You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Despite what it seems, I think ML might be suffering from everyone focusing on the local maxima of backprop-trained affine transforms (CNNs/LSTM/etc). The space of possible approaches is vast, and although this one has turned out to work okay, many alternatives remain possible
If thousands of researchers dedicate their efforts to making incremental improvements to a given technique, it's bound to progress and succeed (within its fundamental limits). But you could have seen similar progress if the same resources had been invested in a different paradigm
I think science moves faster when many different approaches are competing, not when everyone agrees on what works best.
The research community does reward outlandish ideas that *work* even more than run-of-the-mill ideas. Somewhere a researcher is toiling away against the tide and will come out victorious with new methods of learning. They will be handsomely rewarded for opening doors.
I have high hopes for the bayesian framework, projects like Stan are very promising IMHO
What do you think it might be worthy to explore in the future?
By making Keras so accessible and satisfying you did not help much in diversifying ML 
True, but not "everybody"! There is a small but healthy community working on neuroevolution, for instance!
"despite what it seems" ;)
Missing the bigger picture?
There was quite a while around 2013 2014 when dictionary learning and hoglike methods were matching benchmarks if much more expensively trained deep systems. Quoc Le wrote one good paper about cifar. And treating neural net nodes as primitive features works
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
