Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
twtw
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
54 ms
·
511.
▲
Nv-Wavenet: Better Speech Synthesis Using GPU-Enabled WaveNet Inference
(devblogs.nvidia.com)
3 points
by
twtw
9y ago
|
0 comments
512.
▲
by
twtw
9y ago
> What would basic circuit analysis lessons look like 10 years after they become commonplace? Exactly the same as they do now. Memristors aren't linear time-invariant, so it's extremely unlikely that they would show up in a bas
513.
▲
by
twtw
9y ago
21e9 transistors.
514.
▲
by
twtw
9y ago
Does anyone know how this compares to ActivityPub?
515.
▲
by
twtw
9y ago
"You need ... a business plan." I don't agree. What is the business plan of email (by which I mean SMTP, not some webmail provider)?
516.
▲
by
twtw
9y ago
Ditching CUDA would mean accepting at least 10-30% performance hit on NVIDIA cards and losing the ability to use the newest NVIDIA improvements. OpenCL doesn't know about HMMA (aka "Tensor Cores"), so it would be about an ord
517.
▲
by
twtw
9y ago
Source? Double engine failures are incredibly rare, and I've not been able to find anything that indicates Sullenberberger was involved in any prior to 1549. In simulation, probably.
518.
▲
by
twtw
9y ago
... and were completely redesigned starting in the mid-2000s to be good general purpose parallel processors. Modern GPUs have very little specialized graphics hardware. This doesn't mean there aren't better architectures, but this
519.
▲
by
twtw
9y ago
TL;DR: Parametrize your loss function and wrap a normal policy optimization with a random search to find a better loss function. Don't call it "random search," call it "evolution strategies" to make it sound sophist
520.
▲
by
twtw
9y ago
Space has expanded in the interim.
521.
▲
by
twtw
9y ago
GPUs can do things besides machine learning. It uses a maximum likelihood estimator, so no training data.
522.
▲
by
twtw
9y ago
Yes. The vast majority of the folding@home compute comes from GPUs. http://folding.stanford.edu/stats/os
523.
▲
by
twtw
9y ago
I'm a big advocate of "don't tell people what to do." Musk seems to have a pretty good handle on his own limits. He's been doing crazy work hours and sleeping under his desk occasionally for over a decade. Just beca
524.
▲
by
twtw
9y ago
"most people miss" that because that isn't what he says he is trying to do. Musk has said on several occasions that he is trying to fix some of the earth's problems. By all means, speculate on Elon Musk's hidden p
525.
▲
by
twtw
9y ago
I understand that, but a reusable engine is not the same as a reusable rocket. If we are going to discuss engines, it seems that the appropriate comparison is the development of the Merlin-series engines, which started around 2004.
526.
▲
by
twtw
9y ago
Are you talking about the Promotheus engine? Everything I've read about Arianespace's plans indicate that they aren't terribly interested in reusability. Ariane 6, which is slated for first launch until at least 2020, will be
527.
▲
by
twtw
9y ago
This is consistent with my experience. Many papers seem to rely on unstated conventions regarding the problem definition and use of variables. Say I'm not in the "small target audience who have a deep understanding of the problem
528.
▲
by
twtw
9y ago
Many universities have lab materials for introductory electronics courses available publicly online. For example, here are the labs for Berkeley's EE105: http://inst.eecs.berkeley.edu/~ee105/fa15/labs.html I
529.
▲
by
twtw
9y ago
Sure, I was thinking of this specific case only. If someone starts to use tiny cameras in a way such that they are actually useful for remote sensing, then by all means regulate them.
530.
▲
by
twtw
9y ago
Sorry, could you clarify your comment? Who has failed at their job, and who is making excuses and attempting to get press?
531.
▲
by
twtw
9y ago
It's a bit of a farce to classify these cameras stuck on the side of a rocket with horrible resolution and which are clearly not designed to effectively image the surface of the earth as "remote sensing systems." The definiti
532.
▲
by
twtw
9y ago
I read most of the relevant sections of the document, and there are good reasons for regulating remote sensing systems. I'm fairly familiar with remote sensing systems (SAR, scatterometry, etc) and I'm most interested in reasons w
533.
▲
by
twtw
9y ago
The Model T is not a competitor.
534.
▲
by
twtw
9y ago
It falls around 220th in the November 2012 list, and would make it onto the top 10 for the November 2007 list (comparing Rpeak) and the June 2008 list (comparing Rmax). Systems of comparable performance in the top 10 used between 300-500 kW
535.
▲
by
twtw
9y ago
100 GPU rig would not have 512 GB in one address space, accessible by 16 GPUs. Each GPU can directly address the memory on any of the 16 GPUs.
536.
▲
Nvidia Unveils 2 Petaflop DGX2
(hothardware.com)
3 points
by
twtw
9y ago
|
0 comments
537.
▲
by
twtw
9y ago
Maybe something like "meet in person, exchange contact information, communicate via phone, text, email?"
538.
▲
A Model, You Know What I Mean?
(argmin.net)
1 points
by
twtw
9y ago
|
0 comments
539.
▲
by
twtw
9y ago
I haven't done the course, but I'm fairly sure that XOR isn't a universal gate, so you probably don't start out with "nothing more than an xor gate." Based on the name, I would guess you start out with only NAN
540.
▲
by
twtw
9y ago
Work at their intended goal, as in rapidly moving humans and stuff from point A to point B.
More ›