cbr

I think your system is ignoring the color profile attached to the image. They wrote:

    Remember the red square with the faint WebKit logo?
    That was generated by creating an image in the Display
    P3 color space, filling it with 100% red, rgb(255, 0, 0),
    and then painting the logo in a slightly different red,
    rgb(241, 0, 0). On an sRGB display, you can’t see the
    logo, because all the red values above 241 in Display P3
    are beyond the highest red in sRGB, so the 241 red and
    the 255 red end up as the same color.

But if your browser ignores color profiles, and many do, then the P3 -> sRGB conversion won't happen, and your sRGB monitor will be told to display some 241 red and some 255 red instead of all 255 red.

A new city would be much denser than the suburban and exurban sprawl that currently fills our demand for new housing, and so would end up in taking much less farmland than continuing along the current path.

(And the amount of farmland we're talking about, both for new cities and for sprawl, is small as a fraction of total farmland.)

    there's almost no room left on the planet for building
    "new" cities of any significant scale (meaning of
    course not just the meticulously designed city core;
    but the whole meta-city that will grow around it --
    with suburbs, transportation links, etc)

No way. Spend some time in satellite-view in a mapping program and you'll see that there are huge areas of farmland that could be cities. There's a lot of space on the planet.

When building a city from scratch you can cut-and-cover everything, and you can choose somewhere with good soil conditions for what you want to build.

There are many more broke lawyers than broke developers.

(But do think you're right that they're not the bargain they already to be.)

Somerville is in the process of updating its zoning code: http://www.somervillema.gov/zoning/resources/2015-01-22-ordi...

My impression is that one of the goals of updating it is to make much more of the city be zoning conforming.

> I get all of my electricity from a nuclear power plant

If you use more power, that doesn't come from running the nuke plant more (we run them at 100% all the time) or building more nuke plants (politically impossible). Instead it comes from coal/oil/wind/sun: something else.

Sound just has an amplitude at every sample, but this is too low level to be very useful. Instead you would probably make a net process the same kind of features a human has, which is a list of frequencies at each sample, each with an amplitude.

(In humans we have oscillating hairs that perform an analog frequency extraction, while with computers you'd use a fourier transform to do it mathematically.)

It doesn't look to me like that's why Nest failed?

The 1st generation Moto X had a special minimal chip used just to detect you saying "Ok Google Now" (or a custom phrase) to turn the phone on when it was sleeping. But I think phones that only detect a start word when the screen is on probably don't have custom low-power hardware for it.

Reasons to think it's client side:

* Sending all microphone input to a remote server would be really bad for battery life and data usage.

* It still works even in airplane mode.

* On ChromeOS (which isn't Android, but may share the same code) their privacy whitepaper [1] says:

    If you opt-in to the feature, Chrome OS will listen for
    you to say "Ok Google" and then send the audio of the
    next thing you say, plus a few seconds before, to
    Google. Detection of the phrase "Ok Google" is performed
    locally on your computer, and the audio is only sent to
    Google after it detects "Ok Google"."

[1] https://www.google.com/chrome/browser/privacy/whitepaper.htm...

    After all, it's the city you can live without owning
    cars in the United States.

I know a lot of other people in Boston that don't own cars.

    Properties included in the model have total development
    potential, measured by FAR and residential units, of 20%
    or less of the maximum allowed under local zoning law. 
    All properties above this level are considered "built 
    out" and do not factor in to the maximum potential units
    number.

FAR is "floor area ratio" and controls how many square feet of building you can have on a given lot. I'd love to have a slider for this, since increasing the allowable FAR is a good way to make more efficient use of limited land.

(For example, zoning where I live sets a maximum FAR of 0.75, but most properties are over that limit and grandfathered in, averaging more like 0.90. Allowing people to build to 0.90 instead of 0.75 would not substantially change the feel of the area, and would make a lot more units available.)

> worked for black money and collect the social benefit at the same time

This one doesn't apply to BI, since you get the same amount of BI money whether you work or not and regardless of how much you earn.

Reading this makes me so glad I live in a part of the country where there are no HOAs and where multi-unit houses (2 and 3 unit freestanding) are typical. It's a great neighborhood, and the high density means I can live within walkind distance of way more friends than in the part of the city where I grew up.

