VladRussian2

young naive hackers and BigCo's "developers, developers, developers" marketing management drones. Call it a hackathon and they'll come.

I may be too old, yet i just don't get how exciting is the app that builds an itineary for visiting most valuable clients or any other CRM app. Until of course 1M is thrown into the pot. And of course excitement disappears again once 1M happens to not even make it to the pot. :)

btw, closing shop at 6pm isn't a sign of conspiracy - it was already an hour past 5pm as in "9 to 5" :) Talking about canyon deep gap between different mentalities ...

>Another major cause is farming with Roundup, a herbicide that kills virtually all plants except crops that are genetically modified to survive it.

now i know explicit answer why GMO is bad. You never know before-fact, yet can always be sure that whatever good at first look thing appear it will ultimately be turned into real bad one.

because Java arrays length is int, 32bit limited. I think we'll see these articles more and more :)

>even longer checkpoints

yes, that is the point as big flush instead of many small ones would take either the same or, usually, less time than cumulative time of small flushes because of IO ordering and probability of some writes hitting the same data block.

>That's the sunk cost fallacy talking.

nope. Unstable currency is just not very conductive to business. Speaking from experience with inflation. As any transaction has 2 ends, i suspect that high deflation has the same effect as high inflation. This is why Bitcoin as a currency has major built-in defect, while it seems to be wonderful value storage like Picasso or Ferrari California.

no, i'm talking about transaction log ("redo log" in Oracle parlance). Switching log files causes checkpoint (ie. flush - it is when the index datablocks changed by the inserts you mention will finally hit the disk )

http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUES...

MS and DB2 have similar behavior.

>The log writes don't saturate the array itself--but the log file has a limit to how many blocks can be appended--even on fast arrays

yes, the issue usually isn't the transaction log append speed. Instead, it happens too frequently that the log is configured to be too small. A log file switch causes a flush of accumulated modified datablocks of tables and indexes [buffer cache flush in Oracle parlance] from RAM to disk. With small log file size, the flush happens too frequently for too small amounts of modified data - this is where GP mentioned random IO bites in the neck.

>FWIW, many vision researchers believe that the resemblance of the first convolutional layer to Gabor filters is perhaps more a case of selection bias than anything else. The argument goes that were they not the output of the first layer, that paper wouldn't get accepted =)

well, i can see the temptation - the orientation and spatial frequency selectivity are the major characteristics of cells in V1 and the receptive field for the first layer there does look like Gabor

http://www.scholarpedia.org/article/Area_V1#Receptive_fields

I agree that such a good resemblance of the learned kernels to Gabor is too good, this is why i used "uncanny" :) If it is real then i think it manifests very interesting and, no pun intended, deep emerging properties of the neural net learning process (something along the lines "maximum entropy kernels while still doing the job" as the asymptotic state)

Btw, is it really selection or confirmation bias?

And to expand on previous point of convoluting the input with many-many kernels - happens to be at the order of 40 per "pixel":

"V1 contains a vast number of neurons. In humans, it contains about 140 million neurons per hemisphere (Wandell, 1995), i.e. about 40 V1 neurons per LGN neuron. Such divergence gives scope for extensive processing of the images received from LGN."

>some pretty recent results -- most likely, the big breakthrough on ImageNet by Krizhevsky, Sutskever and Hinton (http://www.cs.utoronto.ca/~ilya/pubs/2012/imgnet.pdf), which uses convolutional nets.

kernels learned by the first convolutional layer (the figure 3. on page 6) have uncanny resemblance to Gabor function-modeled orientation-selective cells ("bars and grating cell") in the primary visual cortex. Looks like computers are on the right track :)

http://www.cs.rug.nl/~petkov/publications/bc1997.pdf

"The discovery of orientation-selective cells in the primary visual cortex of monkeys almost 40 years ago and the fact that most of the neurons in this part of the brain are of this type ..."

