1/3 @GaryMarcus @tdietterich
— Murray Shanahan (@mpshanahan) October 25, 2019
On 'understanding' and the foundations of common sense: An agent’s behaviour should be systematically sensitive to the following four facts
1. that there is a world external to the agent to which its representations are answerable
I have a feeling most companies would benefit from thinking of data scientists not as crazy math professors but as journalists finding the scoop in the data: dumb shit the company has been dropping the ball on.
— Erik Bernhardsson (@fulhack) October 24, 2019
The advice to “grow thicker skin” is crap.
— Angela Bassa 🧙♀️🦉🧹 (@AngeBassa) October 21, 2019
No! My “thin skin” gives me empathy, keeps me grounded, helps me stay humble—and these are useful and valuable skills!
So LEAN IN to what makes you good; don’t try to squeeze into a mold that’s actually meant to keep you out. https://t.co/dnTZaBHQXA
I read somewhere that "hackers don't break systems, they merely demonstrate that the system was broken all along".
— François Chollet (@fchollet) October 21, 2019
This is true for democracy, too.
If you are not making progress on a solution, look for new problems.
— François Chollet (@fchollet) October 21, 2019
The only way to know if someone will pay for something is to put a price on it and see if they buy it. Asking “Would you pay $25 for this?” won’t get you an answer - it’ll get you a false sense of security (or insecurity). Why? It costs nothing to say “yes”.
— Jason Fried (@jasonfried) October 16, 2019
I wish there weren't people in our deep learning community that made it necessary for me to write things like this to people that are trying really hard to do high quality work. pic.twitter.com/l7xmbTRC4D
— Jeremy Howard (@jeremyphoward) October 15, 2019
So many things wrong here, but I want to focus on one thing that I see missing in my TL — The root of the problem here is the ML community has “broadend” the scope of publishing from being a means of scientific communication to PR and advertising. Should we be surprised by this? https://t.co/SfR8wnX083 pic.twitter.com/PFZ0uFrX5D
— Delip Rao (@deliprao) October 14, 2019
Machine learning now has an over-agitated culture that IMHO is bad for researcher wellbeing.
— Michael A Osborne (@maosbot) October 11, 2019
You're not a loser if hype, firehose arXiv, FOMO, feuds, overwork & random reviews cause you unhealthily-sustained stress, anxiety & gloom.
We didn't & don't have to have this culture.
When most people read a chart, they take away an overall message or story, not a list of high-precision values that they've read from every pixel.
— John Burn-Murdoch (@jburnmurdoch) October 10, 2019
And if precision really matters for the subject, they'll read a table.
Let's make charts that convey messages, not numbers 🙂 pic.twitter.com/mqNDHmDqMA
Tbh when I first moved to the bay I was perplexed as to why people thought coding was the hard part of data science.
— Rumman Chowdhury (@ruchowdh) October 9, 2019
It’s not. It’s the social science part of data science that stumps people: operationalizing concepts, deciphering variable interdependence, understanding impact.
If you want to be a good data scientist, you should spend ~49% of your time developing your statistical intuition (i.e. how to ask good questions of the data), and ~49% of your time on domain knowledge (improving overall understanding of your field). Only ~2% on methods per se.
— Nate Silver (@NateSilver538) October 6, 2019
