Oh, my bad, that was the other side's policy. I got confused there for a second.

thanks, a lot of truth in that... (Doesn't apply to video games though, does it..)

http://theoatmeal.com/comics/e...

The majority of discoveries leading to drugs sold by Pfizer were made in American labs, by Pfizer scientists. The acquisition history is varied and contains Irish, German and Canadian companies, but that's, all in all, a small portion.

If the distinction "is purely semantic", then it is actually meaningful. That's not what you're trying to say, is it?

I believe they're doing that already (A-GPS) and that's used by cellphones (sometimes triangulation of towers works crudely when GPS doesn't). But that's an augmentation system not unlike WAAS, and what the anonymous expert above explained is very different (A-GPS can't deal with signal reflections off objects, and that's logical).

Super interesting. Thank you.

Such a demo makes sense because you compare the performance against a control (without the software fix). Real-life improvement and absolute performance are simply a different, farther-reaching question.

The other important thing to note here is how well this could perform in combination with the wide-area augmentation system (WAAS), which corrects for atmospheric variation of signal travel time by using information the is generated by ground stations and broadcast by the satellites. WAAS isn't included in standard cheap GPS receivers, but can be had for little money, starting from $100 GPS receivers (and mandatory for certain aviation applications, for example).

Or the 2cm figure already include the use of WAAS?

Sorry, but I couldn't find interesting answers in that Reddit thread. It's mostly that people choose to run OSX over Linux, and why.

The few GNU/Linux users do it for idealistic reasons, or because they're developers, or because they like the latest OS on very old cheap hardware and don't mind to deal with whatever this entails.

The more interesting question is really if freedom exists when you never make use of it. (Do you actually hack the kernel or fix somebody's proprietary binary-only drivers as a GNU/Linux user?)

I've driven thousands of kilometers on the German Autobahn. The safety issue is not so much the speed relative to the ground, but the speed relative to other drivers. If you're going 170kph in the left lane in your BMW, and grandma in her 1990 Volkswagen swerves left to overtake a truck, you've got pretty limited distance to slow down (at high speed). German autobahns are sensibly limited to 130kph in urban areas, for windy stretches of road, or two-lane portions.

French interstates work quite well in that respect. Everyone does about 130kph (that's 81mph), and this speed limit is strictly enforced. Relaxed driving, really.

The other thing to take into account is driver training. The kind of tail-gating I'm seeing here in the states rarely happens in Germany. Yes, you get the angry BMW driver flashing his lights at you when you're in "his" lane (road rage is universal), but that's typically over in a few seconds. Drivers wisely keep their distance.

That was my view as well until learned a few things about this "continuous, analog computer". We know that it is neither analog (neurons can have threshold functions) nor continuous (some important, central processes are quantized - e.g., about 50ms per "decision" in a structure called "basal ganglia").

As for this paper, you seem to neglect that even the supposedly continuous, analog computer will have sub-processes that run in parallel, but are correlated and make a distinguishable contribution to the task the global system is concerned with. If you like to picture a network of neurons, then its structure with will one of many separate clusters (a "small world" network, for instance) rather than a random graph.

It's certainly interesting that the PCA-like analysis in the cited paper comes up with a similar number of subsystems, although I wonder if they ended up matching the Brodmann areas. And importantly, any set of areas is more like a subsystem, in which, if my quick look over the paper serves me well, activations make a unique contribution to task solving.

The question is, does this bring us closer to a computational understanding of how the overall processes work? Localization of function alone doesn't, IMHO. DTI (neuroscience) and cognitive modeling based on architectures (cognitive science) may make better progress.

http://www.skybert.net/emacs/j...

These distributions exist... (shameless plug: Aquamacs 3.1 was released today, based on Emacs 24.4). You're right in that its Java support is sub-par. Packages like JDEE exist, and CEDET has recently been included in Emacs, but you won't get out-of-the-box code completion, nice visual debugging, inline compile error messages and so on.

The PhD is not a degree designed to increase your chances on the market for jobs that do not require a PhD. I think that's the underlying misunderstanding. (I'm an assistant professor at a research university, I have a group of PhD students. Careerism, especially for undergrad degrees, is a common perspective-shift at US universities.)

A PhD really prepares you for a career in research/science, academia. You sound like you're looking to be a programmer (again). Did the reasons go away for which you chose to do a PhD?

Try the big players: Google, Apple, Intel, Microsoft, and so on. They hire PhDs quite happily, and you can apply what you've learned.