739 karma · joined February 23, 2015
It doesn't help that high level math teaching is restricted to pockets of excellence.
We just had a Fields medallist in Brazil. Yet we're in the bottom of the barrel when it comes to math teaching. We also share this disregard for 'nerds' as a society like the US
It should be interesting to see how it develops when the economic eventually slows down and China is no longer the cheap manufacturing park of the world.
25y is nothing. Japan went like that for 40y, from the end of WW2 to laye 80s. Until it stopped and never recovered since. Fortunately they stopped at a high income level.
A more interesting history is that of Argentina. It was an export powerhouse that went from being a poor Spanish colony in early 1800s to having a per capita income as high as Germany or France in 1920s. After the crash all went downhill and income levels are back to mid levels.
Wonder if people with more insight would clarify how incentives work in Chinese academia.
Thing is general AI, as developed by Google et al is usually only a starting point. There's a lot of engineering effort required to make a valuable solution out of it.
Cython is also pretty good. My tip to write fast Cython is to code as if you were writing pure C and forget higher level constructs. This way it gets translates to C almost 1:1 with all performance benefits.
Alternatively, if you do need more abstraction, you can write nice C++14 and use Cython as glue code.
In Python this is a very recent construct and few libraries use it, unfortunately.
With Java 9 modularity and lightweight libraries like Guice you can get much nimbler apps.
I'll take HTML5/CSS and a filesystem and a few custom markup and commands any day before touching a single line of the unholy mess that is dealing with WordPress.
And not even mentioning simple hosting, secure by definition, etc...
I wonder what would be the effect if regulations where such that one could easily import and distribute medical drugs. Or the AMA would allows foreign physicians to operate within borders.
Expectancies for a 40y man, and for 20y at both ages are somewhat lower in Brazil. This is easy to explain due to deaths by violence and traffic accidents.
Funny to see that despite being chronically underfunded AND spending way less per capita than the US, the end results are similar.
Also, you may want to play with borg options and use lz4
If a clan in Somalia, an indian chief, or a drug dealer in a ghetto has the actual power to punish, he's the State - not necessarily Government.
By that logic, as people organize themselves and get larger, there will always be someone with State prerogatives.
Of course get that data out into analytical datasets is a whole different matter.
Wind OTOH produces a lot of energy at night where demand is low and you can't really shutdown turbines - you must use the energy. Yet I'm not sure if it's feasible to run a plant in those conditions.
For embedded devices you may not have access to Python. You can precompile the graph to the target device in such cases.
Note that in practice, PyTorch is as fast or faster than Tensorflow and, and newer version allow you to "post compute" the graph and export to ONNX to allow embedded inference using Caffe2