The difference here is a "number game" - visual cortex contains cells whose receptive fields' positions, eccentricities, sizes, orientation, number of excitatory and inhibitory zones (e.g. Fig.1 in the link) make a reasonable coverage for the space of possible values. Ie. the number of these cells is in the millions vs. 96. Of course it is only a matter of computing power to run all reasonable combinations of kernels emulating the real visual cortex, yet it would put immense computational challenge onto the second and next layers until we understand what [should] happens there.

having the same DSL connection over recent several years my price has gradually risen from $26 to $40 per month. I'd say something is really wrong with a technology sector if it has such negative price/performance dynamics.

well, it is called intelligence fusion centers and license plates is just a minor piece. Pretty spotty. Face recognition info from surveilance cams pretty spotty as well. Credit card transactions - similar. Cell phone position is much better... Once taken together ("fused" dare i say :) - works like a charm i guess.

Guys from Palantir would know better i guess what Suspicious Activity Reports are cooked there:

http://www.palantir.com/_ptwp_live_ect0/wp-content/uploads/2...

"...many Fusion Centers have gradually shifted to a broader mission of providing general criminal intelligence and all-crimes, all-hazards analysis support for constituent agencies."

>The ability for the executive to grab enough power to coopt the legislature and deceive the courts - and get away with it. Were that to have actually come to pass little else would have been important. It's intriguing and relevant (for us) that this overreach was made possible by the knowledge gap generated by rapidly advancing technology

while it make the events relevant to this audience, the technology involvement is a red-herring here as history is full of examples where executive branch was able to perform the power-grab utilizing the technology of the time, frequently an absolute power-grab and with wide popular support in the name of "the greater good" as perceived at the time by the populace. In its current form technology allows to minimize direct violence required to perform the grab, it allows to stretch and smooth the shock of the transition.

just to put things in perspective - is is any surprise that people who is ok with torture would lie? Why would someone expect it different?

just by association with inpenetrable, armor-piercing bullets, etc...

http://media.uow.edu.au/news/UOW118285.html

"UOW researchers have used graphene to develop a new composite material which can produce the toughest fibres to date- even tougher than spider silk and Kevlar!"

union of concerned robots may have some concerns wrt. your comment and attitude it projects

i'm confused - an .edu paper has been already mentioned while "Weird Science" hasn't been yet.

this is why i said 80mph not 30mph.

To inherent instability of 3 wheelers one also need to add inherent instability at speed of low-weight small vehicle with high height to width ratio (an issue arising from aerodynamics) what 3 wheelers are. The best one can get is a T-REX like (http://www.campagnamotors.com/). Now, if somebody managed to sell T-REX at $7K instead of $40K+ - that would be something :)

The main point of HyMotion3 is motorcycle-style whole vehicle body leaning inside the turn - completely different game.

i kind of hate to put myself at risk of overturning at 80mph on mild windy day on mild turn on 280 just because 3 wheeler (motorcycle) is easier (read cheaper) to certify than 4 wheeler (car). Of course that is also because i can afford to pay additional money of "car price", and i'd probably thought differently if i couldn't.

Not saying that such a vehicle don't have its place. Though i have hard time imagining it. If not highway - than cities - European cities are well populated either by small 4-wheel cars or just scooters. Doesn't seem like there is niche for car size 3 wheeler without full benefits of a car.

> As the PG docs say, there is virtually no performance difference at all between all three

Btw, there is a clear [though infrequent] case where using CHAR(X) vs. VARCHAR(X) may cause huge performance penalty with iron platter HDD. CHAR(X) will always contribute X to the row length (even if data value is Y < X, so VARCHAR(X) with such value would contribute Y) thus making it more probable to have your row TOAST-ed ( i.e. have parts of row stored separately in separate table) thus causing a noticeable (when only few rows are accessed) extra disk head lookup on non-cached data. Similar to C/LOB in-row limit exceeding on other databases.

>I have to disagree. You should always put limits on everything

like any orthodoxy it should have a limit put on it.

