mostly_harmless

I, and everyone else, has to pay taxes in USD. That will never replaced.

Isn't that sort of the point of raising interest rates? Makes debt more pressing to slow down money velocity and loan creation? (Thus curbing inflation).

Sure, if you push the rates too high, you'll hit deflation, but the point would be to strike the balance before that happens.

> imagine if you could pay 10 cents to read an article or tip a web site.

This doesn't need crypto. This can be done with something like Paypal.

> How badly does this torpedo googles ad biz?

A solution based on spite is a bad solution.

This is a dumb comparison. An ounce as a unit is arbitrary. A Bitcoin as a unit is arbitrary (but arguably less so). I'm in a metric country so lets use that.

1 kg of Rhodium is trading around $493,189 vs $66,787 per Bitcoin. Rhodium is 738% better if you equate 1 kg to 1 bitcoin! Wow, put all your money into precious metals!

There is no 'true' value of anything. The entire fundamentals of trade depend on the fact that everyone values things differently.

If I'm a rice farmer, maybe I want fish more because I have none, and likewise the fisher would want rice.

In this case, my 'data' has absolutely zero value to me. What am I going to do with it? It's anonymised, so it's not really a security concern. And for the most part it's innocuous stuff like "I like video games". But maybe Google can use it to tweak for better ads, so they see value in it.

On the otherhand, having a robust, availible, free magic box that answers any question I can possibly have (Google Search) has so much value to me.

If you don't agree with the exchange, for the most part there are alternatives like duckduckgo (search) with the trade off of slightly worse results (arguably), or ProtonMail (email) where it will cost $12 per month.

But for me, I consider it a fair trade.

This is incorrect, that's not the reason they are linked. It is using the category box at the bottom of the page, and both Hitler and Gates were voted 'Times Person of the Year' at some point.

It looks like it uses links in the 'category' box at the bottom of the page. Both MLK and Zuckerberg appear in the 'Times Person of the Year' category.

It broke 9000 yesterday, didn't it? That's more than 10% in a day.

Is there any reason it is up, other than speculation?

I made a graph comparison between Toronto Tax and Bay Area tax.

https://github.com/bsurmanski/TorontoSFTax

Torontonians pay less tax for income up to 200k. Note: I also did not include the SF 1.5% income tax in the comparison.

You mention "fake solar observation glasses". do you mind elaborating a bit: how do you tell if they are fake? Is there something missing in the knockoffs? What is the effect of a fake?

Theres other hidden costs as well.

What happens when a harddrive fails in your colo? Unless you have best practice backups, you will lose customer data and trust.

GCloud abstracts the issue of hardware so you can focus on real business development. And to some, that's worth the cost until they can properly afford all the hidden costs of independent hosting.

I don't get the "Google shutting down a product" meme. I can't think of anything they've shutdown without a replacement, except for Reader. Are people still sour about that?

> Zcash is a technology, and like any technology it has multiple uses. I suspect that many of the best applications of this technology haven't been conceived of yet.

> 10% pre-mine to founders

This smells super fishy. Altogether, I've yet to see anything to do with cryptocurrencies be useful. Nothing but scams, hiding illegal activities, and hopeless optimism so far.

so now drivers need to host their own cryptocurrency pool?

My favourite way of looking at it is to imagine neurons in a neural net as analogue nand-gates.

The logic within a computer processor is made entirely with nand-gates, and a processor is able to compute any function. Nand-gates have 'functional completeness' and can implement any other gate (AND, OR, NOT, NOR), and any other high level construct with those functions.

Similarly, neurons in a neural net can output a 'NAND' function if set up correctly.

This provides a logical inference of computational completeness of neural-nets

I imagine this was designed at a higher level, then exploded to smaller parts. For example, usually a 7-segment display controller is used to translate an input value to one usable by the display, but it looks like he soldered one up manually.

Similar to inlining functions in software, but all done manually in hardware.

Ideally, it would be a completely automated system. You don't need to understand machine learning to know that it would be better for the computer to do everything itself without human intervention

ereaders are a really nice platform but the current problem with ereaders is the lack of tactile feedback, and the difficulty of random seeking.

The first issue is mostly just preference. being able to see how far finished a book you are and the satisfaction of flipping a page is nice, but has little utility.

The second issue is a bigger problem. anything but a linear read from start to finish is difficult on an ereader. the only reasonable use case of ereaders in the current form is reading a novel. anything with a more complicated interaction, such as textbooks where you need to jump back and forth for reference to equations; choose your own adventure books; or books with footnotes or endnotes are severely limited.

