tfm

To clarify, seizing of "inactive" funds already occurs -- the change next week shortens the timeframe from 7 years inactivity to 3 years. The money can be subsequently reclaimed through tedious paperwork. This is pretty standard stuff internationally speaking (U.S. seems to have a timeframe of 3-5 years depending on the state, U.K. is 15 years, Canada has a more generous timeframe of 30 years, or 100 years for amounts over a grand!).

The responsible thing for your bank to do would be to notify you at the 35-month period. Not much good if you don't keep your contact details up to date, but if it's something like a investment or trust account then that might be the one address you'd like to try to keep accurate. THIS DOES NOT CONSTITUTE SOUND FINANCIAL ADVICE, PLEASE FEEL FREE TO KEEP YOUR BANK STATEMENTS FORWARDING TO THAT EXPIRED P.O. BOX AT YOUR ALMA MATER

[OT] Woah, takes me back to the bad old days of a.r.k https://groups.google.com/group/alt.religion.kibology/search... The nounal usage "Legos" just wasn't one I heard until I was exposed to this Internet. Correct en_AU usage is to deploy "Lego" as its own plural. Dialects!

Back on-topic, I'm kicking myself that I didn't think of using Lego when I was tutoring TCP/IP, but ... I was also kinda hoping this article would show some sort of robot-based assembly line with motors and (of course) a bit bucket for lost packets. To the Legotorium!

That's beautiful! Just needs a couple more references to Invisible Russians and there's a Sveriges Riksbank prize with speakeasy's name on it.

Well, we can take as a lower bound for information/energy equivalence as the k.T.ln(2) from Landauer's principle[1], and mass-energy equivalence as e = m.c^2; there would presumably be some inefficiency in converting energy to fuck-energy (call the efficiency epsilon), so that gives us

  information =  epsilon . mass . c^2 / (k . T . ln 2)
              =~ epsilon * 3.13e43 bits

or about 3 billion billion yobibytes assuming peak fucking-efficiency. Bear in mind that this is assuming the advice was delivered at room temperature of 300K, which is unlikely given how hot Instagram startups are at the moment, and of course no-one should work in a vacuum, so this is just an upper bound.

[1] http://en.wikipedia.org/wiki/Landauers_principle

"Reality has a well-known liberal bias"

I guess that implicit in the article is the assumption that this self-selection effect hasn't changed much in the last 4-8 years. The author is careful to stick with like-to-like comparisons.

If on the other hand there is a trend away from Google among certain groups then that would seem to be a problem for Google's branding team! Perhaps there is a market here for a left-wary search engine, say DDG hooked up to conservapedia instead of wikipedia at the backend?

It is good to hear this, I had my doubts but no experience anywhere near MRI reading. Looking at those colourful tubes, it is not in the least bit obvious how much of it is pure modelling and how much is cleaned-up scanner output.

"New scanner gives 10x improvement in detail, shows amazing and unexpectedly simple result" is a pretty punchy-sounding headline, too. They should go with that next time!

The limited connectivity is very interesting -- my experience with artificial neural nets is pretty limited, but are there many well-studied ANN models that have a similar structure to this? The spatial restrictions are so simple that it seems like they'd lend themselves to efficient implementation.

To be fair, it's hard to know how much of that sluggishness is the app itself and how much is just a video-encoding artefact.

Certainly the API has some interesting inclusions -- Facebook and analytics built-in.

But then at the end he kept mis-speaking as to whether there were five Directors-General or six. Maybe they're at 5.9 sigma after all!

99designs did well from early investment from Atlassian, but their big recent VC dollaridoos came from ... Silicon Valley.

Continuing the Australia/99 theme, we might observe that 99dresses was almost thumped into oblivion (due partially to lack of investment) before it enjoyed the expertise of YC and the attentions of the YC coterie.

Given geographic inevitabilities and the lack of local investment it is hard to be optimistic about the Aussie startup scene, even given Fairfax media's recent (and beguiling) attempts to toot the local talent's horn.

I don't entirely trust these particular Google Trends figures as they show a minuscule amount of USA-based traffic.

Regardless, though, of the particular instance, or indeed the reality -- the problem for any would-be head-to-head of this type is not generally which startup would "win in a fair fight", but rather which startup any potential investors believe would win. The problem for such Australian startups may be just that the question of "which is going to win?" is being asked at all.

> different way than promised

Once the originally-planned $20k had been exceeded, Inman suggested that he would find some more charities that might benefit from the cash. This thought was abandoned shortly after Carreon cried foul about the "different way than promised". So, good-good for the bears, they'll get a bigger money pit to roll around in.

Hey, jinx! When I heard "US based company", I thought "Telstra". Optus would have saved a few million going with the cheaper crappier solution from some other bargain offshore location.

I'm not happy that the user data munging being outsourced but it is pretty likely that the same activity is rife at the other telcos (or will soon be, once their copy-happy product managers get wind of the feature).

I was under the impression that CUDA is pretty much an nvidia-specific proposition. This is a great shame because it seems there is a bit more activation energy required for opencl projects (AMD's BLAS is great but I don't see it being used anywhere near as widely as Cublas, and there was no Mac support last I checked).

Are there options for getting CUDA going on an AMD/ATI chipset?

It's a genuinely interesting read, even if it does lean on the Astonishing Particular Items Of Proof trope more than a little heavily, but the moment it got to the words "in my opinion, the answer is ..." my eyes glazeth over and skim mode began.

Once an author has weighed in with their personal interpretation, the only refutations available will be ad hominem. "I'm not calling you a liar, but I can't think of a way to end this sentence."

> And, of course, emailing or establishing contact first is > pretty much never a bad thing. Well, as long as one follows > through with what they say they're going to do.

