![]() |
TCHS 4O 2000 [4o's nonsense] alvinny [2] - csq - edchong jenming - joseph - law meepok - mingqi - pea pengkian [2] - qwergopot - woof xinghao - zhengyu HCJC 01S60 [understated sixzero] andy - edwin - jack jiaqi - peter - rex serena SAF 21SA khenghui - jiaming - jinrui [2] ritchie - vicknesh - zhenhao Others Lwei [2] - shaowei - website links - Alien Loves Predator BloggerSG Cute Overload! Cyanide and Happiness Daily Bunny Hamleto Hattrick Magic: The Gathering The Onion The Order of the Stick Perry Bible Fellowship PvP Online Soccernet Sluggy Freelance The Students' Sketchpad Talk Rock Talking Cock.com Tom the Dancing Bug Wikipedia Wulffmorgenthaler ![]() ![]() ![]() ![]() ![]() ![]() |
bert's blog v1.21 Powered by glolg Programmed with Perl 5.6.1 on Apache/1.3.27 (Red Hat Linux) best viewed at 1024 x 768 resolution on Internet Explorer 6.0+ or Mozilla Firefox 1.5+ entry views: 281 today's page views: 190 (12 mobile) all-time page views: 3248337 most viewed entry: 18739 views most commented entry: 14 comments number of entries: 1215 page created Mon Apr 21, 2025 10:37:42 |
- tagcloud - academics [70] art [8] changelog [49] current events [36] cute stuff [12] gaming [11] music [8] outings [16] philosophy [10] poetry [4] programming [15] rants [5] reviews [8] sport [37] travel [19] work [3] miscellaneous [75] |
- category tags - academics art changelog current events cute stuff gaming miscellaneous music outings philosophy poetry programming rants reviews sport travel work tags in total: 386 |
![]() | ||
|
is waiting for the end of the world to come." - Murakami, 1Q84, appropriated by The Big Short Caught The Big Short at Jem on Friday, having already reviewed the book here back in November 2013, partly as I was curious as to how it would translate to the big screen. I mean, for The Wolf of Wall Street, maybe half of it was hijinks, but third-degree derivative swaps? How is anyone going to make a mainstream movie out of that? Oh. Well, the directors figured, we can just call in Margot Robbie, after she did so well in Wolf. And put her in a bubble bath to explain economics, why not? Yeah, that'll do. Stylistically, The Big Short does have quite a bit in common with Wolf, with Gosling (as Jared Vennett) liberally breaking the fourth wall. Content-wise, there's not all that much that the movie has over the book, save that the necessity of clearing up some financial mumbo-jumbo resulted in cameos by celebs such as Robbie, Anthony Bourdain (bad fish as good!), Selena Gomez and Richard Thaler (hot hand fallacy, derivative bets). Oh, there's the usual no-substantial-female-role moanings, which seem to make for a free article on each new release. Story-wise, it's all quite transparent. The U.S. housing industry was booming in the early through mid-2000s, as builders, lenders and regulators cavorted in the same bed (at times literally, it is hinted), granting loans to individuals that had no realistic hope of repayment. However, none of the involved had any incentive to get their hands out of the punch bowl:
"You smell it?" Well, as entailed by the Four Axioms Of Guaranteed Market Crises raised in my original 2013 review, most everybody appears to have forgotten it all already, with Goldman Sachs now peddling BTOs - bespoke tranche opportunities - which are basically CDOs in new packaging (for once, our guys have Goldman beat - unfortunately, they're the HDB); however, the point, as the film tries hard to make, is that these excesses are only to be expected. Almost nothing has changed in that the rewards from financial hanky-panky remain huge, while the risks of getting caught, and the penalties if that happens, remain relatively tiny. As long as the expected returns are so large, why the heck would such practices go away?! Just one more observation - while our cast of anti-hero misfits (headlined by Batman acting as a semi-autistic doctor-turned-hedge fund manager) does successfully get one over the evil banks at the end, it is not easily appreciated as to how nearly all of them got screwed over. Note that part of the reason why the subprime bubble expanded so incredibly was that there was no mechanism to bet against it - you were either in, perhaps against your better judgment, or you were out. To be sure, shorting is often seen as distasteful, because it is at its heart betting that someone else will fail. However, like broccoli, the ability to do so is arguably healthy, if you consider its impact as helping to direct capital to more productive uses. As Cornwall Capital (Brownfield in the movie) did with cheap options, the motley collection of shorters were simply acting on perceived mispricings - in this case, in the entire American mortgage industry. Thing is, not only is it hard enough to time the market right, as more than one legendary speculator has found, the shorters were playing against the house. Since there were no readily-available instruments to short CDOs, the banks had to create them, and they then acted as the counterparty, because banks don't like to pass up free money - or so they thought. And they almost got away with it. The banks, having created the product, also retained the right to price it, and quite shamelessly maintained that all was okay as the markets were tanking. It was the equivalent of, having bought fire insurance and then driving the adjusters over to watch as the house is burning down, being calmly informed that a solid definition of "fire" is needed. Actual casinos are rather more scrupulous about their games being fair, if negative-expectation. As it was, most of the banks didn't even lose that much, by the simple expedient of buying shorts themselves (i.e. foisting the losing end on some other unwitting sucker). It's almost a mystery how they ever lose, with the deck stacked so heavily! 