Apparently the early script drafts had a more plausible explanation: that the spare brain capacity of humans in a dream-like state was used as processing power to run the AIs. One of the editors thought this was too complicated for a movie-going audience to understand and so replaced it with a magic perpetual motion machine.

Translation is like predicting the weather. If you want to do an okay job of predicting the weather, predict either the same as this day last year or the same as yesterday. That will get you something like 60-70% success. Modelling local pressure systems will get you another 5-10% fairly easily. Getting from 80% correct to 90% is insanely hard.

For machine translation, building a database of 3-grams or 4-grams and just doing simple pattern matching (which is what Google Translate does) gets you 70% accuracy quite easily (between romance languages, anyway. It really sucks for Japanese or Russian, for example). Extending the n-gram size; however, quickly hits diminishing returns. Your increases in accuracy depend on a corpus and when you get to the size of n-gram where you're really accurate, you're effectively needing a human to have already translated each sentence.

Machine-aided translation can give huge increases in productivity. Completely computerised translation has already got most of the low-hanging fruit and will have a very difficult job of getting to the level of a moderately competent bilingual human.

Underpowered? The Gameboy had a 4.19MHz Z80-like chip. You could happily emulate it on a 66MHz 486 without taxing the host. The ARM11 in the RPi is massively overpowered for this use! I don't really understand what the feminism you're seeing here is.

The power consumption of the RPi (especially if you're not using the GPU) is tiny in comparison to anything with motors in it. I'd rather trade a slight reduction in battery life for being able to use a rich programming environment than save a few mW and be forced into a constrained microcontroller development environment. It might be different if I were planning on mass producing a few thousand and needed to save costs, but for a hobby project or even a prototype I'd happily overprovision on CPU power.

That's a non-sequitur. The cost is born by society. Government is the name of the body that we elect to represent society. If taxing an activity reduces it, which, in turn, reduces a cost that is born by society, then the government has done its job. The point of such taxation is to reintroduce externalities into the costs, so that the market will correctly adjust.

The advantage of the cars in this model is that they speed up unloading. Go and watch a freight train being unloaded some time, it's a massive endeavour. Now imagine if each of the trucks could just drive off along the roads on its own as soon as the train arrived at its destination.

Because, in the USA especially, lots of Federal government money has been spent connecting the major population centres with roads. Very little has been spent on the rail infrastructure. If you can make it work on roads, then you can take advantage of all of the existing infrastructure cheaply.

Lidar is typically used in situations where you have this kind of interference, but the solution is conceptually pretty simple. You tune the detectors to a specific wavelength and you vary that wavelength between devices. You can typically vary the wavelength in software, so if you detect a lot of interference then you just hop to another frequency. If someone is intentionally blanketing your entire band, then it's not really any different from someone shining a floodlight in the eyes of human drivers - you just point them out to the police.

That's true on the desktop, but absolutely untrue on mobile devices. You typically have an OpenVG accelerator in the SoC that uses a lot less power than the GPU (or is part of the GPU, but uses a lot less power than the 3D mode).

If you are planning on taking Bitcoins in payment for some real product then you want some assurance that you can sell the Bitcoins for more than the cost of that product. You get this implicitly with established currencies, because they have reasonably stable inflation rates and, most importantly, those inflation rates are usually tied to your local economy and so the value of the currency doesn't alter in terms of what you can buy with it in the short term (if you're hoarding significant amounts of currency, rather than investing it in your business, for example using it to pay your employees or buy stock then you're a currency speculator and so have different requirements). With an immature and volatile currency like Bitcoin, you effectively need to buy Bitcoin futures or have some other entity underwrite your Bitcoins.

If you want to be in the business of offering payment services that accept Bitcoins, then you want to be able to do this underwriting. People will be a lot more willing to accept Bitcoins if you can say 'the value of BTC fluctuates relative to USD, but we guarantee that we will always buy them at this exchange rate. If you use this exchange rate for setting your prices then you will never suffer from these fluctuations'. To be able to do this (and not make huge losses), you need to be able to influence the market price by buying and selling in large quantities and by avoiding selling when the price is too low. Having a large initial stash of Bitcoin helps with this.

Paper doesn't scale very well. I have a repository for a project that's been going on for a few years and has a few hundred photos of whiteboards. Trying to find one is almost impossible because there's no full-text search for photos of whiteboards. If you don't need diagrams, then running OpenEtherPad with a machine connected to the projector as a client and just saving the output is much better, but I've not found a good equivalent that supports drawing (especially not free-form drawing on a tablet or whiteboard and then automatically recognising shapes and handwriting, as the Newton's drawing program did 20 years ago).

Enforcement of non-competes varies a lot between jurisdictions, but there is some precedent for requiring the company that demands the non-compete to pay the employee's salary for the duration. This also provides a good incentive for non-competes to scale with employment time.

A contract does not require a signature, it requires a meeting of minds. A signature is one way of demonstrating this. Accepting the pay cheque and showing up for work is another.

How on earth are you buying CFLs with only a 3-year warranty for $12 each? The only ones I've seen that are that expensive come with a 10-20 year warranty. 5 years is about the shortest I've seen and that's on ones that are under $1.

The youngest CFL I've replaced was, I think, 3 years old. I left most of them behind when I moved house after living in the same place for 7 years and most of the CFLs had never been replaced since I installed them shortly after moving in. I was a poor student back then, so all of them were the absolute cheapest that I could find.