HNHacker News
TopNewBestAskShowJobs

fluffything

2,917 karma · joined April 21, 2019

submissionscomments
fluffything··on Amazon gets priority while mail gets delayed, say US letter carriers
Yes, that's the one.
fluffything··on Amazon gets priority while mail gets delayed, say US letter carriers
There is a John Oliver's Last Week Tonight episode from a couple of weeks ago about this topic. Completely worth the 20 minutes it takes to watch it.
fluffything··on Nvidia will build 700-petaflop supercomputer for University of Florida
The FLOPs quoted by national lab supercomputers are usually measured double-precision FLOPs from HPL (High-performance Linpack - at least on the Top500 reports).

IIUC, HPL can use double-precision tensor operations. It might also be able to use mixed-precision for some computations.

Are the performance numbers for this HPC system at University of Florida also from HPL ?

fluffything··on Nvidia will build 700-petaflop supercomputer for University of Florida
I'm a bit baffled.

How high are the tuitions at the university of florida?

This supercomputer is more powerful than many at the national labs and must cost a fortune (multiple 100 millions of dollars).

fluffything··on Report: ARM is for sale and Nvidia’s interested, Apple isn’t
The question is at which price. Those mandates _never_ fix the price. At best, they might restrict how often ARM is allowed to raise it, and maybe even by how much, but that would be it.
fluffything··on Report: ARM is for sale and Nvidia’s interested, Apple isn’t
> So long as whoever is buying ARM is legally mandated to provide licenses

Why would they? With Intel, AMD, IBM with PowerPC, etc. CPU ISAs are not a monopoly.

Whoever buys ARM might want to do that to start cranking up licensing fees, and might not give a damn if this results in customer loses.

fluffything··on Big Sur is both 10.16 and 11.0
They can just not update their browser - which is essentially doing "no work", and is free.
fluffything··on AI in physics: are we facing a scientific revolution?
This looks really interesting.

I think it would be a good topic for a blog post or teaching paper that shows how to do this for very simple problems "end-to-end" (e.g. advection eqt, diffusion eq, advection-diffusion, burgers eqt., poisson eqt, etc.).

I see the appeal in showing that these can be used for very complex problems, but what I want to understand is what are the trade-offs for the most basic hyperbolic, parabolic, and elliptic one-dimensional problems. What's the accuracy? What's the order of convergence in practice? Are there tight upper bounds? (does that even matter?), what's the performance, how does the performance scale with the number of degrees of freedom, what does a good training pipeline look like, what's the cost of training, inference, etc.

There are well-understood methods that are optimal for all of the problems above. Knowing that you can apply these NN for problems without optimal methods is good, but I'd be more convinced that this is not just "NN-all-the-things hype" if I were to understand how these methods fair against problems for which optimal methods are indeed available.

fluffything··on Big Sur is both 10.16 and 11.0
And doing this encourages websites that actually need to detect this platform properly to introduce the next generation of horrible things.
fluffything··on AI in physics: are we facing a scientific revolution?
I see a Poisson solver in the docs.

Is there a paper comparing the performance of this particular solver against the state of the art ?

(if you are using GPUs, the AmgX library has a finite-difference solver for Poisson in their examples - very far from the state of the art, but a comparison might put performance in perspective)

fluffything··on Binary Search: A new implementation that is up to 25% faster
Yes, its essentially a heap. These just happens to be cache oblivious for performing binary searches.
fluffything··on Germany’s top financial supervisor dismissed a decade of warnings about Wirecard
Because people make mistakes, and if you make an accounting mistake in a company listed in one of the largest indices, those mistakes are usually in the order of billions.

So it definitely makes sense to require multiple accountants to check the books of these companies _for mistakes_.

Also, running the simple algorithm is the last step of the job. When somebody gives you their books and transaction data, you need to convert that to a format you can first process, which is 99% of the work. Books, receipts, etc. come in dozens of different formats. Running the algorithm is 1% of the job.

E&G is required to check that, if there is a receipt from an external company that gave Wirecard 2 billion, and another receipt of Wirecard sending those 2 billion to a bank in the Philippines, that those are properly noted in the books and everything balances out properly.

They are required to assume that Wirecard is working in "Good faith", and this is not a German thing: everything else is just impracticable.

