rcar

9 ms·

31.
by rcar 7y ago
The Propublica link includes both TARP and the government's investments in Fannie/Freddie, which provided heavy dividends and is where a large portion of the profits have come from. That aside, the common misperception of TARP is
32.
by rcar 7y ago
For sure. People still get up in arms about the 2008 bank bailout without realizing that it ended up resulting in a $121B profit to the government: https://projects.propublica.org/bailout/
33.
by rcar 7y ago
I was expecting an ad based on the title, but it ended up being an interesting analysis of just how much storage ends up costing them with their focused hardware setup.
34.
by rcar 7y ago
That's a pretty generous interpretation of > implementation-dependent approximation to the sine of x
35.
by rcar 7y ago
Not sure that I'd consider those all that eloquent since it's just the product of 2 different pieces of syntactic sugar (df.a being shorthand for df["a"] and df[<index filter>] shorthand for df.loc[<index filter
36.
by rcar 7y ago
I've worked extensively with credit bureau data, and there's a fair bit of misinformation both in the article and the comments. First, bureaus get very little detailed information about the episode of care that led to the medical
37.
by rcar 7y ago
To offer some help for 2 and 4, you can get a script out of a notebook with jupyter nbconvert <notebook.ipynb> --to python, which you can even include as a cell that starts with ! since that will run shell commands. For part of 3, the
38.
by rcar 7y ago
I'm someone who has been programming for a very long time and has been using notebooks for a reasonably long time (and almost always starts projects with them), my feeling is that they are a bit like C in that they make it easy to acci
39.
by rcar 7y ago
From the History section of the Wikipedia page: Jukeboxes became popular in the 1940s and were often set to a predetermined level by the owner, so any record that was mastered louder than the others would stand out. Similarly, starting in t
40.
by rcar 7y ago
Agreed. The biggest issue for players probably below ~5k tends to be making big blunders, and so talking through moves with a partner would make a large difference at that level but wouldn't be replicable in a normal pair go game.
41.
by rcar 7y ago
My usual would be "the first thing that comes to mind is X". If they have tried it, they know that I'm just throwing out ideas rather than calling them stupid, and if they haven't tried it, it's a gentle enough nudg
42.
by rcar 7y ago
FWIW, I'm not able to replicate that - getting ~70-80 requests with 2.9MB transfer/4.8MB resources on Chrome.
43.
by rcar 7y ago
Facial features can vary from ethnicity to ethnicity (see e.g., https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3074358/ ), and so a machine learning model trained solely on pictures from one ethnic group may not
44.
by rcar 7y ago
I'd argue that these sorts of efforts do help move things towards a better community. The site relies quite a lot on knowledgeable people to review and edit things, and the more grunt work that entails, the more you have to really care
45.
by rcar 7y ago
I agree, though they do make the pretty strong statement of NNAEPR suggests that one may abandon using NNs altogether, and simply use PR instead. in the blog post, which sounds to me like they're going for a finding.
46.
by rcar 7y ago
This is an essentially meaningless finding. Neural networks had the dark days they did in the late 90's/early 2000's because computational power didn't allow for the depth they needed to be applicable to most problems. O
47.
by rcar 7y ago
Firebase was an acquisition too.
48.
by rcar 7y ago
The comparison being made in that article is sort of a weird one and is not really reflective of banks' true reserve ratios (nor is the ~100% M1/reserve deposits that they were excited about true any longer). On https:/&#x2F
49.
by rcar 7y ago
I believe that you missed the /s implied at the end.
50.
by rcar 8y ago
This is a nice follow-on. The one additional thing I'd point out is that when a Class B driver gets the quote from this carrier of $1,684 but gets a quote of $1,100 from another carrier, they're likely to head for the cheaper opti
51.
by rcar 8y ago
This has a lot more to do with the risk pooling of insurance and competition among carriers than it does to do with some desire to mindlessly violate privacy. To give a simplified example, say a good driver costs an average of $100 per year
52.
by rcar 8y ago
There's a great web-based tool for making your own puzzles available here: http://www.keiranking.com/phil/ (with export to the .puz format). Really useful for both creating an initial grid and for suggesting possi
53.
Why we built our core engineering team outside of Silicon Valley(medium.com)
1 points by rcar 8y ago|0 comments
54.
by rcar 8y ago
I think he hit the primary reason in this paragraph from the article: We might need hundreds of hours of CPU time for a simulation, then nothing while we figure out what to do next, then another hundreds hours to run a modification. Now t
55.
by rcar 8y ago
"Most important"? My mother is a patient and detail-oriented person, but I wouldn't hire her as a software dev. To take a less extreme example, depending on the role I'd still rather hire someone who was occasionally imp
56.
by rcar 8y ago
There's a trailer for a documentary about this guy and his family here: https://www.fortydollarsapop.com/ Interesting to actually put faces to the people being discussed in the piece.
57.
by rcar 8y ago
Racial minorities are gong to be greatly overrepresented in a collection of mugshots, so it's almost certainly just the case that it's easier to make mistakes in that direction. If the algorithm is, say, 99.99% accurate on faces t
58.
by rcar 8y ago
There's a fun little programming challenge set up around this if you want to try your own hand at it: https://github.com/mshang/python-elevator-challenge
59.
by rcar 9y ago
There are plenty of things being called a bubble that aren't, but cryptocurrency is almost certainly one right now. The formal definition of a bubble is something that's priced far above its inherent value. Unless you want to argu
60.
by rcar 9y ago
When you're finished, you can use jupyter nbconvert --to python to get your work in script form.

More ›