lightcatcher

Yes, and based on the $0.15/kWh comment, about $450/hour of electricity.

This is more in line with what I would have guessed for 220k cores.

I was about 10 years old and at a bonfire at a summer camp in Texas. It was dark when I was bitten, so I never saw the spider. It felt like there was something between a mosquito and a cockroach on my ankle that was hard to shake off. The bite didn't hurt and I didn't notice it at the time.

Next morning, I'm completely out of energy and feel awful. There's a region about 2 inches by 1 inch on my ankle that is pink, gummy texture, and looks like it's lightly coated in a white balm. A family member recognizes it as a spider bite and then we find the fang marks. Go to the doctor, who diagnoses as either a brown recluse bite with a light to medium amount of venom or a full dose bite from relative of brown recluse. Get some pills (probably antibiotics) and a cream to put on the area and told to stay off my feet as much as possible for next 2 weeks.

Spend the next 2 weeks of summer mostly laying on a couch. Ankle is in mild aching pain, otherwise I feel fine. Watch ~.6cm of flesh of my ankle turn to mush and decide that spiders suck.

End up with a scar similar to https://en.wikipedia.org/wiki/Loxoscelism#/media/File:BrownR... (warning: obviously mildly graphic). Scar has almost completely faded 15 years later, dislike of spiders has not. I still prefer spiders to ticks though.

I think I liked spiders before I got bit by a brown recluse. Now I like dead spiders.

What do you mean that it's "much easier to just pay people a decent wage"? Do you mean that it's much easier to convince companies to pay their employees more than they have to, or that it's much easier to pass higher minimum wage laws?

Neither of these seem easier than "convince the public to toss in a few extra dollars when they use a lowly paid service".

I see "make sure that all working people are paid fairly" as a very difficult problem, and tipping as a moderately effective hack that's easier to implement than the cleaner fix of higher minimum wage of basic income.

> Is it a particularly ethical standpoint to say "I don't pay my driver's a fair wage, but if I did then I wouldn't make money!"?

Not very ethical, but rational, profitable, and prevalent. Cost is a large area to compete in, and people are expensive.

It would be nice if companies acted ethically, but that isn't their reason for existence (especially in fast growing or large companies). If it's not surprising to me when companies act unethically, can I honestly say that I expect them to act ethically? Not if expect means "assign high probability to".

I agree that regulation is needed. It seems that often basic economics (people want to move around as cheaply as possible) lead to unethical situations (paying people as little as possible to move people around). I'm far from an expert of economics, but it seems like regulation is a way to change the rules of the game to have more ethical/fair/uniformly distributed equilibriums at some cost in net efficiency.

So if you were running Uber, would you increase your prices to pay a more fair wage even if it makes you more expensive than competitors? Would you expect to stay in business if you offered a commodity like transport at 20% higher prices than competitors?

I give about 30% of my Lyft drivers a tip, and I'm glad Lyft includes the option in their app. I've occasionally wanted to tip an Uber driver but then not had cash on me.

I don't think it's reasonable to expect ride share companies to pay their drivers a fair wage when they are in fierce competition with each other and self-driving cars are on the horizon.

I agree "mandatory" tipping like in US restaurants isn't a great system. I don't think including tipping in an app with a default of $0 makes it close to mandatory. Lyft could give stats on what fractions of rides have a tip given; my bet is it that it's relatively low. Making optional tipping easier is a good thing.

Agree with your point, but average latency isn't as simple as inverse of throughput, even on a serial processor.

Imagine a process that takes in a request, sleeps for 10s, and then provides a response. If taking in 1 million req/s, it can still provide 1 million responses/s for a throughput of 1 million req/s. Average latency is 10s.

Approximating latency as 1/throughput is only valid on a process that only handles 1 request at a time (no concurrency). I doubt this is the case for Go.

Latency impacts user happiness (did the page load quickly?). Throughput impacts operating costs (I need to buy N% more servers to serve as many requests with Go 1.8 as I did with 1.7.5).

From the original GitHub issue:

             Thread Stats   Avg      Stdev     Max   +/- Stdev
    Go 1.8rc3 Latency   192.49us  451.74us  15.14ms   95.02%
    Go 1.75   Latency   210.16us  528.53us  14.78ms   94.13%

Go 1.8rc3 has both a lower mean latency and lower standard deviation than Go 1.7.5. Go 1.8 decreased latency at cost of decreased throughput.

