This is prasad's Typepad Profile.
Join Typepad and start following prasad's activity
Join Now!
Already a member? Sign In
prasad
Recent Activity
Just a link dump, sorry, but this is a remarkable piece of writing: If books are essentially vertebral, contributing to our sense of human uniqueness that depends upon bodily uprightness, digital texts are more like invertebrates, subject to the laws... Continue reading
Posted Nov 16, 2012 at Accidental Blogger
Ruchira, I'm too cynical perhaps, but I suspect it's more than just speaking fees and foundations. I think Goldman Sachs/JP Morgan, or Microsoft in the antitrust days, or Shell or whoever wants that new pipeline, or RIAA and company, these people are probably making actual honest-to-goodness Cayman Islands transfers to persuadable congressmen, senators, high ranking officials, secretaries and the like. Then again, while Larry Summers is rich, he's probably ten million rich, not half a billion rich. I guess if you have the mindshare and are on the speed-dial, you don't even need the bank account number. John, Nate Silver is a genuine positive side to the attack of the quants in elections. May it continue, and reduce the noise of uninformed or wishful commentary. A dozen or so actual academic statisticians should take up the task too...some guy with a secret spreadsheet isn't enough! In fact, why aren't Gallup or Pew averaging polls themselves? What I'm worried about though is basically people like John Fund or Chuck Todd on TV yesterday, except with more info on individual localities - people who grew up in that state before going to college maybe. Think about it - an army of even two thousand at a 100k a year is 200M, which is starting to be within the budget of these campaigns. The Rove article is very interesting, and there's some truth to it, but Obama didn't invent the Akins and Angles and similar tea party types. The material was there to exploit, not like Swift Boat. I also lol'd to hear 2007 described a year when Americans didn't think of themselves as red and blue staters :)
1 reply
Re. shadow money and real under the table stuff - I have a weak hunch it's much larger than the declared spending. It seems incredible that with the most powerful job in the world on the line, total spending would be only in the few billion dollar regime. I'd think Lockheed Martin alone would spend in the hundreds of millions - they'd be crazy not to. Maybe they think it's more cost effective to buy politicians after election. That sort of spending and bribing is basically unregulated and unregulatable, since illegal to begin with. I don't know whether last night should strengthen my view re Citizens or not - they both spent oodles, and I don't know if that extra money played a decisive role in shaping views of Obama or Romney. There was a long article at the Washington Post about how Obama defined his election strategy, and one of the points made was the use of ad money to sculpt views of Romney, starting with the fundamental decision to focus on wealth and wall street, not flip-flopping or the lack of core beliefs. (Very nice article btw, if you haven't seen it. Supposedly what went wrong in the first debate was sheer overconfidence. Obama didn't prepare. http://www.washingtonpost.com/politics/decision2012/the-strategy-that-paved-a-winning-path/2012/11/07/0a1201c8-2769-11e2-b2a0-ae18d6159439_story.html ) Incidentally, after the results, I'm wondering if "highly number-aware election geek with lots of local knowledge" is going to become a booming and even better paid job description. The real superstars like Axelrod or Rove in '04 might be both too famous and too ideologically committed to be for inter-party sale, but at even one rung lower you might have action. You could see small armies of mathy, data rich election pros for sale to the highest bidder. If you get the best experts on a county, you win the election there, so it's a new competition. Say there are ~250 counties that "matter" so every campaign needs five people in each of them permanently on hire. It could be the new zero sum brain-drain for smart college grads, after high-frequency trading.
1 reply
Hi John, Thanks for the reply and links. - I love these cartogram maps, if only because otherwise the nation (especially if you go to the county level) looks basically all red. - That said, I didn't really see that "morphing by money" was doing anything in that video - if you weighted by predicted/historical margin of victory, or difference in party identification, you'd get the same map. The purpleness of the state attracts the money because that's where spending matters. This is bad, but the root cause of the distortion is the EC, not money. - The way to see the impact of money - in those states or anywhere else - would be to do some regression to relate difference in spending to outcomes, taking into account all the other usual suspects. Even such a relation if found would be tricky, since one reason a candidate might have trouble raising money is an enthusiasm gap, which reverses the arrow. - Scalia ends that interview on a really weak note when he suggests that voters know who's speaking even when they don't know who's funding the speech.
