I think that I tried that. Going from (very fuzzy) memory there was a problem with the combo to switch. It was some like not being able to reliably press both keys at once with a single thumb so needing to use both fingers in the far left. Or something like that.

My main motivation for the phone was a physical qwerty keyboard. But I would say now that when the keys are that small there is no benefit over a touch screen keyboard. There is an advantage to a touchscreen keyboard though: mode switches to alternate sets of glyphs are easier to use.

Yes, exactly. For natural language it wasn't too bad, but for code it was painful. Being able to remap all of the punctuation (preferably with a mode-switch button on the touch screen) would have solved a lot of the issues.

No. I can't speak for the GP but I was definitely in the target demographic and the n900 was shit.
I've been a programmer for about 30 years, my normal environment is vim and a shell. The keyboard and screen size on the n900 were too compromised to make it an effective or productive environment. During the couple of years that I had it I took in on holiday once and spent two weeks on a beach trying to write code on it. After I got back I rewrote the two weeks of coding in about an hour. Trying to code on a hunt and peck thumb board that required escapes for most punctuation symbols was a waste of time.

Since then I've been very happy with an iPhone 4 as the touchscreen keyboard is equally effective / ineffective as the hardware keyboard. My next foray into a portable coding environment will be a glass competitor with better specs, a unix environment and some kind of twiddled interface although I am hoping the myo armband pans out.

Your logic is flawed.

By splitting object-recognition into two cases...
    A: Atomic shapes
    B: Collections of shapes ... you have not proven that algorithms cannot perform object-recognition. You still need to show that one of the cases cannot be performed by an algorithm. As case A is trivial, and case B is what this algorithm does your argument falls apart. Collections of shapes can be recognised by probability of occurrence. There is no need for interactivity, simply enough video to reduce the impact of outliers.

Because university means "a group of people acting as one body", while universe means "everything". The root word meaning "by one" was used in the later case to mean an entire revolution of time.

Better terminology for this theory would be "islands of causality". But scientists tend to be shit at naming things so instead they will overload a sadly overused term instead.

True (to the first part).

But they are not trying to predict the success of authors, they are trying to predict the success of works. Predicting the output of any author would be difficulty, modelling human creativity and all that jazz. But predicting the success of a work is simple(-ish) machine learning. Build a learning bias for style-features in the text and throw an optimisation at it.

For the second part - when do styles of literature experience sudden extreme changes in popularity? I've seen slow changes that peak suddenly due to shifts in demographics, but never sudden changes. Could you provide an example?

Taleb's point was that you can see patterns in past behavior that don't necessarily indicate future performance. He even used literary work as one of his first examples.

No, it really wasn't. His point was that the phenomena that we encounter are modeled by two very different types of distribution. In one kind the past is a good predictor of the future because deviations from the norm don't happen. In the other kind the past is a poor predictor of the future because although deviations from the norm don't happen regularly, when they do the impact of the deviation is immense.

Popularity contests are of the former kind.

Unlikely. Their comparison is the outcome of a popularity contest, which in the terminology that Taleb used is an inhabitant of Mediocrastan. The distribution is relatively smooth as it involves the average opinion of a large population.

I don't really want to know if he had a weak heart or was just quite authoritarian over his daughter...

And that somewhere else was probably also slashdot.

Mmmmmm, dirty non-linear plots. Who's a nasty little exponential function...

No we don't.

Primality testing is easy - the problem is in P. Approximate methods for finding primes are very efficient. Exact checking is rarely used.

Modern security protocols rely on the problem of factoring a number into primes being difficult. Or on inverting exponentiation within a prime field.

If you want to appeal to the floating voters then you need to acknowledge that it is only half a solution: present yourself as the reasonable middle ground candidate, willing to compromise on an engineered super-plague apocalypse.

That is something differnent, each document is a vector and each word is a dimension.

Dat cat was good to roll with, yo?