Even if you were to throw an infinite amount of money and time at the problem to verify that all transactions actually happened as reported, if all sides of the transaction are in on the scam (the company that gave wirecard 2 billion, wirecard, and the bank that stores them), then there is no way for E&Y to find out.

If they are aware of criminal activity, they must report it, and maybe E&Y did not wanted to be aware of that, but who knows, maybe Wirecard was just good at scamming, or E&Y was aware of that and reported it to the German SEC. We do know that a lot of people did report Wirecard to the German SEC over the last 10 years due to suspicions of criminal activity, and the German SEC, which is the only authority that can investigate Wirecard properly, did nothing about that.

fluffything··on Germany’s top financial supervisor dismissed a decade of warnings about Wirecard
Not in Germany, where this company is listed. There, the auditor is allowed to (or even required to) assume "Good faith".
fluffything··on Dijo: A terminal-based habit tracker written in Rust
`rg` is distributed via multiple channels - most distros have packages for it.
fluffything··on Dijo: A terminal-based habit tracker written in Rust
> It it only the most lazy and selfish developers who would not utilize this.

Are you calling the whole Gentoo community lazy ?

If anything, I'd consider them to be the opposite of lazy, since making sure that all software always builds properly on each of your users machines takes a lot of work (IMO much harder than only making sure that it builds on your package manager controlled servers), and Watts.

fluffything··on Binary Search: A new implementation that is up to 25% faster
Not really.

On a GPU for example, if you have an array with 100.000 elements and want to find the index of one value, you are probably best of by using 100.000 threads to find that value.

So that's 100.000-ary search.

On a CPU, using an Eytzinger layout will probably also give you better performance than any of the N-ary approaches.

fluffything··on Germany’s top financial supervisor dismissed a decade of warnings about Wirecard
You definitely do not understand the word, because it is not limited to people, the word "insider" is limited to the information: is the information publicly available? If the answer is "no", then it is "inside" information, and trading on it is "insider trading".

These short sellers were trading on publicly available information. They just gave this information more publicity after getting their short positions, which is what you should do if you discover a market inefficiency that nobody else has discovered yet. That's your reward for making the market more efficient.

BaFin claims that this is insider trading was BS then and is BS now. They just don't understand how markets work, what their job is, and as a consequence did a poor job.

fluffything··on Germany’s top financial supervisor dismissed a decade of warnings about Wirecard
IIUC, EY was paid to check that what was reported in the books balances to zero.

They were not paid to check that what was reported in the books was "correct".

That's not EYs job. If somebody thinks a company lies in their books, they report it to BaFIN, and it is BaFIN job to investigate. They did the opposite of that here though.

fluffything··on Germany’s top financial supervisor dismissed a decade of warnings about Wirecard
And be compensated?
fluffything··on The Fake Cisco
Minus the support from Cisco, which you won't get once you tell them your serial number, which either isn't registered to you, or doesn't exist.
fluffything··on German court bans Tesla ad statements related to autonomous driving
1. Yes

2. Some company will succeed at L5 FSD, eventually. Just because it can be done does not mean that Tesla will do it. Tesla has been selling Vaporware for a decade, and they are investing a lot of resources in trying to get FSD to work with the hardware they sold customers 5 years ago to save face, while right now we don't know, and many do not believe, that such hardware suffices. Companies that have not tied themselves to such promises are much more flexible from a technology point-of-view to make quicker progress.

For all we know, next year somebody could ship a smaller, better, and cheaper lidar systems, that might give every manufacturer willing to use them a huge advantage over Tesla. Tesla would be in a very tight spot to incorporate such technology into their existing customer base.

I don't think we will see L5 in the next 10-20 years and a lot of things can happen in such a time-frame technology wise.

fluffything··on Building Elixir from source segfaults on macOS on Apple Silicon while compiling
> clunky behavior, (no window snapping)

my windows snap, tile, etc. when I drag a window to a part of the screen, it snaps there.

I haven't see a kernel panic on MacOSX in the 15 years that I've been using it, and I haven't seen one on Linux either, but maybe my workload is just different.

fluffything··on Huawei 5G kit must be removed from UK by 2027
UK is in a weird spot now alone in the world out of the EU.

The country is essentially for sale. China just needs to make the "right" political donations.

fluffything··on German court bans Tesla ad statements related to autonomous driving
The law is against fraudulent advertisement.