1 reply
Here's a question for the 6th of November, now that we've all been through two years of electioneering, but before the actual results influence (corrupt?) intuitions: did Citizen's United actually change anything big? There are many ways of getting at... Continue reading
Posted Nov 6, 2012 at Accidental Blogger
10
The Times article starts off on the wrong foot by assuming that software that classifies art (or anything else) by similarity needs to understand art deeply. However, classifying items of X isn't about understanding X; it's about understanding differences among items of X. And this task can be vastly simpler than understanding X itself. I have little or no idea how cars work, but child can tell a bus from a tractor. I use neural nets etc to classify and count physics events routinely, and quite often the machine is actually going to be better at this than I would be. I'm pretty sure my laptop doesn't know more about the underlying physics than I do. Figuring out that someone's singing a bandish (and not a thumri) in raga malkauns (and not rageshri) is a very different task from being able to produce high quality alap yourself. Having made this mistake (or so I see it) the reporter, predictably worried that his soul is being stolen by a machine, gibbers about 'reductionism.' But this is confused for a few reasons. First, there's nothing new or wicked about the idea that genres of music or styles of art have distinctive features. Much of what you do in a typical art appreciation course is to acquire those features - baroque music sounds like this, swing like that, prog rock like something else altogether. Styles of music and art have their rules, both explicit and tacit, and every student learns them before he creatively bends or ignores them. It's strange to forget that fact just because you're worried about your status as a sufficiently magical entity. Also, pattern recognition of this sort in a very real sense is anti-reductive. The "reductive" way of identifying some genre or style is to say "rock music needs to have more than X beats per minute, use at least five of the following instruments, and use/omit various chords." But that's not how recognition (both for people and for machines - BFPAFM) actually works. You indeed start with features like tempo or types of instrument used (if you hear a harpsichord it's probably old music!) or scales. (Pandora probably uses tens of features subclassified to thousands of subtler gradations. People, with better hardware and software, probably use even more.) But you never impose hard requirements on these things - the things you're classifying are fuzzy, multivariate and continuous valued. So you perform some kind of global/gestalt assessment and assign some overall weight. And that assessment involves both design and learning, using very complex ideas connecting different sorts of features. bfpafm there are going to be marginal cases where two or more labels both seem plausible. there are going to be hybrids of classes that ping part of the classification for multiple categories. bfpafm the labeler won't be able to give a crisp, succinct answer as to why any given instance was classified in a particular way - it's done with lots and lots of properties weighed in a complex way, not with a "reductive" simple series of logical decisions. Finally, I don't know what all art.sy does under the hood, but recommendation systems like Amazon or Pandora or Netflix actually get a fair bit of their classification power not from number-crunching the product but from number-crunching the users. If people who like movies X1 and X2 I've liked also like X3 and hate X4, then you can reassess the odds that I'll like X3 and X4 myself. This approach, far from understanding "reductively" or "non-reductively" doesn't bother with understanding to begin with - it's the equivalent of asking my mom what I'd like for a birthday gift.
1 reply
Agree about Geller. The post is my attempt to work out a compatibility between 'X is more civilized than Y' and 'X oppresses Y' that doesn't devolve to 'civilization is a -purely- self-serving metric.' I'm not trying to support Geller, she's a pretty iffy sort of person...
Toggle Commented Oct 21, 2012 on Support the savage (prasad) at Accidental Blogger
1 reply
There's an ad you can see in New York subway stations, that says "In any war between the civilized man and the savage, support the civilized man. Support Israel. Defeat Jihad." The NY subway authority initially refused to allow the... Continue reading
Posted Sep 27, 2012 at Accidental Blogger
One of the patents Apple successfully defended against Samsung recently is one on the pinch gesture to zoom using a multitouch interface. It has been pointed out that, given multitouch, this gesture is about the most obvious imaginable to accomplish... Continue reading
Posted Aug 28, 2012 at Accidental Blogger
16
Image
I had a vague sense that the Aurora movie theater shooting last month became a much larger story than the Wisconsin gurudwara shooting last Sunday. Google trends thinks the same thing, going by search volumes for Aurora and Sikh over... Continue reading
Posted Aug 10, 2012 at Accidental Blogger
Matthew Yglesias at Slate makes the case that Romney should simply shrug and say offshoring is a Good Thing, instead of making silly claims about "retroactive retirement." He says basically that it is good for jobs to be located where... Continue reading
