Feeds:
Posts
Comments

(Im)Mortality Rate

(Our beloved co-blogger, while not fulfilling the blogging part of his job, managed to unearth this treasure trove of win…)

Medical statistics have always been a bit iffy at best, but the current collection of mortality studies apparently pushed some people over the edge. For instance the latest studies gaining popularity, on the health-imparting properties of coffee, are phrased thusly.

So, in a recently published comment in the Lancet, Ashley Croft and Joanne Palmer helpfully contend that the medical studies as quoted cannot be true. They helpfully state:

This cannot be true, however, since the risk of mortality is an absolute. 

As always, to every rule, there is a counterexample, and the careful authors dig up two known counterexamples from the Old testament. However, as they note:

In any event, neither episode occurred in the context of a randomised controlled trial.

To conclude:

We contend, therefore, that the risk of mortality for everyone—prophets included—is 1·0 (1·0–1·0).

The sort of careful science we should all aspire to. (I’ll add a link as soon as I can find one)

Advertisements

Insightful Analysis

As a relative insider to the sport of basketball, I’m not as caught up to the nuances of the sport as the experts. I did, however, fancy that I knew the basics pretty well… for instance I figured it was tautological that winning (see the connection to our wonderful blog?) required one to outscore the other team. Imagine my surprise when the experts at TNT insisted otherwise…

Every now and then, I am called to duty to report something wrong on the internet. I normally don’t respond, but every ever so often the person on the other side is someone I consider a friend, so argument might actually be fun rather than namecalling.

This particular episode corresponds to two pieces by Harsh Taneja – both pointing out interesting flaws in Google Translate (1, 2) that I’ll take his word on. Roughly speaking in the first, he suggests that there’s a problem detecting proper nouns, and if you try to ‘translate’ a proper noun, you could end up with gibberish. In the latter, he points out that GT can’t handle the fact that several languages (Hindi, Spanish) have a ‘formal’ and an ‘informal’ way of addressing people and it often messes up between one and the other. (He suggests that both might be because GT uses English as a go-between language, and makes perfectly good arguments to this end- though why he views this as some sort of dishonesty on GT’s part is beyond me.)  They make for interesting reads, except for one thing. I don’t understand the point he wants to make.

He titles his first piece, in part, the abject failure of Google Translate. This stinks of linkbait. GT can be less than good. For a lazy example, here’s one from the front page of Navbharat times via GT:

New Delhi .. The Secretariat of the Joint Secretary level officer of a female Army Chief Gen VK Singh’s lack of weapons, Prime Minister Manmohan Singh wrote about the ‘secret’ letter has been attributed to leaks. The PMO has denied this. Highly-placed sources told PTI on Sunday night, the female officer was removed from his important post. In intelligence he had rights. sources say that the official IAS (Indian Administrative Service) and IPS (Indian Police Service) are attached to the outside service. This action resulted in the leaked letter from Army Chief, IB (Intelligence Bureau) has been investigated since.
Poor, yes. ABJECT FAILURE?? Come on bro. And besides, this is a learning based algorithm. It gets better as it learns from examples (that’s what artificial intelligence means- like every other intelligence, it doesn’t know stuff, it learns from examples). So, a language which it sees more examples of, e.g. french, is unsurprisingly much more passable.

Francois Hollande will be sworn in as the new president Tuesday, May 15 after the handover ceremony held at the Elysee Palace in from 10 hours.

Then begin a timed course in Paris, which will start with the Arc de Triomphe , where Mr. Holland will pay tribute to the Unknown Soldier at 11:45, said a statement from the press service of the Elysee. This tribute is a tradition in the V thRepublic. In January 1959, when he took office, Charles de Gaulle had brought Rene Coty, the last president of the IV th , rekindle the flame of the Unknown Soldier, Place de l’Etoile in Paris.

