derf_

21 ms·

211.
by derf_ 6y ago
> They could even do this accidentally/nonmaliciously Designing a development process that naturally achieves a malicious outcome is just as bad as actively making malicious decisions: https://www.zdnet.com/article
212.
by derf_ 6y ago
You can already sort of see what direction this would go in with education. Just look at student loans. They're available to basically anybody, and as a result you see a growing industry of for-profit colleges and continually rising tu
213.
by derf_ 7y ago
What if there are two (nearly) equally likely predictions? What if there are N, and they are not close together? "A single prediction that you can take the difference with" makes some assumptions about the shape of your distributi
214.
by derf_ 7y ago
The America Invents Act was signed into law in September 2011. This was exactly the time to do something about patent reform. There was a lot of discussion about it among tech companies at the time, but little leadership and no unified me
215.
by derf_ 7y ago
There is an extension to TLS 1.3 that supports encrypted SNI, and it has been deployed on some popular services since 2018. Of course, your DNS is likely still unencrypted. But maybe one day we'll get there.
216.
by derf_ 7y ago
This is a nice result for the single-scalar case, but it is not immediately obvious to me how it would scale to support SIMD, given its use of per-element indexing and a periodic conditional fallback to a slowpath which does not trigger at
217.
by derf_ 7y ago
> the Japanese were portrayed as this unique society where you need to say "excuse me" before approaching a stranger on the street asking for directions. I agree that part is not really unique, but if you are an obvious forei
218.
by derf_ 7y ago
Hello, fellow brilliant engineer here. In my experience, it is not as often as you might think. The key issue comes down to brilliance coupled with selfishness . If you look for it, you can see some common themes in these comments: ".
219.
by derf_ 7y ago
But then you've spent more money. You could have just bought more hardware in the future. For a fixed budget, assuming that Moore's Law still holds (dubious, these days), you should wait until the computation will take 26 months o
220.
by derf_ 7y ago
A lot of institutions used to run their own e-mail. Over the years I've watched as my e-mail addresses (both universities and my current employer) have been replaced by Gmail on the backend. All of them stopped being willing to manage
221.
by derf_ 7y ago
> everybody, no matter how big, needs to toe the line. Do that, and you won’t have any problems. And maybe you're fine with that today. But once you've ceded that ground, what happens when they move the line? Because they alw
222.
by derf_ 7y ago
No worries. It is rather the date that you compare it to that changed (filing rather than invention). I could have expressed that more clearly. It actually does change the dates for when patent documents are considered prior art (as describ
223.
by derf_ 7y ago
One might easily believe that based upon the name, but a number of changes to how prior art worked in the US were required to implement it. Please consider https://www.uspto.gov/web/offices/pac/mpep/s2152
224.
by derf_ 7y ago
So, my understanding of first-to-file is that any public disclosure counts as prior art for anyone else. So the only time you would have to prove the troll literally stole the idea is if your disclosure to them was non-public. In either c
225.
by derf_ 7y ago
> Computers, so far, don't have the context to do this As someone who did their Ph.D thesis on the statistics of shape using models based on the medial axis (i.e., a skeleton), I would beg to differ. Whether these models are as ea
226.
by derf_ 7y ago
It's possible, but with adaptive probabilities, the cost for never using a feature is pretty negligible (roughly O(log N) bits of overhead to learn that the probability is ~zero, where N is the number of times you code the symbol... an
227.
by derf_ 7y ago
It is easy to dismiss what Cisco demonstrated like this, but if you listen to the talk, they argue that having more tools available gives you more room to optimize for quality per bit per cycle , in the same way that having more choices
228.
by derf_ 7y ago
> So we are very differently positioned than we were when Firefox/Chrome overtook IE. I would agree, but not for the reason you state. IE was the standard back then, too. The difference is that once it was "good enough",
229.
by derf_ 7y ago
To be fair, BBC's sole reason for investing in Dirac (to my understanding) was to fit HD quality video into the pipes for an SD production system. As far as I know, it satisfied that purpose for them, so the project can be considered a
230.
by derf_ 8y ago
As someone who harassed you about VP8 support in the past [0], this is really great news. Thank you for all of your hard work. I am also really happy to see the transition to Unified Plan. [0] https://news.ycombinator.com/it
231.
by derf_ 8y ago
Many RTC stacks (and specifically openh264) also only support H.264 baseline profile. So you are likely to get higher quality with VP8 (subject to variations in the quality of the implementations on each endpoint, of course).
232.
by derf_ 8y ago
The part where he drove it for the maximum legal time he was not required to affix his license plates to the car (six months), and then leased a new one [1]. [1] https://www.itwire.com/it-people-news/enterprise/506
233.
by derf_ 8y ago
> ...fear can have a really positive effect when it comes to learning something. This has a built-in failure mode caused by the fact that the student is optimizing the objective function, "Reduce the potential for public embarrass
234.
by derf_ 8y ago
At one point I had a boss with (roughly) 50 direct reports. To paraphrase him describing how he did it: "badly". The company later was able to hire more managers and take the load off of him (and in fact he went back to being an I
235.
by derf_ 8y ago
The Firefox Reality team is much smaller than the FirefoxOS team was (i.e., by more than an order of magnitude).
236.
by derf_ 8y ago
I pretty much gave up on Matrix after the following exchange: https://news.ycombinator.com/item?id=8697151 Perhaps I am overreacting, but that sort of attitude on matters core to basic privacy in the post-Snowden era seems
237.
by derf_ 8y ago
It depends on what operation could potentially overflow. Mozilla has a CheckedInt (C++) type that checks all operations for overflow. You can see examples of how this is actually implemented here: < https://searchfox.org/m
238.
by derf_ 8y ago
That was the main reason. It also means that you don't need to pass around (a pointer to) a big table of function pointers throughout all of your functions that need to use accelerated routines. Instead you just pass a single integer i
239.
by derf_ 8y ago
One (non-generalizable) solution is to avoid using writable function pointers, to help maintain control flow integrity. For example, in libopus we have platform-specific SIMD optimizations that are chosen at runtime. The common solution is
240.
by derf_ 8y ago
Yes, it is the "Family-wise error rate". You need to perform "correction for multiple tests". The simplest, most conservative form is Bonferroni correction [1], where you just divide the required significance threshold b

More ›