svantana

63 ms·

601.
by svantana 11y ago
Except that after those ten years, the value of the property will exceed $0. In fact it's rather likely to be worth more then than now, so it's actually an investment. As Tony Soprano said, "buy real estate cus God ain't
602.
by svantana 11y ago
Agree, they should tone down the 80's sci-fi vibe IMHO. But I think you may overestimate the control your average industrial worker has over their work situation. That is, if it helps the bottom line, these will be bought and deployed
603.
by svantana 11y ago
Parallax is common enough, but the way the background adapted to the content of the article was really well done IMO.
604.
by svantana 11y ago
As a remix artist myself, my experience from youtube (which I believe have similar deals in place) is that rights owners will tolerate (i.e. monetize) straight copies of their work, but many remixes, mashups and cutups get blocked. Which is
605.
by svantana 11y ago
I would argue the desire is not so much for the stuff itself, but the social status that the buyer hopes the stuff will bring them (i.e. status symbols). Proof: a fake diamond ring will be frowned upon, even though noone but an expert with
606.
by svantana 11y ago
Not only that, but saying "the top 68 percent" is flat out wrong. Even if a normal distribution is assumed, it should say "the middle 68 percent" or something to that effect.
607.
by svantana 11y ago
Even worse, the numbers are differences , which can take on negative numbers as well, so it doesn't even make sense to talk about a distribution (in the probability sense). I read the AVC blog once in a while as I feel he's a pre
608.
by svantana 11y ago
A machine-learned system is only as good as its training data, at best. In this case, it was trained on natural-looking images, so results on "unnatural" images will be unpredictable/random/wrong. One way to fix this wou
609.
by svantana 11y ago
The results are comparable to human transcribers, they note -- which is more a testament to the low quality of Mechanical Turk work than the high quality of this system. Surely a word error rate of 8% (for clean speech) would be unacceptabl
610.
by svantana 11y ago
This is similar to the Moravec Transfer [1], a proposed method of (slowly) transitioning to a non-biological brain. Indeed the whole concept of "the self" is important to maintain, even if it is to some degree an illusion. 1. htt
611.
by svantana 11y ago
I'm fond of saying that founders don't need an idea, they need a plan. An idea is just a pie in the sky, but a plan can be executed and is therefore more easily dismissed as unrealistic or flawed. So if anyone wants to tell me the
612.
by svantana 11y ago
Not only that, but the specialness is totally in the eye of the beholder - one man's trash is another man's fine art. And if told that trash was painted by a certain Mr Picasso, perceptions might change rather quickly. There'
613.
by svantana 11y ago
That's not really fair though, since any deterministic function can be "back-propagated" using the chain rule (or even automatic differentiation), even though it's not really necessary for simpler models such as GMM and
614.
by svantana 12y ago
This is very nice, however I wish they would have used a traditional rendering technique (e.g. raytracing) for the decoder stage. It would have been more difficult to compute the gradient, but maybe not too bad if employing some type of aut
615.
by svantana 12y ago
Interesting, I've long wondered why parametric nonlinearities are not used more often. It adds very little to the overall parameter count (the number of units is most often dwarfed by the number of connections), but it should increase
616.
by svantana 12y ago
Is it just me, or did this article completely assume that it's reader is male?
617.
by svantana 12y ago
It may be interesting to note that here in Sweden, "jaywalking" is not a crime, and we have what may be the lowest rate of driving-related deaths in the civilized world. People here do indeed jaywalk more than in many other countr
618.
by svantana 12y ago
This seems wrong. My understanding is that the ~200 spikes/second limit derives from the cell needing to "reload" before firing again, rather than some built in latency. Relaying a spike can be very quick indeed. A better con
619.
by svantana 12y ago
It wasn't written in 1980, it's a tongue-in-cheek response to an article written by a 2015 teenager that went viral yesterday: https://medium.com/backchannel/a-teenagers-view-on-social-me...
620.
by svantana 12y ago
I've noticed similar behaviors in my own sleep patterns. But for me, another problem arises - there's a huge chance I might wake up at 3 am, hungry and unable to go back to sleep.
621.
by svantana 12y ago
See also: Acid House. An astonishing number of classic tracks in the genre are based on "random" sequences on the Roland TB-303 synth, repeated hundreds of times. (The synth's sequencer is tenuous to program, so just pushing
622.
by svantana 12y ago
Not only that, but 10x or more money is loaned out or invested than is actually in the accounts. That's why we have financial crises.
623.
by svantana 12y ago
Halting has nothing to do with being bug-free -- lots of buggy programs halt, and some obviously bug-free programs may not halt (e.g. a brute force search to determine Goldbach's conjecture)
624.
by svantana 12y ago
But the article itself states that this difference was not statistically significant, and should therefore be attributed to randomness?
625.
by svantana 12y ago
Interesting work, however training on data seems unnecessary; chess would be perfect for unsupervised learning - initially it could be trained against an existing chess program, but as the models improve, they could start competing against
626.
by svantana 12y ago
Very neat stuff, but seriously, who uses (tiny) screenshots of code in 2014? Also, too bad it's CUDA only - i.e. no support on most mainstream computers. There's not even any software fallback -- execution fails on my macbook, for
627.
by svantana 12y ago
Combine that with a karma system à la HN and you should have something...
628.
by svantana 12y ago
The results are quite impressive. But why are they not combining the gyro data with that from the accelerometer? That should theoretically give superior results (i.e. sensor fusion). Or maybe they do, but simplified it for this write-up.
629.
by svantana 12y ago
This is really beautiful. However, be warned, the webGL widget completely froze my Firefox (on OSX), I had to kill the process to recover. No problem in Chrome though.
630.
by svantana 12y ago
One of the biggest problems of modern urban life is: there is minimal to no cost (in the long term) of being an asshole, because people you encounter at random you will rarely meet again. Under such circumstances, it's reasonable to re

More ›