宴无好宴, 还是不去算了 - President of China, Xi Dada "Pssht, I could stand in the middle of Fifth Avenue and shoot somebody, and I wouldn't lose any voters, OK?" - next President of the USA, Donald "The J" Trump ![]() 唉, 早知道先跑人了! (Source: scmp.com) [N.B. Under new ownership, to "promote wider views of China's rise as a global economic power", with "continued editorial independence", *ahem ahem*] The movie has inspired many wannabe Michael Burries to identify the next big market dislocation, with U.S. student loans, charter schools, auto loans and even Big Tech coming under scrutiny (Burry himself is gunning for water). However, the most conspicuous bubble candidate is so glaring that it is largely left unspoken, all the more as popping this particular one has all the makings of a murder-suicide. From the "bad debt that is plainly never gonna be repaid" angle, it's likely that China's ongoing mess dwarfs America's circa 2006. Of course, it doesn't make for a direct parallel, as the arrest of the luckless director of China's National Bureau of Statistics demonstrates; while America's top titans of finance probably outrank almost all of their more-or-less elected politicians in practice, there's no question who wields the axe in China, and they're not squeamish about chopping heads. Okay, minor correction: debts do end up repaid - or more precisely, settled - one way or the other. The question is who ends up doing the repaying. In the case of America's subprime crisis and eventual bailout, it was ultimately the taxpayer, who railed about it for a bit before tuning back to Real Housewives of Atlanta. This was helped by the loss being relatively "invisible" - many of the affected had not put in that much, and walking away from underwater deals was always an option. Problem one for China - their real estate is more often than not paid upfront (at least for commoners), which makes any losses, well, real. Pettis, who is the best authority I know of on the subject, tackled this issue in a 2014 editorial, and outlined the three major sectors to which bad debt could be assigned, and the projected outcomes:
![]() By the way, we make concrete shoes, just sayin' (Source: imgur.com) Of course, it's extremely unlikely that we'll get the Chinese edition of The Big Short; Soros may just be big enough to stare down the PBOC - and by extension, the CCP - for a bit, but consider that the worst Jamie Shipley faced was being laughed out of the room (okay, and going bankrupt). In China, one could disappear like a Hong Kong bookseller. Increasingly many respectable analysts are now mumbling about 4% GDP growth, as we've suggested for ages, but one suspects that it's not about the figures any longer - if ever it was. Locally, The Party has guaranteed a minimum of 12 Opposition MPs by NCMP top-up, in a move calculated to dampen calls for more alternative representation, while maintaining a stranglehold on all voting actions. Personally, I'd rather they just dispense with all this rule-lawyering and simply return to single seats with a neutral bureaucracy, but fat hope of that. They were at least kinda reasonable over Lee Li Lian's vacated NCMP seat, so there's that. Trumpmania is enveloping the United States in the meantime, with The Master skipping out of a Fox News debate, instead hosting his own fundraiser for veterans that was aired by CNN and MSNBC. If there was any doubt as to whether this was yet another masterstroke, the Fox News host made herself the star of the show as the other also-rans ganged up on his closest competitor Ted Cruz. It was simply beautiful. That guy, he knows which banquets to skip. Programmer's Lament It's not just Trump who's not turning up; as the tech sector may be heating up - ever so slightly - once again, articles have sprung up in the local press bemoaning a lack of local programmers, especially with those in the line aspiring to quickly climb the ladder to "project manager". Definitely, the IDA - sorry, IMDA - can go around handing out pom-poms as always, but the bare fact remains that coders, and engineers in general, are paid pretty crappily here, for filling the same job scope as elsewhere. Apologists may point out the higher tax rates in the USA and UK etc, but all considered, a techie that's good enough to get hired in the West can probably still take home and keep double or more what he can get locally, even without getting into the other career benefits such as exposure and market access, which makes the choice nearly a no-brainer for most such talents. The engineering