varelse

98 ms·

661.
by varelse 11y ago
I hate to agree that this sort of manipulation is inevitable in a free market, but I do. It's either that, or broadly, no more cheap #$%^ on Amazon. So what do you pick? Occasional sociopaths or social good at the expense of everyth
662.
by varelse 11y ago
Amazon can apparently go up or down 50% depending on whether a butterfly sneezes in Australia (it's wildly danced between 284 and 580 in the past 12 months alone). In the background of such high volatility, it would be hard to pinpoint
663.
by varelse 11y ago
I'd throw stupid money at NASA any day, just maybe not so much for manned exploration until it's cost-effective. We both want more space probes I suspect. But from the above articles, I see two trends. From (1) Nonmilitary scienc
664.
by varelse 11y ago
No, you can pass the controller states and timestamps to whichever node is acting as its server. You then extrapolate everywhere else until a correction arrives from the server. It worked magnificently and we used to watch efficiency impr
665.
by varelse 11y ago
Moore's Law w/r to clock frequency ran out over a decade ago. And that sucks, because it's trivial to make single-threaded code reproducible. But parallel code is how one continues to benefit from Moore's law w/r t
666.
by varelse 11y ago
If your parallel reductions are consistent, you're cool in my book. That said, I've solved a lot of dynamic load-balancing behavior with 64-bit fixed point reductions and decisions. Maybe that will work for you? Unless the timin
667.
by varelse 11y ago
To get this for free (for various values of the word "free"), you'd have to encapsulate reduction across arbitrary distributions of processes and threads. Good luck with that(1)... As for the second part, one needs a great d
668.
by varelse 11y ago
As long as there is one true arbiter of the game state, I can't see how. The default these days is to have a master server, but back in the day, I developed a distributed server state that let each node determine the true state of its
669.
by varelse 11y ago
Neither the instruction sets nor floating point are irreproducible at all. What are you talking about? In fact, FP32 is commutative (go ahead, check it out). It's just not associative. If you can't figure out how to reduce sums
670.
by varelse 11y ago
So unless you want your core talent to flee for startups and begin the cycle anew, one has to find a middle ground, no? The moonshots seem like a reasonable attempt to do so to me. But perhaps that because I'm 100% progress-driven. P
671.
by varelse 11y ago
Really? IMO as a former Ph.D. academic who chose industry because there was no future there, science funding in America has dried up into a shriveled husk of its former self in an age of bottomless pockets for overseas military adventures
672.
by varelse 11y ago
So right now the moonshots are privatized and the liberals are unhappy about it. When the pendulum shifts, the moonshots will be publicly funded again and the conservatives will be unhappy. Takeaway: happiness is unattainable.
673.
by varelse 11y ago
I think this problem takes care of itself, but it takes time, more time perhaps than people are willing to allow. 1. As organizations metastasize, the negative tail of the competence distribution becomes more and more populated. In respons
674.
by varelse 11y ago
There's a clear and massive movement of data forward for prediction in a CNN, and all that plus saving intermediate state and then moving the same amount of data backwards when performing backpropagation. But don't believe me, don
675.
by varelse 11y ago
Fixed as of Maxwell to the best of my knowledge. But even then, I found them more efficient for reduction operations in global memory than any other method (using fixed-point math for places where a deterministic sum was required).
676.
by varelse 11y ago
GPUs aren't cache-coherent yet. That said, they have fantastic atomic ops performance relative to CPUs. I'd guess that the lack of performance benefit so far is because it's possible to write many algorithms with the assump
677.
by varelse 11y ago
BMWs, Audis, Teslas, Mercedes, Porsches and I'm sure I missed a few others: these are all sacred cow status symbols. You're going to get downvoted by their respective followers saying anything bad about them. I personally really
678.
by varelse 11y ago
Except for the people who take these things to 11: http://www.hennesseyperformance.com/2015-dodge-srt-charger-h...
679.
by varelse 11y ago
Little irks me more than the occasional driver who comes to an unnecessary stop to merge onto a highway (I rear-ended one of these geniuses once, neither car was scratched fortunately). Or the guys who slow down to 20 mph when one is suppo
680.
by varelse 11y ago
This is where I always thought HUDs would come into play. And yet, apparently, mostly, no... One could probably have integrated this into Glass as well, but given the snail's pace of Android Auto (and the raging quality issues of Andr
681.
by varelse 11y ago
If all you care about is 0-60, the Dodge Charger SRT Hellcat does it in a nearly as stupendous 3.4s and beats the Tesla by 0.4 seconds on the quarter mile for half the price... http://www.caranddriver.com/dodge/charger-
682.
by varelse 11y ago
The Internship and Mr. Robot have about as much in common as Twilight and Fight Club (all IMO of course)...
683.
by varelse 11y ago
Is it possible that the reason Quoc Le never published the code is that it got tangled up in the word2vec patent? At GTC 2015, Ren Wu stated (paraphrased) that the reason DNNs are advancing so quickly is because all the practitioners are s
684.
by varelse 11y ago
I like to refer to DNNs (specifically back-propagation on them) as the world's most lucrative application of the chain rule from Calculus...
685.
by varelse 11y ago
And yet gas is still over $3 a gallon. Funny that...
686.
by varelse 11y ago
Very strange, there are videos for the 2014 edition on YouTube, but not this year... https://www.youtube.com/playlist?list=PLhoHEZlJjdQKI1cs5yPRU... Because while I'd love to fly to Italy, I suspect I'd be complet
687.
by varelse 11y ago
Wow, that's a 10 year-old article and still most of the code I see is single-threaded Java and Python. I rarely even see C code anymore. Methinks that explains the popularity of Hadoop: just break your task into manageable subchunks a
688.
by varelse 11y ago
Thank you, seriously, thank you, for pointing out the obvious that eludes so many... Have an upvote...
689.
by varelse 11y ago
What can I say? I respect a shipping product and nothing short of it. He entered the conversation pitting his powerpoint processor against shipping hardware. Have we not tired of all the broken promises of Kickstarter and IndieGogo yet?
690.
by varelse 11y ago
IMO it would really inspire confidence to see one of these chips beat traditional neural networks at something ... Speech recognition, image recognition, tumor detection, whatever. I don't care, but something ... Right now it seems l

More ›