natalyarostova

11 ms·

31.
by natalyarostova 6y ago
Leaders that make decisions that have ramifications that are trivially worth 9+ figures are willing to pay the extra 500k to take their trust in the word of a consultant from 95% to 99%.
32.
by natalyarostova 6y ago
Anything that buys you the time you need to work through free resources. Time is the hard part for self learning.
33.
by natalyarostova 6y ago
It is what it is. If you want to be productive as a modern tech team just deal with it and use it and learn it.
34.
by natalyarostova 6y ago
Up until a few months ago I agreed. I have since been convinced that the power of Twitter/FB etc in connecting people who have never been connected is something we need to think carefully about...
35.
by natalyarostova 6y ago
You should use git. — someone who shares your job.
36.
by natalyarostova 6y ago
Great. Go work at a firm that reserves time for activism under a common cause. Plenty of us don’t want to do that at work, and will look for firms that let us focus on engineering for our time at work, so we can reserve our free time to do
37.
by natalyarostova 6y ago
My first instinct is to hate it. My second is to respect it and doubt my first instinct.
38.
by natalyarostova 6y ago
Depends on the problem space. For clinical drug trials you need people who can do statistics well. For some ML problems you just need engineers who can properly tune and deploy a DNN or forest. Then there is a large world in between, where
39.
by natalyarostova 6y ago
Is the implication here that subdivisions of humans on attributes should all be equally valid or invalid, irrespective of the attribute? Because that seems silly.
40.
by natalyarostova 6y ago
It depends on the inference you want to make. Average treatment effects are integrated over the entire sample. But that means it can “miss” or be biased for a subsample. You can then explicitly model that subsample, but sub samples by defin
41.
by natalyarostova 6y ago
The risk profile of vitamin d is different than chloroquine. Taking vitamin D on weak evidence still passes cost vs benefit.
42.
by natalyarostova 6y ago
I tried, and had the same experience as you. Lots of little things in Pycharm, like ctrl-clicking into arbitrary libraries and debugging in them, is hard to give up.
43.
by natalyarostova 6y ago
It took me four years before I built up the experience and confidence to advocate for myself and the length things take. Now I'm trying to protect newer data scientists" "Hey, new_hire, can you build this model by Friday?&quo
44.
by natalyarostova 6y ago
It'll be fun and motivating for a month or two, then the real work will set in. Unless you are prepared for that, tread carefully.
45.
by natalyarostova 6y ago
Election forecasting isn’t useful for those who don’t understand enough about statistics and forecasting to have the correct context to know how to read and use the information. That’s not a criticism of you, it’s a problem in the field. I
46.
by natalyarostova 6y ago
There is always more work to do, and your employer will take whatever you're willing to give.
47.
by natalyarostova 6y ago
Bertrand Russell may or may not have been right, but he was definitely an elite who felt he knew what was best for the masses, even if he didn't take his own advice. This is the guy that dedicated portions of his life to proving mathem
48.
by natalyarostova 6y ago
These US is just as bad as China hot takes are so stupid.
49.
by natalyarostova 6y ago
No twitter spam.
50.
by natalyarostova 6y ago
> significant life events can help one take a step back and see bigger pictures in life. I still get anxious some days. But nothing close to before a family member was diagnosed with cancer. Of course, everyone has their own response to
51.
by natalyarostova 6y ago
Maybe you're right. I felt the over-confident claim that it would fail to make it cross the line. If we discourage people from trying 1/100 or even 1/1000 chances because it's overwhelmingly likely they fail, we miss out
52.
by natalyarostova 6y ago
:)
53.
by natalyarostova 6y ago
Classic modern HN comment. The hard working person with a history of success who is trying a risky ambitious project will fail, because they didn't consider the few things a commentor instantly spotted as problems.
54.
by natalyarostova 6y ago
You'll always have people across the industry who want you to go work with them :)
55.
by natalyarostova 6y ago
> Machine learning is genuinely over-hyped. It’s very often statistics masquerading as something more grandiose. Its most ardent supporters are incredibly annoying, easily hated techbros This sort of fashionable disparagement of a group
56.
by natalyarostova 6y ago
Some people enjoy doing full stack DS and get paid well for it.
57.
by natalyarostova 6y ago
Deliberate practice in abstraction. Forcing yourself to solve abstraction problems at the edge of your current ability. This should be hard and frustrating . If it’s not, you aren’t solving hard enough problems. Do this daily and after a
58.
by natalyarostova 6y ago
I’m working on a project/product now with no established value or customer, yet. Imagine if I told sr. Mgmt “no, we still haven’t provided value. But we have ensured the design is robust and will last” lol. I’d be fired so quickly.
59.
by natalyarostova 6y ago
This is over intellectualizing and already hyped up topic. Predictive “ML” here is closer to an excel spreadsheet with crimes per block, sorted from largest to highest, with more police sent to the high ones. It’s not some weird deep net tr
60.
by natalyarostova 6y ago
I don’t know anything about the specific dynamics of the problem — so perhaps you’re right. But I have a general problem with the catalyst to being removed being a twitter mob. Whether it should have happened long ago, was deserved, or not.

More ›