There's the implicit engineering problem right there! You've always do a bit of implementation/thinkening before even going that far. Don't want to be the Flakey Sir Promise-A-Lot, do you? But of course it is sometimes hard to draw the line if it's a reasonably interesting problem ... and that's where hearts get broken and egos get bruised.

I thought I was the only one! (sob) I've tried folding, so many times, but it just seems like I'm fighting against it. I don't know that it's the shape of the code that's being hidden that spoils it, maybe I'm just a bottom-up kinda guy.

Maybe there is a need for a support group here.

Fine line to walk, isn't it? Taking ownership of your code would seem to be pretty closely related to taking pride in your work, and the motivations that entails. Having worked with and for $BIG_COMPANIES I would have to agree with your observation that the majority of developers are not, uh, enjoying a personal association with the fruits of their labour, and it shows.

Maybe own it till you're proud of it, then let it go! Schedules don't always allow every codule to be the shining star in your portfolio, but there should maybe be pride to be had in a really dirty hack done on an insane deadline ^_^

Aw dang, I missed Pi Day again! Stupid timezone. I guess tau will be getting my support this year ...

I like this analysis better! For bonus points, why would this generation be any more risk averse than previous generations?

Managing risk is a learned behaviour, and we've kinda got our recession on... may be they were paying attention. Heck, I couldn't be prouder of these cottonwool wrapped kids. Let's give em a low carb imitation lollypop.

North Dakota may have problems housing the itinerant Nevadans once they arrive, too. http://www.nytimes.com/2010/04/21/us/21ndakota.html

It seems likely that people will move / have moved interstate to take advantage of the job availabilities in ND, but it's not clear from the article why we'd expect unskilled youth to take on this mild inconvenience, as opposed to any other demographic group.

Census data does indicate that interstate migration rates are falling generally (albeit gradually), but isn't that to be expected? I hear this new fangled "Internet" means that the telecommuting cybernautic generation can work from their oxygen-controlled humidicribs, thus saving the burden of investment in loathsome, expensive transport infrastructure.

> Having only human generated content in Google+ adds a certain quality to it.

Very good point. There is a slightly weird aspect though to the divide between the personal posts being poured into G+, and the purely algorithmic results being pumped out the other end by Google Plus Your World et al.

Certainly, a "cohesive strategy" would be to give a more human feel to the search results (or, easier -- Kill All Humans) but in the meantime it certainly doesn't feel like different aspects of a single magical piece of huggy software. Perhaps five or ten years down the track it may seem more organic, meanwhile it has this disjointed feeling which may be contributing to the lack of love for the Googles at the moment. I pour my heart into you, robot. Why do you give me nothing but cold metal?

Sure would love for an API though so that enterprising developers could hack up some interfaces other than the one-size-feels-itchy-on-all.

An interesting read is an interesting read, there are occasionally nuggets of gold even on the most poorly designed slideshow ad trap. I like to imagine that the "inverted pyramid" technique frontloads the best stuff on the landing page and this discussion page (a man can dream, can't he?).

Etiquette might suggest a heads up in the submitted title (paywall, slideshow, obnoxious popover ad that breaks the article on mobile, &c). If the article is in fact paywalled then it's unlikely to rise far enough to get in many peoples' way.

sed deserves a lot of love, I still prefer it for a lot of command-line munging, greps everywhere, bit of awk thrown in. That despite the fact that (say) perl -pe might do a more comprehensive or efficient job as basically a drop-in replacement with benefits.

Does/should a UNIX course nowadays bother training up students on these microprogramming languages if they're almost certainly going to have perl or python available anyway? They only marginally fit into the UNIX philosophy, so is it just their old-school cred that keeps them around?

I find this a really interesting area.

As a programmer rather than a gamer I'd tended to have this background awareness of GPUs getting more and more powerful, and saw all the Amazing GPU Stunts for {SETI,Folding,..}@home or Crack Your Neighbours' WiFi Using Just A Perlin Render-Renderer, and had this vague idea that the graphics card was this generic powerhouse that just needed a little love to get going. The C-like syntax of your typical GL shader language is particularly encouraging in this regard (not to suggest for a moment that Looking Like C is a generally positive attribute).

Recently I have been giving it a little further inspection, and of course it is not quite so cut and dry. Given the Single Instruction Multiple Data (SIMD) nature, you are somewhere pretty good, roughly between a programmable DSP and Full Gory Petaflop Gigacluster.

BUT. You get:

- No recursion, but maybe we can just be careful and collapse everything into loops.

- No memory management (yay, maybe), but that makes sense since we're trying to execute everything in parallel. Perhaps we can squeeze our dynamic programming into a local 4x4 matrix?

- Expensive branching, especially particularly very expensive if it's data-dependent branching which goes different ways in different kernels. This definitely makes sense if you think of the GPU as a hungry hoarde of soldiers marching along in lockstep, slaying the data, then somebody has to stop to handle an edge case and everyone else waits around ... I'd kinda envisaged it more as an N-microthread model with a bunch of workers churning through vertices and fragments as they become available. Probably it is time to take a break from web services until I get over this, or develop an SOA GPU In The Cloud (GPUaaS?).

- &c. &c.

So, obviously there are some tasks which are well suited for this, and idioms which are well supported. Other tasks and other habits don't fare very well at all. Maybe an amazing recursive raytracer which was optimised for C is not going to survive a transition to GLSL without some very clever trickery, but perhaps there are subtasks which can be handed off instead.

The very interesting work being done on WebGL (Webkit, Firefox, Opera thankyouverymuch) seems to have the potential to bubble the GPU to the surface where it can be used and abused with a very small activation energy. I think it would be nice to have some shader skills in the toolkit just in case the right problem comes along.

TL;DR: GPU programming is a lot more like learning a new programming language than I thought.