vs. finance salary gripe has been lingering here for a long time, with a major complaint being that those who actually do the shit, don't get paid, in contrast to those who simply talk a good game. Then, given that both fields can require brutal hours, but with the environment, remuneration and potential upside of finance so much better, it is hardly surprising as to why our top minds were hardly champing at the bit to break into the technology and engineering fields. That said, one might guess that plenty of SMEs are merely after somebody to pipe their Excel macro outputs to Word documents, and maybe clear the company's Hotmail backlog once in a while, and there's nothing wrong with that. However, software development is an area in which more is definitely not automatically better, and where the 10x value aspect holds; not only can projects become unwieldy as the number of developers increases - changes in one part of the code may duplicate or even silently break other regions, etc, before going into all the coordination headaches - one can further argue that much of good developing (as with trading) is all the hard-learnt domain-specific tricks that add up, as well as knowing when not to do anything... which can be tricky if you've got a dozen guys worrying that they're not justifying themselves, or worse - who are getting paid by the line, as if they were production workers seated at an assembly belt. Sure, programming for one may seem a highly outsourceable service, but there's still a reason why talent congregates in the Valley; The State's Times has highlighted Google being on a hiring drive for software engineers here, while not even bothering to try to figure out how many positions would be available. Tech In Asia has flagged it as two, which says it all really. Heck, ISIS is offering US$10000 for good hackers, who are in short supply here, from the manual brute-force hack of nearly 300 SingPass accounts that didn't reflect particularly well on either the offender or the SingPass sysadmins. You've gotta salute the tenacity, if nothing, though. Well, when it comes down to it, all I've got to end with is that there's certainly no lack of tech and engineering staff... but also that there's nothing that says sincerity like a fair wage, rockstar titles and other feel-good platitudes my ass. Gogogo! Speaking of Google, they have just released a Go engine, AlphaGo, that has thrashed a top human player, right on time to rub it into Facebook's face. Ignoring for a moment just why a search engine and a social network - sorry, two advertising firms - are battling over an ancient Chinese board game, it remains something of a breakthrough for the deep learning paradigm, which if I do say so myself, is getting quite a bit of attention these days. A teeny bit of background here. Computers, as you might suspect, are excellent at games of calculation. In chess, for example, your humble desktop PC is practically no longer beatable by humans. The number of options in Go is, however, on another level - whereas there might be at most a couple dozen possible moves each turn for chess, there are usually hundreds for Go. Worse, there is often no easy way to evaluate who's winning, while for chess, one can state that losing a queen or rook is generally a very bad idea. As such, past Go engines have been notoriously weak, being little better than amateurs, and no match at all for the pros. Enter convolutional neural networks and deep learning. It is oft stated that Go admits on the order of 10170 legal board positions (you know the number of atoms in the entire universe? Well, imagine that each of those atoms is a universe in itself), which seems to make it obvious that a brute-force search won't get one very far. Instead, one has to generalize patterns... which, as it turns out, are one thing that convolutional neural networks are unreasonably good at. Indeed, while AlphaGo employs an undoubtedly well-designed Monte Carlo tree search, the training of the actual low-level value networks appears extremely straightforward - beginning with a database of 160000 games between top human experts (data is king), totalling nearly 30 million board positions, the engine was first trained to predict expert moves, and then turned on itself to learn improved strategies, roughly said. Poignantly, this came in the week of Minsky's passing, but while this may certainly be ahead of the expected timeline, I wouldn't be joining the "human intelligence is within reach" hype train as yet. Although the topic deserves a far more detailed look, there are at least two tantalizing questions: First, how competitive would this approach be on chess, particularly as the method appears easily portable? Second, is much of the "received wisdom" in Go, such as opening on or around the 4-4 and centre points, actually sound? Linky Linky
Kinslayer Still in a fitness centre far, far away... P.T. Instructor: ...twelve laps before rolling over isn't too horrible on your first session, I suppose. Tell you what, Mr. Ham, let's get our massage therapist onto you. [On reddit] Next: Monkey Business
|
![]() |
||||||
![]() Copyright © 2006-2025 GLYS. All Rights Reserved. |