sh33mp

not bc of performance, but bc I hate OSXs Cmd-Tab ordering

That's a trivial but surprisingly reasonable reason for using two browsers. OSX has some annoying quirks sometimes.

The 20,000 for advanced degree holders already holds in the current system, and it already over-subscribed. So this would not reduce the number of H-1Bs issued.

The above commenter probably spoke a little too hastily. "Building Machines" is indeed not a paper about a neural network method, but a survey of problems that we should expect neural networks to do better on, but they currently aren't. That said, the paper isn't down on deep learning, but rather says "we need extra stuff, either additional methods or more inductive biases in our models".

Could you guys elaborate on the relationship between PyText, torchtext, and AllenNLP? I've briefly used the latter two, but with how quickly things are moving it'd be nice to have a quick answer from the devs themselves.

I grew up in a place practicing one-candidate votes (i.e., you choose between "in favor" that in the end will show 95+% and "against" with no alternatives) so acquired immunity to "citizen votes solve all problems" mentality.

I grew up in a place with effectively single party rule, and it was imbued into our culture that it's pointless to vote because you'll never unseat that party anyway. A big recent change was when, in a semi-recent election, an opposition party won an electoral constituency, worth a mere 5% of parliamentary votes, by the thinest of margins. I couldn't tell you what the opposition platform was today. But regardless of the fact that the opposition still wielded absolutely no legislative power, this led to an era of what many would consider a very electorate friendly legislative push.

This "don't vote for the lesser of two evils" business is nonsense. Use your right to vote relentlessly: to punish arrogant politicians, to press on single issues, to fight for the lesser of two evils because it is the LESSER of two evils. Finding and balancing the lesser of two evils is your job as a voter.

Even if your desired choice has no chance of winning, grinding down the margin, year after year, makes the other side nervous and more willing to compromise. Even if your desired choice has no chance of losing, expanding the margin gives them more room to take less "centrist" stances and push for the things you want. If there's a lesser of two evils, keep voting until the more of two evils has no choice but to compromise and become less evil. Lather, rinse, repeat.

I want to commend you for trying to learn more about the immigration process. More often than not, I find that Americans tend to not know hoops and travails that internationals have to jump through to work in the US, or even just to keep working in the US. Too often, I've heard "H1-B is for cheap foreign labor - just apply for an EB-1/2 or something."

"Evidence suggests that the heat death of the universe is unavoidable."

To put a futurist spin on it: advertising is the commercialization around the information bottleneck in the information age.

ULM-FiT and OpenAI's Transformer* are quite different. Both are pretrained language-models, but ULM-FiT is a standard stack of LSTMs with a particular recipe for fine-tuning, whereas the OpenAI's Transformer uses the much newer Transformer architecture, and no really fancy tricks in the actual fine-tuning. I suspect the difficulty is with the Transformer model itself - this is not the first time I've heard that it is difficult to train.

* = To be clear, this refers to OpenAI's pretrained Transformer model. The Transformer architecture was from work at Google.

Self-documenting code base, right?

I guess I was a little unclear in what I meant. I meant that everyone (prior to the much later 2DS) had to pay for the 3D tech that most people didn't really want. I absolutely turned my 3D off about a week in and never really used it again.

Just a gamer here, but there's something to be said about the success and failure of the 3DS and PS Vita.

The 3DS, based on hardware/system alone, should have failed. It was tremendously underpowered, and it forced a terrible 3D technology on all its users, which never really took off or became anything other than a novelty while raising the cost of manufacturing.

The Vita was a truly next-generation portable gaming device, with a beautiful screen, incredible graphics, properly analog sticks, an extremely modern interface, and great connectivity and human-interface (camera, capacitative touch) features. Even incorporating the cost of a proprietary Sony memory card, for the amount of power you got, I think it was very reasonably priced. Comparing the 3DS and Vita was like night and day in terms of a modern gaming device. Even today, I think it can stand head-to-head with the Switch in terms of portable gaming.

But Nintendo continued to pour resources into developing top-tier games for the 3DS, slashing its price to bolster adoption, and sticking to its still unorthodox 3D screen/touch screen combo. Whereas for the Vita, Sony quickly got spooked that the Vita didn't perform as well as the PSP, pulled first-party support and general marketing support, and major 3rd party developers (particularly outside Japan) fled the device.

