andars

12 3 bit characters.

The drum stored 16384 36 bit words, equivalent to 72 * 1024 bytes (72 kb).

12 "character" words of 36 bits.

Those markets are also not demanding support because many of them have no display at all. This is definitely true for HPC GPUs and will be true for deep learning servers.

I don't know about Indian trains, but in the U.S. most locomotives are diesel-electric, meaning that they use a diesel powered electric generator coupled to electric motors. Definitely plenty of electricity.

I believe you are thinking of The Idea Factory.

Looks like the JS parser is created using Jison, a parser generator, and the generation of GLSL from the AST is performed in src/backend/webgl/function-node.js using a single-pass tree walk.

I do not see how sample size has any relevance. There was no claim.

I also use a iPhone that is a few years old. My experience has been that Google sends me to AMP pages in Safari and scrolling literally doesn't work. It just does that springy thing (as if there is no more content) on both top and bottom.

Seawater is a pretty great absorber of EM energy at all frequencies above tens of kilohertz, which is probably around the switching frequency of the drive inverter.

The key words are "in classical mechanics." Nobody wondered whether the mass of two stacked bricks was equal to the sum of the masses of the individual bricks. On the other hand, burning cloth seems to "leave behind" a lot less than you started with. Lavoisier's contribution was to demonstrate that even in chemical reactions, with shocking changes in color, state of matter, etc, mass was conserved.

Wilczek is discussing the ingrained assumption about the bricks.

> In reality, our 5-volt CMOS system will consider anything above 1.67 volts to be a 1, and anything below 1.67 to be 0.

Worth noting that the region from 1.67 V to 3.33 V is undefined and systems in practice will not behave nicely for signals in this range. A CMOS logic 1 needs to be above 2/3 Vdd to be reliably recognized.

He probably got the 1970 ETH during the presale for ~500 USD.

They started running three shifts some months ago.

5 years ago, over 100 million tablets had been sold worldwide and the iPad was entering its 4th generation. Your description is more apt for 10 years ago.

Mobileye.

AOT compilation of js is probably not worthwhile.

Brilliant.

I worked on getting rust to run on an arm core a few years ago, and the process was honestly pretty miserable. This is so far ahead from what I had going, it is hardly comparable. Looks like a very streamlined and elegant process. Excellent work.

Decent idea, but unfortunately unconvincing examples.

Mendel has been the subject of controversy regarding whether his results were fraudulent. Hardly a great choice for an exemplar, whatever the truth is.

Linus Pauling (described as an "amateur biologist" in a caption) was hardly an amateur interested in the structure of DNA, having written his PhD on x-ray diffraction. In addition, he is introduced in the text as "Linus Pauli," a Nobel-prize winning physicist. This seems to be a Mendelian cross between Linus Pauling, the Nobel-prize winning chemist, and Wolfgang Pauli, the Nobel-prize winning physicist.

Francis Crick, who made the title as an example of the power of amateurs, was pursuing a PhD in the field of x-ray crystallography, which sort of disqualifies him an "amateur" in my mind. Those "careful biology experiments" were studies of x-ray diffraction images of DNA, making Crick's physics background entirely relevant.

I have. By and large, I agree with you, but there are some arguments to be made in favor of allowing monopolies. I find the story of Bell Labs (mentioned in the article) especially striking: a frankly unreasonable number of innovations and ideas were developed there. The creation and continued existence of Bell Labs as a dominant force would likely have been impossible without Ma Bell's monopoly.

Peter Thiel also makes a case for monopolies in Zero to One (do not interpret as an endorsement).

Maybe, but unlikely. Lasers need line of sight and would generally be a bad idea at the power levels required (basically a weapon), and microwave power transmission requires large receiving apertures. Current "wireless charging" technologies are pretty much universally near field.

"consider stationing employees who need to concentrate outside the city"

This reminds me of the Bell Lab's old Holmdel research complex. A ton of early microwave research was done there in the 50s when it was basically just a farm. It wasn't till later that they built the facility that now stands there.

Without disclosing the cost.

That's all fine and good, but the only source NYTimes article you suggested is SpaceX, so you're not getting commentary from any outside experts. Instead, you are just getting quotes and paraphrases from this and previous SpaceX press releases.

"small performance increase"

FT can get 30% more mass to LEO. That is a pretty big performance increase, especially considering that it enables SpaceX to carry the maximum payload of v1.1 in expendable configuration and still land the FT rocket. Even if we assume that half of that performance boost came from the other improvements, 15% is a significant increase.

Lots of people do write in Java for performance. It is all relative.

JIT performance benefits are highly dependent on the language.

Something is fishy. Author claims 10x "faster" than Node, which seems to imply that they are beating V8. I find this highly unlikely given how fantastically resistant JS is to AOT optimisation and how many talented engineer-years have been put into the development of V8. I am very interested in seeing some numbers and hearing about the techniques employed.

> Wrong.

I'd say it's more like "not quite". ~10 years ago, the statement was pretty much correct. GPUs were entirely SIMD, so they couldn't truly branch, but could fake it with predication. Longer branches would involve executing both branches on every thread and only committing the results of the active branch.

Modern GPUs can do much better, since individual warps/wavefronts can truly diverge. Within warps, however, it's still a bit of a mess.

This strikes me as a pretty useless distinction. Github frequently refers them as "sites", i.e. websites, and advertises using them to host Jekyll-generated blogs.

This is why we can have nice things. People always push up against limits as long as they aren't shut down.