iraphael

I also liked this explanation of the Typicality assumption

http://scientifica.wikia.com/wiki/The_Typicality_Assumption

My understanding of the post is it would host learned models. You'd train them wherever, but host the learned model in algorithmia, which exposes it through an api, making it easy for others to use your model.

I don't know how much time you want to put into it, but the first few videos of the Stanford iOS development course (available on YouTube [1]) has a good explanation of how to layout things. I am, however, also a beginner to iOS dev, so maybe there are better resources I am not aware of.

[1] https://www.youtube.com/watch?v=GOEPVM5OzJk

I was wondering the same thing.

I'm pretty sure these kinds of regulations have existed in the US for a while too. According to [1], having an algorithm be interpretable (which NNs aren't) is a legal requirement in any financial decisions subject to anti-discrimination laws.

In this case, it doesn't necessarily mean there will be a HUGE CLASH. We will either turn to more interpretable AI, or continue developing more and more interpretability techniques for deep learning, which has been an area of very active research in the recent years.

[1] https://www.cs.princeton.edu/picasso/mats/Book_Schapire.pdf page 664

Interesting to see more js in hn, esp with collapse-comments.

For a while now, hn has (in)famously had only a couple of functions: https://news.ycombinator.com/item?id=11307758 (Not that the new features changed this dramatically).

Is there a statistical difference in accident rate between high and low income populations? If so, do we know what the fatality/miles rate is in the group that would be able to afford a tesla? When I read the blog post, I couldn't help but think that while tesla's rate might be smaller than the national average, it is also not a fair comparison to draw conclusions from.

Yeah I don't understand the reasoning behind this at all.

Truces are declared to some extent with patent pools. Not a lot of action towards lobbying Congress for patent reform, though (which is very needed).

http://www.wipo.int/export/sites/www/ip-competition/en/studi...

actually the second line won't be cleared. A hole will be left where you're thinking each of the top blocks would fall into.

Ohh, in that case, we could train the LSTM only with code that was used to generate it. Unfortunately, I think that's just not enough data to get interesting results. Maybe we could train it on multiple different kinds of implementations of LSTMs, but that may or may not be cheating...

Oh yes, those experiments would have been fun. But let me explain what I had in mind.

Potential Literature is about finding meaning in things. Meaning is human-made anyways, so anyone can find meaning in the most random of things. Having a randomly-generated, but still structured, text can allow for new meaning to come through. Think SubredditSimulator - generating random stuff, and letting the users find/create meaning for it.

I thought the same. Perhaps we can think of a way to do it? The only thing coming to mind is either using a markov chain or a char-RNN to generate code files. The latter has been done here: http://karpathy.github.io/2015/05/21/rnn-effectiveness/

That's an interesting solution. I'm on OSX too but I use Terminal.app. I'll try to find a similar feature for it or look into iTerm.

Thanks :)

Is there a mailing list for discussions of currently-proposed RFCs? I couldn't find it on ietf's website.

e: I may have found it: https://www.rfc-editor.org/mailman/listinfo/rfc-interest

My linkedin URL has a prefix "br." because I signed up when I still lived in Brazil. That makes Glassbowl crash and alert me the URL is invalid...

I understand the limitations, but I was excited to be able to read sites I usually visit while still immersing myself in the language.

Demand that control from the content provider. The man in the middle shouldn't touch your content.

When I first saw the article I was excited for it, because I thought Readlang had an "immersion"-type feature, where it translates every page I visit to french and un-translates on demand.

Not saying that would work better, but it's something I thought would be cool.

I think the idea is that, if you retake the test, your system will be in a different state (our organism's biochemistry fluctuates naturally). So if something in your system caused a false positive, there's a chance it won't be there when you retake the test and you will get a true negative.

Basically, you're thinking of your blood composition as stateless, when it may also be counted as an "external factor".

> The closure f does not capture the value of i at the time when f is defined, changing i after f was defined make nonsense of the idea of a closure.

I have an incomplete understanding of functional programming. Are there other uses for closures (besides readability and practicality) that still work if they don't capture the value of i? If not, is this something someone should suggest as a change in swift-evolution?

Apparently, Whatsapp is only end-to-end encrypted on Android, where it uses TextSecure. On iOS, it still uses RC4.

According to their pricing page ( https://www.jukedeck.com/pricing ), you can use the track free of royalty, but if you want to actually own it, you have to pay $199.

But this falls kind of in a gray area. If the AI created the tracks, why does the company own the copyright (and thus, have the right to sell it)? In December 2014, the United States Copyright Office stated that works created by a non-human are not subject to U.S. copyright (see: https://en.wikipedia.org/wiki/Monkey_selfie#Copyright_issues ). So, in theory, AI could also own copyright.

Moreover, do they actually check every newly generated track to make sure its not too similar to previously-sold tracks?

Another cool article from this website: http://koaning.io/canvas-trees-and-communities.html

I also appreciated the fact that they overrode the browser's default scrolling to a less natural speed, and in doing so, also prevented me from being able to use gestures to return to hacker news.

I wouldn't place it in the uncanny valley. It just sounds highly scripted and acted.

My bad, you're right. I mistook Netflix's family price for the starting one I think

The method used for mimicking is not the same as the learning method in supervised learning.

From this article (which OP's article links to): http://www.eurekalert.org/pub_releases/2015-12/uow-url120115...

> research has shown that children as young as 18 months can infer the goal of an adult's actions and develop alternate ways of reaching that goal themselves.

> In one example, infants saw an adult try to pull apart a barbell-shaped toy, but the adult failed to achieve that goal because the toy was stuck together and his hands slipped off the ends. The infants watched carefully and then decided to use alternate methods -- they wrapped their tiny fingers all the way around the ends and yanked especially hard -- duplicating what the adult intended to do.

I think the idea behind mimicry is goal identification (i.e.: see human wash a single dish, and identify the goal of 'getting all dishes clean'), not 'learn the way humans would perform this task and then do it'.

Or even the word socialism

Would having movies and TV shows make Youtube Red appeal to you, at $9.99/mo? It's cheaper than Netflix.. But I guess that will depend on the shows they get the rights for.