Posted Jul 19, 2012 at Accidental Blogger
Uma Thurman look-alike pacifies the hordes with deep breathing, sukhasana and gyan mudra: There have to be cleverer ways of saying "The EU is a force for peace in the world." These guys really need new PR people. Continue reading
Posted Jun 23, 2012 at Accidental Blogger
15
As the high-end wine market confronts the problem of counterfeiting, a professional wine "detective" and sommelier explains how she identifies counterfeit wines: fraud detection has nothing to do with the taste of a wine, Downey says. “If you’ve got something... Continue reading
Posted Jun 16, 2012 at Accidental Blogger
18
The following is more of a set of musings than an argument, which musings I come by via this Slate article about giving men a choice in supporting children they didn't want in the first place. My interest in that... Continue reading
Posted Jun 8, 2012 at Accidental Blogger
33
Image
Jonathan Haidt is a moral psychologist best known for his work on moral foundations, the basic dimensions along which peoples' moral intuitions vary. These include care/harm, fairness/cheating, liberty/oppression, loyalty/betrayal, authority/subversion, and sanctity/degradation. Interestingly, Haidt's research suggests that while conservatives bring... Continue reading
Posted Apr 24, 2012 at Accidental Blogger
33
Just stumbled upon this marvelously loopy interview with the head of a Swiss Federal Ethics Committee. Said committee issued guidelines in 2008 pertaining to the dignity of plants. In Switzerland it is now permissible to harm plants, say for agriculture,... Continue reading
Posted Apr 14, 2012 at Accidental Blogger
12
I Did Not Attend the Funeral, But I Sent a Nice Letter Saying I Approved of It - Mark Twain If you gave him an enema you could bury him in a matchbox - Christopher Hitchens ( about Jerry Falwell)... Continue reading
Posted Apr 3, 2012 at Accidental Blogger
17
Michael Sandel has written a long, thoughtful, but frustrating article that raises questions about the intrusion of markets and market values into new domains. He begins with a long list of problematic goods and services that can now be bought... Continue reading
Posted Mar 16, 2012 at Accidental Blogger
29
Long piece at Salon about Jonathan Franzen, the internet, and sincerity, in the context of some recent controversy regarding things he's said about Twitter. This bit is unsettling: But it’s the discussion of a last conversation with his mom that... Continue reading
Posted Mar 12, 2012 at Accidental Blogger
Sujatha, I think that's the surprise for me...that there was so much more opposition from the cultural left than there was support from the cultural right. I suppose they miscalculated, but more than that I think they simply goofed; they should never have made so inept a decision, but having made it they might as well have stuck to their guns. Now, as you say, no-one likes them. Re Wilkinson, I know of his background, though it doesn't "bother" me particularly when it comes to reading him regularly. In any case, he's a left-libertarian in the sense of being socially liberal, so on abortion he's on my side.
1 reply
- I don't understand why Komen didn't do the obvious thing to start with and simply clarify that their funding of PP was narrowly tailored toward breast cancer prevention and screening. - I'm struck too by how much less clout pro-lifers have in the breast cancer pinkness gig than their opponents. Presumably Komen gets a lot more money from pro-choice people and organizations than it does from pro-lifer ones. This is surprising, the more so since abortion support/opposition itself is not really gendered according to polls. - Will Wilkinson made an amusing observation before Komen did backtracked: By itself, the predictable way this episode has everyone hunkering down in the familiar culture war trenches is completely boring. What's interesting is that Komen and Planned Parenthood are both largely admirable organizations doing necessary and urgent work for women's health, and this controversy is going to work out well for both of them. Had Planned Parenthood and the Komen Foundation colluded to profit by riling our culture's warring tribes, they couldn't have done it much better.
1 reply
The total amount of nerditude in the universe is fixed; it can only be converted from one form into another. Continue reading
Posted Feb 4, 2012 at Accidental Blogger
If you enjoy a good debunking, this one's a doozy. A scientist and contemporary violin maker conducted blind comparisons, getting professional violinists to try and choose among three old Cremonese violins, including two by Stradivari, (total value 10M) and three... Continue reading
Posted Jan 15, 2012 at Accidental Blogger
33
Travel to tribal portions of the Andaman and Nicobar Islands is fairly restricted, and access to particular local tribes is (in theory) hard to accomplish. Rationales for isolating these peoples are probably varied - lack of disease resistance, the desire... Continue reading
Posted Jan 11, 2012 at Accidental Blogger
17
The New York Times reports reports that a panel of scientists has recommended more research into geoengineering methods to control the Earth's temperature. Members said they hoped that such extreme engineering techniques, which include scattering particles in the air to... Continue reading
Posted Oct 7, 2011 at Accidental Blogger