The 3DS is now seen as a major success for Nintendo, while the Vita died (or is still dying) a slow and unceremonious death.

Sometimes sticking to your crazy guns works.

Agreed. I think it comes down to the presentation/interpretation of results. The response to "My classifier gets score of X" can be either "wow, that's a good score for a classifier, this method has merit" or "but X is not a good measure of [actual objective]".

So I think it's come down to conflict between

1. Which the author is trying to present 2. What an astute reader might interpret it as 3. What an astute reader might worry an uninformed reader might interpret it as

And my feeling is that, given all the talk about hype in pop-sci, we're actually on point 3 now, even when the author and reader are actually talking about something reasonable. Whereas personally I'm more interested in the research and interpretations from experts, which I find tend to be not so problematic.

This thread is a microcosm of this whole issue of overhyping.

On one hand, we have one commenter saying he can train a model to do a specific thing with a specific quantitative metric, to demonstrate how deep learning can incredibly powerful/useful.

On the other hand, we have another commenter saying "But this won't replace my doctor!" and therefore deep learning is overhyped.

The two sides aren't even talking about the same thing.

Ah, I was under the impression that RFs choose from a subset of features, not just one feature.

In any case, I agree with the thrust of your original comment that the specifications of the RF algorithm can be relaxed, usually for performance reasons, and still retain strong performance. But this goes back to my original comment that the performance considerations of random forests often aren't highlighted to new learners (whereas introducing ERTs to a beginner would probably shock them - how could you take totally random splits and still get any reasonable performance!)

I believe the formulation of random forests requires you to find the optimal split, albeit over a subset of features.

What you're talking about, where you simply generate a set of random splits across features, is Extremely Randomized Trees (https://link.springer.com/article/10.1007%2Fs10994-006-6226-...).

A funny thing about decision trees (or random forests) is how conceptually simple they are, but in terms of implementation they're very non-trivial.

There's always a point in the lecture or explanation where they go

So we just find the optimal split/feature based on entropy

which no one talks a ton about, but naively implemented is something on the order of O(kNlogN). For each split. Multiply that by the number of leaves (2^depth), and multiply that by the number of trees in your forest.

I learned this the hard way when I tried implementing random forests on GPU for a class (would not recommend: efficiently forming decision trees seem to involve a lot of data copying and shifting around). I actually learned a lot from reading sklearn's implementation of decision trees in Cython - it uses quite a number of neat tricks to make things really fast.

Mortal Kombat is very, very fun. I'm not sure I'd call it a good film, but I enjoyed it a lot.

Another good example is the Phoenix Wright movie (in Japan). Both are excellent love letters to their fans.

To this day, I'm still trying to understand why my freshman calculus class (proof-based), was as effective as it was.

It was 30 students, 1 lecturer, and a whole bunch of black board panels.

I think part of it was that it felt like a conversation, as well as a game. He would lay out the pieces (assumptions, definitions), and then point us in some direction ("now how would we show X?") We'd throw out ideas if we had any, and he'd either rebut us or nudge us in the right direction. I was 100% engaged in that class - no checking of phones or surfing the net - and it was just myself and my notebook, unlike in other classes that were slides-based and I picked up the bad habit of zoning out when something familiar was being covered. You can't do that in a conversation! The material and strategies I learned in that class completely built the foundation for my math major.

I am still not sure if math is one of the few topics you can take this approach for, slides are the devil, or if the lecturer was just secretly brilliant.

Facebook run fun psychological experiments to see if they can manipulate their users - without calling for volunteers first or anything I recognise as ethics.

I've always wondered: How is that different from A/B testing, or any other marketing experiment?

On the other hand, these criteria seem like they would encompass a good deal of all fiction.

Is this true though?

Sure, for a casual observer, these new methods for generating videos appear convincing, but is that the right bar to judge "ability to fake evidence"?