Francois Hollande day continues with a tribute to Jules Ferry in the Tuileries Gardens at 13:45 and another at Marie Curie , the Institut Curie , at 14:30.Welcoming the former Minister of Education of the III e République and the Nobel Prize for Physics (1903) and chemistry (1911), the new head of state intends toring immediately “priorities of his five years” , hammered during his campaign, that are “youth” , “education” and respect for “secularism”

This, automatically done in about a second, is better than what the average student who studied french in high school would be able to serve up. In short, I’m missing the abject failure. Due to our omniscient Google, I now know President Hollande’s first week schedule, perfectly. Hurrah.
Harsh follows this up with an even more bizarre blog post. After pointing out that GT might mess up going ‘formal’ versus ‘informal’ second person (e.g. vous vs. tu in French)  he serves us this:

Anyway, till Big Brother achieves greater perfection so that humans can only learn the newest newspeak  (ref: Orwell’s 1984) – the language of the mouse,  I urge you to continue enrolling in real language classes and turn to real people for humanistic tasks!

 The best one could say is this is attacking a straw man, but that’s far too nice. Would the author like me to mail my Israeli friends and ask them to translate what their friends are posting in hebrew on a heated fbook discussion on Iran-Israel policy (debates I currently join via GT if I feel like it)? Or perhaps I should take a French class before I attempt to   book tickets on SNCF?
In short, if you need to speak a language well, or send a formal letter in a language you don’t speak, please don’t use machine translation just yet. It’ll get better, it has been for the past 20-30 (if slowly). But if you need something to get the rough point in a language you can’t even read, instantly, there’s never been anything in human history that does better. It’s got to be one of the greatest ‘Abject Failures’ to date. It’s free (if you dislike el Goog for some reason, Microsoft offers up an equally servicable alternative), easy to use (Google’s browser, Chrome, will offer to automatically translate pages not in your language, I’m sure IE9 can be made to do the same) and as I offered above, passable for most tasks other than blog snobbery.
As someone who loves the power of AI and the cloud to make my life that much better, I love reading people’s posts on flaws. These can only make these techniques better. Indeed, machine translation is not quite as far along as one would like, and these  points make for interesting reading. But I’d urge them to keep their “abject failures” and “google translator exposed through fundamental evidence” subtitles to themselves.
(edited shortly after posting to make the closing a little less combative)

Spring-ing to Life

In George RR Martin’s nowhere close to completion fantasy epic, Game of Thrones, the house Stark of Winterfell stays ever prepared for the long winters in their world with the house motto, “Winter is Coming.” After a long winter (nearly a year since our last post), we’re quite happy to declare that spring is here. Plenty of long dormant ideas for blog posts about spoonfuls of win will ensue.

Hopefully our readership will return… maybe you guys even missed us. 

Social Discrimination

Everyone here has heard about price discrimination. I know something about your willingness to pay (from other data about you or people like you), and use that to charge you a ‘better’ price. This has mostly been restricted by some combination of ethics, vague legal standards and technology to use ‘coarse’ information, e.g. your age (student/ senior discounts), your address (mailed coupons), and so on. As we pointed out a few months back there are cleverer methods on the way. But today, I think I’ve seen the best yet. A company called Klout (indubitably with the cooler K-based variant of the spelling) looks into your social network and offers a ‘score’  estimating the influence you have. Some geniuses have decided that one’s ‘Klout score’ might be a good way to discriminate on what website you see (and indeed, what free swag you get offered): http://mashable.com/2011/06/22/klout-gate/ . A few concerns:

1. Are we headed toward a world where along with my credit score, I need to monitor my Klout score? This seems several hops, skips and jumps away, but it’s all too easy to imagine a world where access to things I might want is contingent on my Klout score. Of course, one can argue that access to a lot of those things is currently controlled by things like looks, status, wealth with which one’s Klout score will be presumably correlated.

2. If we are, can a system like this account for my offline world in any meaningful way. On logging in, Facebook often prompts me with something useful like “xxxx – you haven’t talked to him lately. Write on his wall”. Where “xxxx” is my brother. And we talk (by phone) almost every day. While “I’m not on Facebook” is the new “I don’t have a TV”; it’s also true that me on facebook does not equal me in person.

