I'm sure that there are applications that would wish to put February 31 into a date field in a database, so why restrict them?

http://en.wikipedia.org/wiki/February_31

I'm not a geneticist (IANAG?), but while reading the article, I thought that they just discovered that rare variations found today are mostly 5000 years old or less. That doesn't mean that humans evolve faster, it might also mean that rare variations simply don't survive more than about 5000 years (perhaps they get disappear, or somehow get masked, or simply cease to be rare)...

For example, when you put a drop of colored paint in a bucket of white paint, the colored paint is initially rare. But if you wait 5000 years then look at the bucket, it will be a bucked of off-white color without a 'rare' drop of paint...

"C on evaluating the likely accuracy of the data..."

Don't forget that currently available barometric data from other sources is quite sparse in most locations.

You have to start somewhere. Apps can be updated with improvements...

A lot of the data issues you mention occur more in urban environments than they do in suburban environments or out in the country, and the currently available barometric data from other sources is probably a lot more sparse outside of urban environments than it is in areas with subway systems and tall buildings (people do have smartphones with enough data coverage for sensor networks well outside of large cities).

The app can certainly also look at other sensors such as the acceleration, positioning, and motion sensors to help with data issues. The 'network' can find bias or accuracy issues from nearby measurements of other devices. The 'In a car with open windows' situation is easily detectable by the high rate of variation in pressure even for a short measurement duration (and even so, a long term average might still yield useful information, bias 'when in motion in the car' can be measured when the vehicle stops (using the motion sensors)). The app could even learn, f.e. "When near this wifi, report the measurement with a potentially previously learned pressure bias", or "When near this bluetooth, then it's in a car". With the light sensor and positioning/acceleration sensors, the app can even be quite aware of various state of the phone (sitting motionless on a flat surface, or being carried inside of something, or being carried open, or being held up to an ear, etc). I'm pretty sure it is possible to make an app that collects a tremendous amount of useful data for weather prediction that would otherwise be unavailable (actually, temporal variations from motionless devices in a location with unknown bias can be very useful raw data, and the strengths and positions of the windgusts around buildings may even be useful raw data).

Also don't forget the power of large numbers, you don't have to use all the data: Outliers or inaccurate measurements/devices can be detected as such when surrounded by enough other measurements, and snr improves with filtering of multiple measurements (also from multiple devices).

"Scientific method is rather simple"

It is in the field of math, some aspects of physics, and in theory, but in the real world filled with people (scientists) working in real world organizations, regarding the field of medicine, the public 'outcome' or 'status' of scientific method is strongly influenced by politics and economics.

In addition to that, human health status is such a complicated system to measure that pretty much every medical field study uses very sparse data points. It should be no surprise that in the field of medicine discoveries are made regularly that certain things have been previously misunderstood.

And as a result, not everybody follows the herd. And that is simply how the world, filled with human beings, works.

"it's not that people don't trust science, it's that they don't trust the motivations of the people practicing it."

Bingo, we have a winner.

"Keep in mind that any hardware you design will still have a fully programmable core in it,"

Say what?

That logic does not hold up if you have a new product that you intend to sell at full price and that is meant for the same market as the old product, because selling the old product at a discount cannibalizes the sales of the new product in the same market, causing it to take longer to 'recoup' the costs, etc, etc... You don't want to be in that spiral.

Umm... Where is your ground wire when you're in the air?

See subject.

That "radiation that a functioning nuclear power plant releases into the surrounding environment" is a tightly controlled quantity and should be as close to "none" as possible. That number does not include the radioactive waste that the plant generates, because the intention for the waste is to not release it into the environment. Sometimes, however a nuclear power plant does release more than the normal amount of radiation, and then usually it's bad enough to be referred to as a 'disaster', with greatly increased releases of radiation into the surrounding environment (chernobyl, fukushima daiichi). How often has an area been declared a nuclear disaster area from an "event" caused by a nuclear plant?

So... venting the radioactive waste into the air would, well, poison the planet faster than you can say slardibartfast.

Maybe it's still cheaper to build that bioreactor.by growing a cow? You don't slaughter the cow to get the milk...

"It is entirely conceivable that in the next decade we start 3D-printing buildings and electronics."

Wheee, and let's 3D-print a planet in the decade after that!

3D-printing works because they use plastics that melt when heated. They are cool, but they are basically robotic glue guns with fast drying glue.

How would you 3D-print something that needs to withstand heat?

Or how would you go about 3D-printing reinforced concrete, or an economic and equally strong replacement?

You can't extrapolate a glue gun to that many materials/properties.

Obligatory xkcd reference

http://xkcd.com/327/

If your platform has stdint, then you can use ptrdiff_t instead of int/long in your code... It's always the right size integer for pointers.

5) turn that nice phone into something pretty much useless

Or in other words: You're mordac.