>But if there is multiple interfaces (such as a REST api etc) to your database then you have to remember to put them in place everywhere.

multiple interfaces going directly to database - that's a much bigger problem that the rest pales before it. In particular, multiple updates of limits would constitute only minor share of headache causing updates and the multiple updates itself would be just a minor share of the whole headache.

> the crowds are truly wiser than an expert.

it depends on whose opinion one asks - crowd's or an expert's :)

at this thickness it would be something like aluminum, not steel. Good as a frame material, while softness make it not that good for point penetration resistance (it would be somewhere like 1mm steel equivalent, so seems i was wrong above). 5mm Al cover for 10ftx6ft battery would weight 80kg - doesn't look like there is much room for increasing.

So it doesn't sound like they did a screwy job - that's good to hear (as i said i'm a fanboy of Tesla and think nothing can stop them but their own hubris - when/if it appears), yet it also may mean that there is no easy, "low hanging fruit", way out.

>What exactly do you think should happen when a jagged metal object collides with your car with enough energy to rip through critical components? Do you think susceptibility to telephone poles and meteorites is also a defect that auto makers need to fix?

>Short of encasing the car's primary energy store with 6 inches of plated steel, what are you looking for, exactly?

that is a bunch of 6in fallacies. Car fuel tanks are made of something like 1-2mm steel. As far as i understand Tesla doesn't have a 1mm steel sheet below the battery. That is the defect made obvious by these 2 accidents.

Btw, modern cars are much better at hitting telephone poles and does tend to explode less when being hit by "meteorites" (say bullets) than they were say 40 years ago. So, yes bad crash results ("susceptibility to telephone poles") for example would be a defect.

with all due respect to Musk (kind of hero of mine for fighting the uphill battle to really advance our civilization), and with Tesla S being an object of my desire, this sounds like a defect that have to be fixed :

( http://www.theverge.com/2013/11/19/5121706/tesla-elon-musk-r... )

"it was damage to the cars' undercarriage and batteries, caused by striking metallic objects on the road while at high speeds, that caused both fires."

The software update to increase suspension clearance looks like just a [cheap] band-aid to me.

The third car fired as result of crash - that seems like a real and hard to avoid thing for lithium batteries and does seem like Tesla is ok here - the fire looked to be pretty limited.

the autorotation requires a highly trained pilot and it is an emergency situation. Failure of 1 motor/propeller out of 16 is like a "low gas" warning, time to think about where to conveniently and casually land, instead of immediately autorotating down to whatever lake or heavy forested mountain you happened to be flying over.

i'm wondering whether the fact that it is in UK is important, ie. wrt. who paid for the [experimental] surgery. As we all know, while coming up with idea is great, an actual implementation faces the issues of resources/money.

http://www.yasamotors.com and similars. 75kw/27kg

the advantage here is the ability to have several (like 8-16 motors) so the failure of one motor isn't leading to a crash). It is the same paradigm shift as when horizontally scaled commodity hardware systems with expected multiple points of failure have almost everywhere replaced the highly-reliable Big Iron systems.

Another advantage is the possibility to have the propellers in ducts/shrouds (eVolo didn't do it, yet nothing prevents it) that is not doable for one big helicopter's propeller, thus such machine can fly near buildings/trees/etc... where helicopter can't. There is also issue of control in tight quarters where air flow is hard to predict and is affected by near by objects - multiple electric motors are much more flexible and faster reacting than one big engine+propeller.

Like with electric cars, the main issue is "range". It is solvable by having a series hybrid drive (especially once new gas engines specifically developed for series hybrids - ie. higher efficiency and lighter weight at the cost of power band - come to market). While such system willn't beat efficiency of 787 or fly transatlantic, it would be unbeatable for Santa Cruz to Mountain View commute.

> $12,000 - too steep for a hobbyist

market segmentation. They used the "industrial" keyword. So it could have been anything between $300 for Parrot and $35K (airframe only, the whole system is $200K+) for Raven.