I'm not sure if either of these issues can be fully solved, but they will definitely need to be mitigated in the UI for ereaders to have a more sustainable demand.

In a more recent example, the F35 Jet is having delay and cost overrun problems. I hear that one of the reasons is that there is 3 different version of the same jet and it may have been cheaper to design 3 completely different planes oriented toward each niche.

The 3 configurations are:

A) Conventional takeoff and landing (CTOL). A normal fighter jet, intended to replace the F16.

B) Short takeoff and vertical landing (STOVL). This model essentially has a big fan in it to allow it to hover and land on aircraft carriers.

C) Similar to A, but with larger, foldable wings and a stronger tailhook for landing with a carrier arrestor cable.

Model B is intended for the Marine Corp, and C is intended for the Navy.

Now I admit that I am fairly uninformed about military hardware and design, but I think something can be concluded from the F35 program issues.

not quite. It is similar to longjmp except that this can jump down the stack. these labels can be jumped to even after the function returns

But programming is not equivalent to stenography. Stenography works on a stream of input speech converting it into text, whereas programming is separated into the 2 modes of browsing&thinking and actually editing.

But I think it comes down to preference of whether you want the universality of commands, or conceptual separation of modes.

The biggest problem I see with these global addressing schemes is that they have nothing to do with the location, and they dont have any locality. Snake.Dog.Rabbit is in a location with no snakes, dogs or rabbits and is right next to Lamp.Chair.Guitar? Theres no visible logic to it. its just easier to say 123 River Road. By the name you know its likely to be near a river, and it will be right next to 124 River Road

I think the idea is that the language is focused towards games. Games generally have some inherent shared qualities, mainly that they need to be real time and the language should be minimal overhead. On the otherhand, a language focused on enterprise software would sacrifice speed and low overhead to instead add some safety features like a garbage collector.

Its not a language only for games, but it is a general purpose language where the main feature set is chosen to benefit professional game development.

personal plug: I've also been working on a programming language in my spare time, called "OWL"; My language is, coincidentally, also meant to be a cleaned up C++ for game development. It is an object-oriented systems language, without a garbage collector where it's most notable features are:

* classes are reference types to avoid slicing (but may still be stack allocated)

* classes are reference counted

* arrays are sized

* semicolons are optional

* tuples are first class types. (essentially acting as anonymous structs)

* basic one-way compatibility for importing C headers

The compiler can be found here: https://github.com/bsurmanski/wlc

currently it compiles for both Linux and Windows, requiring LLVM as a dependency.

a game written in OWL for Ludum Dare 30 can be found here: https://github.com/bsurmanski/ld30

If it's a GC thing, you should also try using various D compilers to compare performance.

There is:

dmd (official D compiler),

ldc (llvm D compiler),

gdc (GNU D compiler).

When I was messing around with D about a year ago, dmd compiled code the fastest, gdc by far the slowest, but gdc produced the most performant code. ldc was middle ground for both compile speed and program performance. Perhaps the GC issue you are having is just a quirk of the current compiler you are using.

I think it would be interesting to be able to travel up the subnet nodes to bigger boards.

eg, with subnet A.B.C.D, being able to visit A.B.C.X, A.B.X.X, and A.X.X.X.

Additionally, subnets are a bit opaque. what does each one correspond to? if you wanted to extend this out even more, It would be nice to have a map showing how big, and where your current subnet is.

I think this ability to compute any function is important, and useful in understanding why neural nets work. Neural nets are essentially a blank computation matrix that, through training, best approximates a function.

comparing neural net neurons to logic gates in a computer circuit or FPGA, gives a nice intuitive understanding of why the training works in the first place. The training essentially sets up a custom virtual circuit for whatever function you need.

Given this property, it seems intuitive that they would be able to learn any function, since any function is representable as logic gates.

I wrote about this a few weeks ago:

http://serialprog.blogspot.ca/2014/07/neural-networks-like-c...

My point of view was that neurons in neural nets are essentially analogue logic gates. Given that combinations of logic gates are Turing complete, combinations of neural net neurons should be also.

My writing is not quite as nice or rigorous as the parent post, but the whole point of the blog is to get better at self-expression and explaining things.

I've used the leap motion, and so far the accuracy is nowhere near what they show in the video. simply forming the index finger and thumb into an O-shape was enough for them to disappear from existence.

If this is an active demo of their next generation software, I'm very impressed with the improvements, but I suspect its a mock-up.

I feel that 1-99 should be broken down into a more granular scaling.

something like: 1-10, 10-40, 40-100