrad-level fluency in two or more fields.
I don't know about *good* students, but *great* students are those that will never again think about their grades—and their greatness has been recognized when no one will ever again ask to see their grades. https://t.co/bvGQSwAvXs
— Zachary Lipton (@zacharylipton) July 31, 2019
"we should move tech away from the center and towards the sides – as a tool to help people. Centering the algorithm feeds into our worst inclinations towards tech solutionism – the idea that we should find the 'one true method' and apply it everywhere." https://t.co/hkUuw54gQ7 pic.twitter.com/T3PS9pCfOc
— Rachel Thomas (@math_rachel) July 30, 2019
In my experience, this works best when you take the approach that "if one person in the meeting is remote, we’re all remote." Make the meeting a videocall (Hangouts, Skype, whatever) and have everybody call in.https://t.co/ylTqA77UWu
— Julia Silge (@juliasilge) July 29, 2019
Problems of racial bias & discrimination in new technologies will persist if we don’t address how racism actually operates in society. -- Rashida Richardson @AINowInstitute @techreview https://t.co/wJUJrdFYZa pic.twitter.com/cR2b55QSbF
— Rachel Thomas (@math_rachel) July 28, 2019
I now see it as a personal challenge to see how many different presentations I can sneak this slide into 😂
— Data Science Renee (@BecomingDataSci) July 28, 2019
(It was a favorite from my early "Becoming a Data Scientist" talks)
This is "what every data analyst and data scientist should be able to do" pic.twitter.com/PKAWKnBO1s
“I need a data scientist to build me a dashboard” is not a problem statement. It’s a solution statement!
— Angela Bassa (@AngeBassa) July 26, 2019
“I want help figuring out if my stuff is working the way I want it to“ is a proper problem we can solve together. pic.twitter.com/wM9vLQdM8e
