Bruce Campbell. No matter what happens, he's seen it all before.

I own a mineral deposit in a central Nevada mining district, though not with any intent to exploit it. I am quite familiar with the regulatory details of mining in the US. It is very different than the caricatures spoon-fed to the public by activist organizations.

Environmental impact studies are fine and necessary. Archaeological impact studies are mostly bullshit; the region is littered from end-to-end with artifacts leftover from the Lake Lahontan civilization (http://en.wikipedia.org/wiki/Lake_Lahontan), you can find stuff everywhere if you know what to look for. So everyone just pretends that there are no artifacts.

There are two big problems that really make it impossible to profitably mine US deposits. First, there is an environmental lawsuit industry that thrives on delaying the opening of mines until the companies run out of money to deal with them. The lawsuits are mostly bullshit about hypothetical habitats for endangered species and the like; they aren't credible, but that isn't the point and some courts are willing to entertain them indefinitely.

Second, a big problem is that if you pick up a rock, you own it. In the western US mining districts, those rocks are laden with natural concentrations of all sorts of low-value heavy minerals that are magically transformed into "toxic waste" the minute you touch it. This has arguably been the biggest killer of new mining. The obligation to scrub natural mineral formations of elements with no economical value very substantially increases the cost because you end up "mining" metals that have no value. This is particularly problematic for things like rare earth metals -- the mineral complexes are intrinsically "toxic waste" under standard regulatory regimes. It doesn't matter that they are natural, the mining company is obligated to treat nature as a superfund site.

Regulations regarding arsenic in the water have been similarly exploited by environmental activist groups to shut down mining. In many places in the western US, the background levels of arsenic in the groundwater is naturally several times higher than the EPA limits because of the local mineral formations. The way it works now is that if you do mining near those formations, you become responsible for bringing the natural background levels within EPA guidelines -- a fool's errand. So mining companies avoid areas where the local arsenic levels exceed EPA guidelines, lest they become responsible for cleaning up arsenic they didn't produce.

Environmental activists have very cleverly created a regulatory framework that holds mining companies responsible for natural mineral distributions even if the mining companies are in no way responsible. This has effectively outlawed heavy metal mining in the western US because the environment is naturally full of heavy minerals.

You are misunderstanding the problem. The mining companies would *love* to develop the rich mineral deposits in the western US -- all mining is a long-term investment -- but it is politically impractical. Not only are there many years of regulatory overhead before you can even get permission to start (archaeological clearances, environmental impact studies, etc), you also find yourself plagued by routine lawsuits by environmental activist organizations. In short, you can waste decades trying to develop a new mineral deposit with nothing to show for it but a lot of well-paid lawyers. There are difficult regulatory problems even exploiting existing rare earth mines.

It is cheaper to explore and develop countries like Australia and Chile, both of which have mineral deposits similar to the western US, than it is to develop existing US resources that we already know exist. This is not the fault of the mining companies. Indeed, the free market is working precisely as it should when one supply is priced far beyond what is reasonable due to political intervention.

The vast mineral districts in the western US are very expensive and risky to develop even though there are large, high-value mineral deposits available there. While old mining sites are sort of grandfathered in, developing a new mining site is prohibitively expensive for regulatory reasons such that the value of the resource has to be atypically high to offset the regulatory overhead. In short, opening a new mine in the western deserts has become kind of like trying to build a new nuclear power plant. There are so many lawsuits and interminable amounts of politics and paperwork that it has become effectively impossible even if it is theoretically economical.

Instead of developing US mineral wealth, most mining companies are developing mineral resources in countries with less regulatory overhead and fewer environmental lawsuits. It is not that the mining companies do not want to develop US mineral wealth, it is that the US government has made it all but impossible to do so as a practical matter.

When writing C++ I usually test two implementations of an algorithm/data structure, the STL version and my own purpose-built implementation (usually not that difficult after all these years). About 50% of the time, my implementation destroys the STL implementation on performance and the rest of the time they are roughly similar. In the former case I use my implementation and in the latter case I use the STL.

The problem is that no library has enough knobs to fully parameterize likely use cases, they have to make assumptions. In many cases, those assumptions will be poor. For some applications it doesn't make that much difference, but for many of the applications you might use C++ on these days (e.g. server engines) it can make a huge difference. Over time, I have accumulated a large standard library of source that I can reliably adapt to a variety of use cases. It is not a library, it really is more like cut-n-paste, but it allows a lot more flexibility in dealing with design issues than simply using the standard libraries. Of course, this is predicated on actually understanding the nuances of algorithm implementation, something that seems to be an increasingly rarified art.

The performance lost in generic libraries is not trivial.

Part of the problem in Silicon Valley is that the venture capital community has become noticeably more risk averse than it was many years ago. Many (most?) firms act more like investment banks than high-risk, high-tech venture funds.

Additionally, I think the rise of social media has biased venture capital deals in strange ways, steering even more money toward social network and media whores than actual tech ventures.

On one side you have a bunch of people who, like the old tobacco company, will swear blind that the current system is perfectly okay despite it killing an estimated 45,000 people a year.

We need health care reform, but let's use legitimate arguments instead of specious talking points to make the case. Otherwise, you are just supporting the people against health care reform by damaging the credibility of the argument for it.

You need to put that 45,000 figure into context. If, for example, the US had the UK healthcare system, there would be 300,000 more cancer deaths in the US every year as a simple matter of statistics across the respective populations.

Even if 45,000 was an accurate number it is pretty insignificant statistically. The US loses that many people to auto accidents ever year and no one cares. The differences in survival rates across a large number of common medical conditions between Europe and the US may be small as a percentage, but when scaled to a third of a billion people it can become a difference that dwarfs the number of deaths nominally caused by the current system. Hell, plain old medical incompetence kills far more than 45,000 Americans every year.