Sorry, you're right, thanks for the correction.

(I don't know how I got to my initial reading!)

Ok, but then the supportive housing and eviction prevention money divides to ~$22k/person.

    better at parsing the Penn Treebank than the best
    natural language parser for English on the Wall
    Street Journal

I'm pretty sure "the 20 year old Penn Treebank" and "the Wall Street Journal" are referring to the same dataset here. In the early 1990s the first large treebanking efforts were on a corpus from the WSJ, and they were released as the Penn Treebank: https://catalog.ldc.upenn.edu/LDC95T7 People report results on this dataset because that's what the field has been testing on (and overfitting to) for decades.

(I worked on a successor project, OntoNotes, that involved additional treebank annotation on broader corpora: https://catalog.ldc.upenn.edu/LDC2013T19)

You get better transcriptions and better search results if you give your search backend a probability-weighted list of possible transcriptions and let it combine that with what it thinks you might be trying to search for given the context it has for you.

The combination of the title and phrases like "simulations using a mid-latitude climate couldn’t get the job done because the rocks didn’t weather fast enough" did make me think initially that they were proposing using this as geoengineering to pull CO2 out of the atmosphere.

(But yes, that's not what they're saying. And my guess is it wouldn't be anywhere near worth it.)

    15- 18% AC annual avg. capacity factor, orders
    of magnitude better than East Coast installs

Huh? It looks to me like the east coast is only ~30% worse than SF? http://www.c2es.org/docUploads/Solar-1.png

If you figure out how to hash 1000x more efficiently than other people and put that into practice without sharing your knowledge then you can cheaply get and maintain a hashpower supermajority.

(I'm not saying this is likely, just more likely than finding a large gold asteroid.)

The cheap way to scan books is destructive scanning, though.

    we'll never find an asteroid that's miraculously
    made almost entirely out of Bitcoin

I think we're more likely to find a very fast way to compute sha256 hashes than a 10T+ gold asteroid.

> I bet you could infer the energy usage of certain > appliances such as hair dryer, curling iron, stove ...

https://sense.com/ does this, figuring out what different things you run at home and knowing when they're on.

Trust works differently selling electricity. At each step in the chain from generation to distribution to consumption everyone involved can easily measure what they're getting. This means that each entity only needs information about the entities that it buys from or sells to. With ads this sort of verification is far harder, and passing lots of information back from the browser to each party involved is the only way the industry has figured out for keeping fraud down.

There's actually a weird way that control over your devices and privacy on your devices are at odds here: if your device was running signed code from the ground up, where you were given a browser and had no ability to modify anything, then we could build an ad system that leaked no information to the various ad industry companies but still let them be sure that a view was a real view and a click was a real click. But of course we do want to be able to control the software we run, and we successfully rejected trusted computing for most uses in the '00s, so instead we have all this tracking.

But once there are strong consequences for downtime the service provider is going to set up training, monitoring, oncall, etc to make sure things stay within the SLA limits. So you are effectively negotiating uptime.

Typically beta/stable is about stability, not about features or platform support.

Set your site up to be able to handle option A and option B at the same time, randomly assigning visitors to each. Now you don't need to control for anything, because you can just compare the A-visitors to the B-visitors.

In your case, option A started doing a lot better when you started running the test, which is moderately surprising, so maybe you messed something up in user assignment, tracking, or something. But if your framework is good, then it sounds like an external change. You still want to look at how A's rate and B's rate compare for the time period when you were randomizing, so 6.5% vs 5.9%, with the earlier 5.7% being more or less irrelevant.

You do need to do significance testing to see how likely you would be to get this wide a difference by chance. The easiest way to do that is to enter your #samples and #conversions into a significance calculator [1], and the p-value tells you how likely this is (assuming you didn't have any prior reason to expect one side of the experiment to be any better).

[1] Like this one https://vwo.com/ab-split-test-significance-calculator/ , but multiply the p-values it gives you by 2 to account for them incorrectly using a one-tailed test.

The idea is that you should prioritize based on impact-for-effort and not sweat the small stuff.