Looks like current technology is 1000x1000 per eye at 90Hz: http://www.digitaltrends.com/virtual-reality/oculus-rift-vs-...

8k x 8k per eye at 120Hz is 64x more pixels at 1/3rd increase in frequency ~= 85x more processing power. Making the (maybe faulty) assumption of doubling processing power every 2 years and that current setup is processor limited, this sort of processing power is ~13 years away.

Same computation but with 4k x 4k per eye predicts ~9 years of progress needed.

$3000/person is around or larger than the world's median individual annual income: https://www.givingwhatwecan.org/post/2016/05/giving-and-glob... .

I'd also expect the total wealth among the 100 billionaires to be well over 100 billion, considering just Ellison + Zuckerberg + Page together have over 100 billion.

In a similar vein to these two figures, the richest 62 people in the world hold as much money as the poorest 50% of the world: https://www.theguardian.com/business/2016/jan/18/richest-62-... . As a direct consequence, if these 62 people gave all of their money (except for a couple million each) away immediately, 50% of the world would have twice as much money.

(edit: I'm not suggesting that billionaires instantly give all their money away as a direct cash transfer. Just providing a counterargument to "billionaires don't have that much power")

It's slow, but working through good books is a valid path. It's a huge boost if know a math grad student to give you a bail out / explanation when the exercises get difficult (or you can't figure out why the book puts focus on specific topics).

> Transportation is not the largest energy expense that goes into a product like a car; the largest energy expense is extracting and refining the raw materials into processed materials that are ready to be used for final production.

Sure, maybe the GP's point is that (she thinks) (environmental impact of new Prius construction + n years of Prius driving > impact of n years of driving an existing pickup truck). I have no evidence for or against this, but I wouldn't be surprised if it's true for n=5 years and typical mileage/day. As you point out yourself, besides bringing the parts from all over the world, you need to mine the resources, spend the energy to cast them, etc.

> both apple and google have massive purchase orders for commodity nvidia gpus

source?

> So at best he seems to have gotten around 30% of the voters to vote for him.

What do you mean by this exactly? Trump lost the popular election with 47.5% support to Clinton's 47.7%. How did you come up with the "around 30%" figure?

> If that happens, gasoline cars will stop being produced overnight.

Some people will still pay for range or range at speed or less charging time.

http://jalopnik.com/they-drove-a-tesla-from-la-to-new-york-i... is a decent look at the state of technology. Driving across the US takes at least 12.5 hours of charging time, and averaged 65.4mph while ICE record averaged 98mph.

Even when "electric cars are cheaper on a capital basis", the technology will still be well behind ICE in some ways (and well ahead in others).

> the consequences can be much more dramatic

How are the consequences much more dramatic while flying? I can think of plenty of scenarios while driving where you go from perfect autopilot conditions with no visible risks to death of passengers or pedestrians in <5 seconds.

Any recommended resources (textbooks, course lecture notes/videos) for someone with moderate science background (say intro college courses in chemistry, electromagnetism, circuits) to learn about batteries?

Good to know about the JSON spec. The only JSON parser I used before jq was the Python stdlib parser. Python JSON correctly handles big integers, which lead to my (out of spec) expectations for jq to do the right thing.

> Given that we can't depend on people to do simple things like handling escape sequences correctly, that seems pretty important. (Also, the 'jq' command is nice.)

jq stores all numbers as doubles, which means it can't handle 64 bit ints (or larger integer numeric types).

It's otherwise a great tool and I wouldn't point out a negative such as this, but you did explicitly mention that you know JSON won't be fed to a bad parser.

There's something to be said to be familiar enough with the optimization that you can predict assembly that is not a direct translation of what you're writing.

For instance, when writing a scalar-processing loop over a fixed length array with no dependency between iterations you can reasonably expect the compiler to convert to SIMD instructions and do some unrolling. It's not too difficult to check the binary that this happened, and if it doesn't happen you just write it yourself with intrinsics.

I often think of C + compiler's optimization passes as a "scripting language for assembly".

Also from Houston.

Even if you get to the point where the heat doesn't bother you, it doesn't stop the heat from making you sweat and smell. I didn't mind going for a walk (or run) on Houston summer afternoons, but there's no way I could walk >.5 miles or bike to work without a shower at the destination.