3. It seems to me like there’s an easy way to ‘up’ my Klout score– link to every controversial piece I can find, or write about my thoughts on Gay marriage, Abortion, politics, and whether evolution should be taught in schools (be sure to click on that)- stuff I assiduously avoid airing, like most people. Whether or not this ‘influences’ people, enough people will comment to make me seem more Klouty.

Game theorists obsess about the difference between mutual knowledge (when everyone knows something) and common knowledge (when everyone knows something, and knows everyone knows ad infinitum). It matters a lot when it comes to figuring out how to behave in a strategic situation. If you don’t quite know what I’m talking about, consider reading (1), (2) or (3) from the masterful Jeff Ely at Cheaptalk on why this matters. (Actually read those pieces anyway, they’re some of my personal favorites and if you’re on spoonful it’s very likely you don’t have much better to do).

So anyway, there’s two recent examples of something that is (apparently) mutual knowledge being made common knowledge, and people finding this really troubling. The first was Obama’s statement regarding the 1967 borders. From what I can tell it was well known among all the participants that this is what any agreement would amount to. But I can also imagine that pro-Israel factions were unhappy with the symbolism of making such a public statement, and how it might effect the balance of power in any negotiations that might break out.

The more interesting one is a little brouhaha today regarding Duke Nukem Forever. DNF is one of those seemingly cursed games, having been in development for well over a decade (half a century in human years). It finally released this month, and was universally panned by video game critics. But then something weird happened- the PR company threatened to hit back.

(lifted off wired.com, full article linked below)

Now in general this seems like a heinous attempt to subvert free press by threat of reprisals. But to quote the wired.com article where I first learned about this:

Anyone who has done this job for any amount of time has suffered through a dry spell after giving a publisher a bad review, but this is the first time the threat of a blacklist has been made public.

Ok so it’s a troubling equilibrium in the video game review industry (as opposed to, from what I can tell, the gadget review industry). But beyond that, no harm no foul, right? They just stated something everyone in the industry knows is going to happen. So why the outrage? I understand the tug at the heartstrings reasons, but it doesn’t seem to make any logical sense. Someone just told you via a tweet something you already knew was going to happen. I’m missing something, but what?

Footnotes:

Original Wired Article

 

Final outcome


As a user of Facebook, and completely (I’d like to believe) divorced from my e-political/iPolitical views- I’m a little worried about the drop off in quality of the news feed lately. Roughly speaking, the new feed mostly seems to contain boring stuff from people who’re distant acquaintances, rather than all the cool stuff my close friends are up to. This has also resulted in a drop in my daily fbook usage to ~5mins a day (I think), where it previously would have been larger than I’d care to admit to on this blog.

4 hypotheses. In no particular order:

1. Stage of Life Theory: The feeds are the same ‘quality’- it’s just that I’ve aged a little (nearly 30, almost married) from the stage of life (early 20s, single) where I care about what ‘certain’  other people are doing (trendsetters, opposite sex), and this reflects in my interest in the feeds.

2. ‘Friends’: Where I once had maybe a 100 or so friends, I now have nearly 800 ‘friends’ on facebook (and I’m not even that popular). Roughly anyone I have ever met times the probability that one of us was proactive about friending the other. It could be that this is simply too many more people than I care about, and facebook can’t separate the people I care to know more about from the others. Maybe it’s a bit like approval voting : voting for everyone is like voting for no-one.

3. Quantity vs Quality: It used to be that people had to go to facebook and type stuff in, so they only did it when they had something interesting to say. Now, it’s a free for all- people share articles right from the page they’re on, there’s 4square checkins, FPlaces checkins, facebook likes being reported, tweets pulled in, instagrams posted and other technologies even I haven’t heard of. And as marginal cost goes down, so does average quality. (HT:PP)

4. Nothing’s changed with the feeds: I’m just a hater making stuff up (quite likely, but I try not to be too obvious in public).

Which is it? I’m genuinely interested…

%d bloggers like this: