chris_va

43 ms·

541.
by chris_va 10y ago
A couple of points: Referring to an industry as "them" is not constructive. Comcast has a horrible reputation for internet service, does that mean we are all a part of "big tech" out to screw the little guy? Pharma compa
542.
by chris_va 10y ago
I get really skeptical whenever someone uses the term "Big Pharma". 99% of the time the person using it is selling something (hey, look, a $28 book). How dare a pharmaceutical company sell a product? How dare doctors decide to pre
543.
by chris_va 10y ago
The campaign, which cost an estimated total of $300 million, employed a strategy called ring vaccination that was credited to the American epidemiologist William Foege. Even with inflation, best 300M ever spent.
544.
by chris_va 10y ago
That was a nice read, though I do think they should've gone into more detail on sourcing and decision making. VCs (unless you are Sequoia) can get really stressed about good deal flow in sourcing. The 100:1 pitch:invest ratio is tellin
545.
by chris_va 10y ago
Languages: At some point in your career, the languages that you know stop being as important. Once you know 5-10, you'll pick up whatever your current company uses. You may have favorites, but beyond a certain point further specializat
546.
by chris_va 10y ago
I was assuming the bottleneck was protein output, but I cannot find the info in the paper at a glance.
547.
by chris_va 10y ago
If I am reading the original paper correctly, there is a bottleneck with the vegan diet that essentially requires you to only use cultivated cropland to get enough of a balanced diet. The vegan curve never flattens out, because it is entire
548.
by chris_va 10y ago
It is a very very old protein, stuck in local maximum more than anything else.
549.
by chris_va 10y ago
Point, but there is generally a specific timeframe for documents to combat that tendency. Usually ~25 years. The document in question here is only 20 years old, even if the information it references is older.
550.
by chris_va 10y ago
No one (well, almost no one) ever got fired keeping something confidential. Agencies always get referred to as "The NSA" or something, as if the collective consciousness of everyone working there should be merged. Let's repla
551.
by chris_va 10y ago
Kudos sir
552.
by chris_va 10y ago
They are the most sophisticated in the industry. They also handle something like 7T in assets. The insurance companies themselves are fairly empty pass-thru companies for a lot policy types. They all use the same forms, call centers, models
553.
by chris_va 10y ago
Heh :). I was thinking that the current "best" model/architecture may change fairly frequently. Obviously you wouldn't want to download 100MB every time the application starts, but maybe amortized every time there is a s
554.
by chris_va 10y ago
Well, this is fairly manual. More like: my_model = registry.get("tensorflow://github.com/asdf/models/imagine/latest") ... my_model.push("...")
555.
by chris_va 10y ago
Question: Does anyone know of a library for loading models/weights from a registry of some sort?
556.
by chris_va 10y ago
Combined with a form of fusion, we have all the energy we would ever need
557.
by chris_va 10y ago
lloyds of london/private investors
558.
by chris_va 10y ago
Reinsurance companies (which backstop pretty much all insurance in the US) also do this, with significantly more private data available. I would imagine there are quite a few companies that try this.
559.
by chris_va 10y ago
I would recommend looking at these programs: http://cty.jhu.edu/
560.
by chris_va 10y ago
And private tollways
561.
by chris_va 10y ago
+10 to Name of the Wind. But don't read it if you aren't willing to wait 5 years for him to finish his trilogy. The Martian was probably the best fiction in the last year for me.
562.
by chris_va 10y ago
This is not a well researched article. Various sweeping comments like: "Rent control is in effect, but that has just increased the incentive for evictions" ... ignore so many issues that it makes it hard to believe the core argume
563.
by chris_va 10y ago
I suspect we'll see supervised seq2seq generated code first. Like: --- Programmer inputs on left: compute std dev of x please AI on right proposes edit to code: + import numpy as np ... + stddev_x = np.std(x) --- Not super complica
564.
by chris_va 10y ago
If you had a specific design in mind, it might actually be more satisfying to generate the STL triangles directly. You can fit your parabola in 2D (by, say, computing reflection angle to a fixed focal point), then just rotate and dump out S
565.
by chris_va 10y ago
Not a bad question, and they attempt to address it in the paper with their multi-sample calibration: It is highly unlikely that oxygen could have been injected into halite inclusions without evidence of morphochemical disruption, and we mu
566.
by chris_va 10y ago
This sort of thing + headless browser makes for nice regression testing.
567.
by chris_va 10y ago
That's not entirely accurate. Microsoft's offer included patents, Japan, and Alibaba. Today that is all worth more than $45B, so their current exit is technically better.
568.
by chris_va 10y ago
Composition of matter patents got a lot harder for naturally isolated compounds circa 2014. We'll see if it gets granted. Having said that, a patent for general anti-biotic+matter is probably valid and it's unlikely that this comp
569.
by chris_va 10y ago
Looking at page timing, people seem less jarred by the slight contrast difference, even if they would have a hard time distinguishing the two. As for why, I am not sure.
570.
by chris_va 10y ago
As a takeaway from many many experiments, I would recommend something like #030303 over pure black, even if you don't want to embrace grey.

More ›