The channel used to perform the advertisement and how expensive that channel is are completely irrelevant.

For example, if Tesla claims in their website that their cars can drive themselves and they can't, that's false advertising and illegal in Germany. If Tesla organizes a concert in some city somewhere, and the singer states that Tesla's cars can drive themselves and they can't, that's false advertisement.

If Tesla distributes stickers to their car owners that claim that Tesla's can drive themselves, and their car owners stick them in public bathrooms where people can see them, that is, as well, illegal advertisement, even if Tesla did not stick those stickers themselves.

The law basically requires all companies selling products in Germany to be honest about what their products can and cannot do. This is good for consumers, and good for companies doing business there, because everybody is forced to play by the same rules.

The definition of being honest and what communication means etc. are all super loose, so most companies don't risk lying about their products. There are dozens of consumer protection organizations that'll sue a company for you due to false advertisement. The main consequences for the sued company are usually damages if there are any, and mainly the fines designed to discourage false advertisement. Most of the money ends up on the tax payers accounts, so consumers are really encouraged to report these times of crimes.

fluffything··on German court bans Tesla ad statements related to autonomous driving
When a plane auto-pilot fails (e.g. Boeing 737-MAX), the manufacturer is responsible for the failure and its consequences.

That has always been the definition of auto-pilot.

Tesla might be able to redefine the term in the US, but this ruling is proof that other countries disagree with that.

Whatever Tesla is shipping isn't an autopilot, and calling it that is just deceptive marketing.

fluffything··on German court bans Tesla ad statements related to autonomous driving
> Tesla has no idea if they'll need to replace or upgrade hardware either.

They have upgraded the hardware 3 times already.

Maybe the hardware in the last generation of Tesla's suffices, but right now, nobody knows.

fluffything··on Building Elixir from source segfaults on macOS on Apple Silicon while compiling
> You get them as Web apps.

Microsoft web apps suck. The outlook webapp does not even support smart folders, and dozens of other "basic" features that the native apps do support. Same for Word, excel, powerpoint, etc.

This might be an option in the future, but right now, paying microsoft for their webapps is just a waste of money.

fluffything··on German court bans Tesla ad statements related to autonomous driving
That marketing was not misleading, it was and still is a lie. They have upgraded the hardware at least 3 times already.

The first time this lie might have been just sheer incompetence. But after three upgrades, I can only assume that the lie is malicious and negligent.

Musk has been claiming that his "fleet" of 10,000 robotaxis will arrive next year for a couple of years already. I mean, he is going to be right eventually.

But Tesla's competition, e.g., NVIDIA et al., are claiming that for full-self driving we'll at least need hardware with an order of magnitude more TOPs than what the current generation provides, and that we won't see that hardware before 2022-2023. They are also not promising that this hardware will suffice, only that they are pretty sure that it cannot be done without it.

I'm skeptical of Tesla's claims to say the least, but I guess we'll see. Statistically speaking, Musk predictions about when we are going to see full-self driving have been 100% wrong to date.

fluffything··on Building Elixir from source segfaults on macOS on Apple Silicon while compiling
> and to navigate to a specific folder directly, it is a nested menu item

Cmd+G - G for Go, and I agree that having to go to the menu bar item called "Go" is not super discoverable, but I found it once, and been using Cmd+G since then. I guess users that don't use shortcuts have more problems with these.

> Copy+Paste is not?

Cmd+C Cmd+V ? I am not sure this really needs an icon. I mean, if you go to the "Edit" menu, you see the shortcuts for Copy, Paste, and Cut, which are the same as on Windows or Linux, but using Cmd instead of Ctrl (which is consistent with the rest of the OS).

> tabs were finally added very recently

Tabs were added a long while ago, 5 years at least. But you don't even need tabs for this. I just do Cmd+Shift+Left to tile one finder window to the left of the screen, and Cmd+Shift+Right to tile another window to the right, and that's it: two windows, side by side.

I mean, it's not i3 or sway, but works fine.

fluffything··on Building Elixir from source segfaults on macOS on Apple Silicon while compiling
> Solaris and HP-UX are enterprise UNIX,

By "enterprise" a lot of people understand native Outlook, Powerpoint, Word, Excel, ... Adobe tools, ..., etc. support. On HP-UX and Solaris you get none of these.

← PreviousPage 9 of 31Next →