Poor machines scapegoated for the pervs that use them to peek.

Not for defrost...

... and it's nothing that can't be kickstarted from some minimally-supervised hands-on experience and a short while of book-crunching. It's much harder to train a person to think in the right way than to give him some `skillset' that's just a hill of facts that has an expiry date. The latter is more akin to learning about the functions of Microsoft Word than actually writing an essay. I agree a large portion of people who come out from CS curricula have no clue about either computers or science, but those who prove to be capable of what they're taught and what they've explored often have what it takes to learn what you throw at them.

I don't think counting photons is going to be unbiased either. Randomness is different from unbiasedness. There must be a distribution, with some more values more likely than others. Unbiased-ness is however not necessary. You can use an unbiased coin to simulate a biased coin, and vice versa, and any continuous distribution can be turned into a coin flip. If you want a more efficient `unbiased randomness translator', there's research on stuff called randomness extractors, which aim to generate as many approximately-unbiased random bits as possible from a random source.

Perhaps except Terrence Tao; a famous math prodigy, who also became an incredibly successful mathematician, "Such is Tao's reputation that mathematicians now compete to interest him in their problems, and he is becoming a kind of Mr Fix-it for frustrated researchers. "If you're stuck on a problem, then one way out is to interest Terence Tao," says Charles Fefferman [professor of mathematics at Princeton University].". Also Erik Demaine, who finished PhD and became a professor at MIT at 20; he has a less impressive history than Tao, but still a fruitful career.

With cash, many transactions can happen between getting from a checkpoint (banks) to another, and some bills simply don't go to banks so often. Sometimes a bill just used for payment is given out again as change. Sometimes a transaction can happen in private, such as in internet auctions. Sometimes there's a vendor that doesn't go through banks so much. Finally, shops simply don't take note of your own identity, so even when the police traces the money back to a shop owner, the shop owner won't be able to tell which bad guy gave him a $20 bill with a certain ID on what day.

In Bitcoin, every single transaction is traceable, it's just not so easy to identify the endpoints. Up till now there have been many ways to avoid being identified, but as more regulations and infrastructure are brought in, eventually it's going to be less anonymous than cash.

Unless you're being sarcastic, no. Hell no. When any power plant is built there's absolutely no revenue, and absolutely no KHW, so to you it's a loss, and we won't be building any power plants. This is classic beancounter-level thinking, and I hope you are no where near running any business or voting for budget matters.

The point of having a power plant is to provide electricity, which, even if all you care about is money, supports economy output. The overall cost is the cost of the electricity it produces, and it is a justification of building this plant over another plant, among weighing other pros and cons.

If it works out that we're getting cheap (and clean / safe etc.) electricity after all, it's good to have built this power plant. Perhaps we can discuss how they should have priced the electricity to cover these costs, or where the original profits should have gone; building another power plant isn't a bad way to go, as with funding the decommission of another plant. Either way we can't let this kind of stupid short-term accounting get in the way of providing essential long term services.

It's a lot of rather common techniques put together, but it's a meaningful result. But the summary is rubbish: on the same benchmark for which this algorithm gave 98.52%, there was an older algorithm that gave 96.33%, and a graph showing small incremental improvements up till then; saying that older results "never came close" is a blatant lie.

The methodology seems legit, in the sense that it's consistent with those used for previously accepted results. On the other hand, I still have the same old concern on all machine-learning type research. I believe the model is useful, but getting experimental results like this is about tweaking details to fit the same test data, and as soon as the accuracy hits a new high you can publish. There are certainly ceilings that a model can't break, and the results certainly show that their model has a higher ceiling for a specific dataset than before, but it doesn't prove that the algorithm is universally better than previous ones, certainly not humans.

Even if the proofs are true, the physical implications were drawn too far. It doesn't account for the possibility that quantum effects do involve (super-)exponential classical computation; that the universe doesn't care if its rules can be enforced "efficiently" by a Turing machine.

Not if you can't or don't plan to spend it until 30 years later... Sure there's a judgement in risk, such as Bitcoin going down to nearly nil and gold keep having some ups and downs like it always did, but due to the factor of long term risk the proportion choosing US dollars would be much lower.

Mainstream languages do something like this: uniq ( sort( grep ("foo" , cat file) ) ).

We're not waiting for another Einstein or Newton. They've always existed in the world of academia; tons of great geniuses, many of them on par with some of greatest men in history, have been working tirelessly over the last century. It just takes more than one man and one lifetime to prepare the soil with knowledge and insight for a breakthrough like relativity.

Time complexity aside, insertion sort makes sense to laymen, bubble sort not so much. At least not that much easier than mergesort, quicksort and heapsort. There are ways to organize code that make those slightly more advanced algorithms easier to read.

Bubble sort has a problem that it is not painfully obvious you'll end up with a sorted list; a layman only sees a lot of swaps and it takes a little bit of non-layman thinking to believe in it.

With heapsort you first use function names like InsertValue and RetrieveMinimumValue to encapsulate the voodoo, and if the reader believes in these functions he'll understand the purpose. The voodoo part is less straightforward, but it still can be organized in certain ways that's easier to understand. You can name operations with the proper level of abstraction and granularity: RemoveMinimum, FindNewMinimum, SwapParentChild etc. A layman would most likely get stuck only at the lowest levels, that is understanding how FindNewMinimum works, and I don't think that part is harder than bubble sort.

Mergesort and quicksort are more easily understood by laymen iteratively, when you go bottom-up, rather than top-down by recursion. With similar abstraction like above you can reduce both of them to a structure that makes sense to the laymen except possibly at the bottom level; I personally think for mergesort even the bottom level is quite straight forward (merging lists), while with quicksort the proof lies slightly more outside the scope of the code.

These, of course, are not the most desirable codes for professional programmers.

There must be some direct cash made somewhere by an insider. Perhaps some people were secretly holding some shares of Nest, or the VC behind Nest?

Just like OOP eventually breaks down into CPU instructions, and you can use any Turing complete language to implement anything, I think he has a fair point. It sounds like a way to design a language so that complex event/trigger type stuff are easier to implement and debug systematically, perhaps using a slightly different way of thinking. It might not be a fancy thing, but I don't think there's anything wrong about giving it a name.