In short, that 45,000 is supposed to sound big but it falls in the statistical cracks when we are talking about death rates across various medical outcomes and fatal injuries. If the medical outcomes elsewhere were statistically better or close enough that 45,000 was a big difference it would be a more compelling argument, but none of the literature in the medical journals supports that. Consequently, the 45,000 is pretty much a non-argument for changing anything. You'd save far more lives clamping down on incompetence and malpractice.

Infrastructure is primarily paid for by the individual states, not the federal government. What the US federal government does or does not spend is largely irrelevant; it is not as though the individual states have 'defense' as major line items.

Why a US government agency needs an "investment arm?"

Investment vehicles like In-Q-Tel are not redundant with conventional venture capital and were created to fill some clear funding gaps in the existing technology venture markets.

First and foremost, they tend to invest in ventures with technologies that are sufficiently advanced or unusual that normal VCs will promptly ignore the venture. This came out of a realization that really advanced computer science and hardware technologies that the agencies needed were being routinely ignored in the traditional venture markets because VCs don't understand technologies that don't play buzzword bingo or which don't follow the herd. Investment vehicles like In-Q-Tel have a much stronger long-term technology vision and in-depth technical competency than traditional VC firms, which can be beneficial if you are building a startup based on serious geekery.

Second, they provide an inside track into organizations that would otherwise be very difficult for a startup with no ties to the defense industry to sell into. There are big customers of advanced computer and software technologies in the defense organizations, but getting a product in front of the right people is no easy task if you are an outsider.

Sorry, but CS is not engineering by any stretch of the imagination.

Algorithm design, the part of CS that is subject to patents, most certainly is engineering. There is a single mathematical concept of "sorting", but an unlimited number of computer algorithm designs that accomplish "sorting" under a wide range of real-world constraints. This is an engineering process.

If computer algorithm development was mathematics, there would only be one sorting algorithm for all applications.

A stone flying through the air can be described by the math for a parabola. Language that symbolically represents that mathematical relationship is math, whether it's machine-readable or not.

Per the basic theorems of algorithmic information theory, the "stone flying through the air" is math and can be literally instantiated on any Turing machine. It is a finite algorithm and can be manipulated as such; most of the theoretical interest in high-order algorithmic induction revolves around this very topic. Physical machine designs are literally high-order algorithms right down to a particular implementation of the abstract design, and increasingly such machine designs are being directly manipulated in their high-order algorithmic abstractions.

In short, all patents are discrete high-order algorithms relative to the algorithms they describe. The move to higher-order abstractions that do not directly reduce to physical machinery has been going on in physical engineering for decades, though it has happened first in chemical processes and computers because the engineering math is easier in those fields. This is every bit "math" as computer algorithms are, with the advantage that these algorithms are vastly more tractable for real-world application problems than enumerating your naive concept of patentable material. You are essentially suggesting that only inefficient representations should be patentable.

This is not a defense of patents per se, just a repudiation of the fallacious idea that computer algorithms are meaningfully distinct from any other patentable subject matter. Just because most software geeks are ignorant of the ramifications of algorithmic information theory does not mean that we should perpetuate that ignorance by pretending it has legal standing.

It is using mathematics to derive algorithms that solve our problems.

This describes all engineering disciplines, it is hardly unique to CS. System and process models are (often quite literally) high-order algorithms, but they nonetheless remain patentable in conventional engineering disciplines. This is some of the most valuable intellectual property in engineering.

Any argument loose enough to classify algorithms as mathematics is necessarily loose enough to classify *all* patentable subject matter as "mathematics". I'll see your Howard-Curry isomorphism and raise you algorithmic information theory.

The Howard-Curry argument is essentially that anything that can be described on a computer is "math". Unfortunately, there is no patentable subject matter that does not have this property.

Even ignoring that, the part that is disingenuous about the Howard-Curry argument is that it also is directly applicable to electronic circuit design and chemical process patents in the same way it is applicable to a computer algorithms. I would find the argument less shady if it was not applied selectively by opponents of algorithm patents.

"Let's take a hypothetical this way, then. What would you imagine would happen if every software patent out there were rigorously enforced, without exception? Do you really believe things wouldn't slow to a grinding halt if this were to be the case?"

The problem is that "software patent" includes multiple unrelated concepts. What are you actually enforcing or getting rid of? Eliminating business method patents are no great loss, and would have minimal impact. Eliminating computer algorithm patents would necessarily invalidate many other classes of patent that most people seem to fine with since there is a literal equivalency in practice. Most legitimate computer algorithm patents that are in force today are not likely to cause much damage -- these are the only ones that are enforced as it is, patent trolls notwithstanding.

All patentable areas contain loads of frivolous patents that are either unambiguously obvious or for which there is prior art. That the patent office in the US and abroad puts so little effort in filtering out these types of patents is a big part of the problem.

Business method patents are different from computer algorithm patents in that the latter always has a strict machine specification and the former does not. The only way to make business methods have a strict machine specification is to severely reduce the scope to the point where the patent would have no real value in the sense that everyone's business method would have a different specification and therefore be their own unique inventions. Can you define the "shopping" in "one-click shopping" in terms of logical bit transformations? Computer algorithms can always be trivially defined in such ways.

Business method patents as they exist today were achieved by effectively eliminating the requirement to have scope that is strictly bounded by the machine specification. Computer algorithm patents have scope that is strictly defined by a machine specification. Business methods have much more ambiguous scope because there is no specific machine required for the patent.