As far as I know, there have always been more sophisticated techniques and forensics to determine if an image is doctored, and likewise for video. I've not seen any research tackling fooling those methods yet, and I would bet that naive implementations of neural networks for generating videos would leave very obvious "neural network" artifacts. Of course, this is still new technology, so it will obviously get better at fooling our other tools over time too, but as of right now, I don't think the clamoring for "all evidence can now be faked" is all that justified.

Meritocracy is very much in conflict with equality. Meritocracy means you are judged on your merit. It is completely orthogonal to a nature-vs-nurture argument.

For example, this is exactly the sort of debate we've been having in Singapore, which prides itself on being a highly 'meritocratic', which consequently people have criticized for exacerbating income/social inequality.

But this runs into a problem - if you don't know that the piece was created by a machine, you might assume intent and judge it as an art. Does it stop being an art when you learn that there was no intent? What if you learn that about a human piece of art?

I don't see this as that big of a problem? Something can be beautiful without being art, but art tends to elicit a different kind of discussion specifically because there is intent behind it.

As a thought experiment, consider a particularly beautiful natural formation (like corals or intricately eroded rock). It can be beautiful without it being art. If someone then told me that it was actually sculpted by a person, then to me it's now art. I can ponder what the person was trying to express when they created it. If someone else then told me the previous person lied and it's really just a natural formation, then it's back to being not art.

If we were able to detect "intent" in an algorithm (although that is a hairy discussion in and of itself - arguably we could consider objective functions intents, but in that case I defer to the individual's interpretation as to whether that's intent or just clean study of a mathematical process), then yes, it would become art.

Here's a no-nonsense video chronicling the progress of the Donkey Kong world record: https://www.youtube.com/watch?v=LAQ2vEHq2N4

This guy has also made videos chronicling similar speedrun records for other games, and they're fascinating to watch.

My view for the longest time is that most of mainstream Economics is best seen as "argument by rigorous analogy". The post-Samuelson heavy emphasis on math has allowed for really rigorous analogies, and at least provides a common language for discussing economics rather than wild philosophical/sociological speculation, but in the end analogies are only as good as their original abstraction.

That said it should be noted that there are parts of Economics that do work really well when applied (e.g. certain market mechanisms where profit-maximizing rational agents are a reasonable assumption.)

My biggest concern with Economics is that there is too much money in Economics. Unlike something like Physics or Chemistry, there are too many stakeholders with significant financial resources that have an interest in having the mainstream thought lean one way or another and the big Economic systems are too complex for "wrong" theories to be definitively "debunked". I'm not insinuating any academic misconduct, but what is considered successful/fruitful/popularized research is surely affected by money, in terms of funding departments, funding research, or well-financed employment at various institutions. Think taxes, financial regulation, discretionary monetary/fiscal policy, and so on. You will probably find that the "boring" parts of Economics like education, food and transport to be relatively level-headed. My advice for anyone interested in any Economic theory is to ask of any claim 1) Who are the stakeholders, who stands to gain? 2) How easily is the claim falsifiable?

And of course always remember the Gell-Mann Amnesia effect.

He spent a couple of years in Singapore, so I'm going by the pronunciation there. (May be different in different regions, and I'm not sure about his preferred pronunciation now). It would do something like this:

Start with the word "urn". Now don't drag it out, make it short. Make the "n" and "ng" sound at the end (urng). Now take out the "r" sound (uhng).

It seems like Americans do tend to pronounce it "ehng" instead.

pandas was first developed at AQR Capital.

http://pandas.pydata.org/community.html

There's a great deal of misinformation and wrong-headed insinuation in this comment.

People love to talk about how bad Duerte is in the Philippines (he is), but in Singapore, 18oz of marijuana gets you a mandatory death penalty

There's a great deal of difference between endorsing extra-legal killings (Duerte) and having laws with extremely harsh punishment (Singapore). The reason for Singapore's harsh drug laws is that, at least up to 1-2 decades ago, it was a major transport/business hub in the region of significant drug trafficking. The overtly harsh policy was meant to fend off any chance of Singapore becoming any sort of drug trafficking hub.

Singapore is a 1 party system that has been controlled by the same family (the lee family) since what, the 1920s?