Chicago has a variety of housing that is relatively affordable. From the US cities I have lived in (LA, SF, grew up in suburbs of Houston, spent ~2 months in Queens and a little time working in Manhattan), I found Chicago to be the most reasonable to live an affordable urban lifestyle in.

Chicago has the advantage of plenty of land as opposed to New York or SF.

There are massive sprawling suburbs, but I get the impression people move there when they have kids mostly due to better schools, not being priced out or leaving for safety.

Of course, Chicago has the most murders in the US (but still about half the murder rate per capita of New Orleans or Detroit), bitterly cold winters, and far from any mountains, but nowhere is perfect (where being affordable is part of being perfect).

> It sounds obvious in hindsight, but you can't exactly run 60 atari sims on a gpu.

Why not?

It's certainly easier to run parallel Atari sims on CPU, because CPU programs are typically written as single-threaded or with parameterized number of threads.

Running parallel Atari on GPU is completely possible, with either running an Atari game on the each of ~30 SMs or each of the 32 * n_SMs ~= 1000 warps. However, because GPU code is typically written and delivered as kernels which utilize the full GPU, this type of embarrassing parallelism over SMs or warps typically can't be gained from using an existing library.

> So if I understand what you're saying, a 1x1 convolutional layer for collapsing 100 channels to 10 channels would take a 100x512x512 tensor and collapse it to a 10x512x512 tensor?

Correct. As I understand it, this would be applying a 1x1 covolution with 10 filters to a 100x512x512 tensor.

I tried (and failed after a month of work) earlier this year trying to do ~20TB shuffles with Spark. I felt both relief and frustration reading this post.

Relief: I'm not an idiot, and the problems in the shuffle were likely in Spark and not just me being a beginner user.

Frustration: I wanted to group about 100 billion x 200 byte records into 5 billion groups. This seems like exactly the problem Spark is designed for and is advertised for. I had great difficulty even getting example Spark SQL code (or my own RDD based code) working. I hear so many great things about Spark as a tool for big data, and also "your data isn't big". I considered 20TB on the "low end" of big data, but a seemingly popular and widely used big data tool can't shuffle it without numerous bug fixes and pain on the part of the user. Shuffling 90TB was worth a Facebook blog post! This makes me ask: To all the people using Spark for "big data", how painful is it and how much data are you handling? It appears the answer for >=20TB is "very painful", and for <=5TB I think you're generally in "handle on single node" territory.

> Why would you write networking code in C in 2016, when there are better alternatives available (like this one)?

Support for kernel bypass networking libraries like ibverbs, DPDK (has an old unmaintained Rust wrapper) and other IO kernel bypass libraries such as SPDK and IOAT.

If Rust supported these libraries, I'd much prefer the future based Rust code to a massive event loop in C.

Another indicator of the article's quality: The company is called Metamako[0], not Metamoko.

[0] https://www.metamako.com/

I agree there's a lot of (potentially dangerous) ambiguity in what the name "autopilot" means and implies.

Personally, I'm not a fan of the name "autopilot" for 2 reasons.

Firstly, the name is sexy and filled with hype. Even for a fully autonomous "driver go to sleep" level 4 driving system, "autopilot" would be a great sounding name (much better than "full autonomy" or anything more explicit in my opinion).

Secondly, the name deviates from industry semi-standard naming such as "cruise control" or "lane assist". I don't believe Tesla's automated features are more advanced than the rest of the industry[0], but the name sounds more advanced. Maybe this is good marketing, maybe this is dangerous and deceitful, maybe both.

[0] I'm not an expert on the state of self-driving systems. If anyone here who is reads this and has some evidence that Tesla's autopilot features are more advanced that what's offered by Audi (for example), I'd love to hear why / see some sources.

Where in the US do you think there is a greater density of "Ayn Rand followers"?

> Aside: I seem to remember seeing something about the Arabian Peninsula soon becoming so warm you cannot maintain core body temperature by evaporation of your sweat, at which point it will be uninhabitable outside of modern buildings with A/C ...

No source on this, but I remember reading somewhere that in the hottest conditions recorded (Persian Gulf heat + humidity), the humidity would actually condense on a human like a cold water bottle on a hot day.