Could this RNN model perhaps be used to filter click bait headlines from HN automatically? Perhaps one could perform some sort of backward beam search to figure out how likely a particular headline would've been produced by it. If there are words in a headline that the model doesn't know, one could perhaps just let it replace it with one that it knows.
Maybe I’m mistaken, but it seems to me that Reddit would be much better off as a dontation-based non-profit. It seems ethical that nobody should make money off of user-generated content except the users are compensated for their time (which is unpractical); at least not beyond what’s necessary to keep the system up and running and occasionally add handy features to it. I bet the amount of time it takes to develop and run Reddit fades in face of the collective amount of time there is spent on creating and curating content by millions of people. I’m actually surprised that none of the comments here are touching on this point.
I’m not sure what this site is supposed to be for: Is it (a) to generate more ad revenue for reddit to be finally self-sustaining or (b) an attempt to come up to the expectations of Reddit's VC shareholders (i.e. a separate startup to generate more revenue than necessary for self-sustenance).
I think a cargo cult also has something to do with signaling, sort of like a status symbol ("They can't really be Y if they are not X!" -> "Look at us how X we are!"). It's a self-reinforcing meme that is used as a heuristic for value estimation, but usually fails catastrophically because of its heuristic and self-reinforcing nature.
Then lets steelman this point of critique: Why doesn't Apple keep their OSes in beta for a couple of months more to release something that is really stable? Is the sample size of beta testers really too small? I doubt 6 months or so later would be a big disadvantage to the competition, quite the contrary, I think releasing more stable 1.0s could be a considerable advantage.
I think the same applies to TeX distributions. It awfully slow for some reason, perhaps it double checks each file to avoid bit errors in system critical files?
I'm not sure whether evaporative cooling [1] could be a result of that, i.e. that if the sane quit, you are left with the insane. If that's the case, such a campaign would make things worse.
Helvetica Neue is not optimal either, especially due to the light weight used in iOS 7-8, but it I find it still acceptable.
For my taste San Francisco looks a bit too blocky due to the flattened curves along the vertical axis. It's not bad, but I like Google's Roboto typeface, for example, better.
Your grand-parent comment was about refuting the conjecture that it's due to helicopter parenting or a weakness of the students; then your parent comment was written as a supporting anecdote, and finally your comment claimed that the anecdote is not in support of the refutation at all, however without showing any efforts to explain why ("nananana").
I agree with that. San Francisco seems optimal for button captions, app names and other more unique, short pieces of texts, but seems sub-optimal for running text.
I think a similar development can also be seen in countries without any tuition fees, but my sample size could be too small to say that with certainty.
You've got no clue what you are talking about, do you? The thing is that many mega projects fail because of ridiculous reasons; mostly because high ranking people make dicisions outside the range of their expertise. Startup projects have very different characteristics and different stakeholders (i.e. mostly private). If 90% of the money could be saved then that's pretty bad.