Lee Kuan Yew took office as Prime Minister in 1959. Also, this notion that the country has been "controlled" by the same family also needs to be put into context - Lee Kuan Yew and Lee Hsien Loong have been two out of the three Prime Ministers in Singapore's short history as a nation. Many members of the extended Lee family hold prominent positions in the private sector and government-linked corporations. So does the family hold a lot of influence? Highly likely. But is it "controlled" by the family? I'd like to see more evidence of that.

there is a $700 fine if you happen to spit it on the street Which is akin to a littering fine.

They very much censor media / free speech. If they didn't, this article wouldn't be news. Do they, though? There are 4 broad classes of "censorship" you could talk about:

1) Yes, there are strong laws against inflammatory statements relating to race and religion. These are significant, do curtail freedom of speech, and should be challenged or at least debated. Unfortunately, Amos Yee ran afoul of this. This is the only reason he's in trouble.

2) There is a requirement for websites having viewership above 50,000 a month to register as a "news organization"[1]. I personally strongly disagree with this, but also this has to date never been used to curtail speech. (Note: I'm not a fan of the "but this law/policy hasn't been abused yet" as a defense for a bad law/policy. I'm just stating the fact that it hasn't.)

3) Outright incitation of violence. I'm not sure how this varies across different countries/states, but I don't think this uniquely Singaporean.

4) Libel/Slander. These laws are standard. Yes, the Lee family has gone after opposition members who make factually false attacks.

But nested in the claim that "free speech is censored" is the insinuation that people cannot criticize government officials and policy. There is almost nothing further from the truth. In fact, Singaporeans are regularly self-deprecating how much they complain and rag on the government for bad policies. There is nearly no curtailment of the freedom to criticize, attack or badmouth policy, government or government officials (up to libel and slander).

There is also no illusion about the biases of media Singaporeans consume. Everybody in Singapore grows up knowing that the major news outlets have a pro-government bias, being closely linked to people in government. Everybody in Singapore grows up hearing the cliche that the history lessons are "government propaganda".

Singaporeans are among the most internet-savvy country-population in the world[2][3]. The Internet in Singapore, barring a laughable "symbolic list of 100 websites blocked"[3] is unfiltered. People who want to find alternative views, who want to organize against policy, can.

Littering or any petty theft can result in a caning, which depending on the severity, is split up over two days. This is to keep the punished from bleeding to death (seriously). Littering only results in fines. Vandalism can and has resulted in caning. No doubt, caning (especially in the prison form) is harsh, antiquated corporal punishment. But harsh is different from unjust.

Only the government elite can afford to drive vehicles. This is due to the taxes and registrations for even a 20 year old car being around 100,000 local currency. They force the public to use transit.

The first statement is so blatantly false I'm not going to bother refuting it. Cars are expensive and are a middle/middle-upper-class good. Singapore has a system whereby ownership of a car requires a license that costs up to tens of thousands of US dollars. This sounds over-the-top - but is also the correct solution from an economists' point of view. The government wants to reduce the number of cars on the road, so it taxes car owners. The cost of the license is determined by supply and demand via a market mechanism (e.g. the price has gone down during recessions). Meanwhile, the government plows significant funds into expanding the public transport network. Of all the Singaporean policies to complain about, transport is among one of the best thought-out, unless you already have a notion that a car is something everyone should own.

They require iris scans of every citizen and permanent resident (as of Jan 2016) As of Jan 2017[5]. And is this really different from finger-print scans? Is this any more intrusive?

---

There are many reasons to criticize Singapore, but please stick to facts.

[1] http://www.straitstimes.com/singapore/mda-rolls-out-licence-... [2] https://data.gov.sg/dataset/mobile-penetration-rate [3] http://www.internetlivestats.com/internet-users/singapore/ [4] https://en.wikipedia.org/wiki/Internet_censorship_in_Singapo... [5] http://www.channelnewsasia.com/news/singapore/authorities-to...

So could someone clarify this for me? Is is that the TPP negotiations are occurring in secret, or is it that the TPP will be passed without it being available for the public to view (which should be the case for most bills?)?

The former seems very reasonable to me (that negotiations happen in secret, and then the final product is presented to the public/congress for a vote), but up to this point I'm still not sure which is the case.