![](/static/253f0d9b/assets/icons/icon-96x96.png)
![](https://fry.gs/pictrs/image/c6832070-8625-4688-b9e5-5d519541e092.png)
Fuck it, let’s all just start winding our own magnetic core memory arrays.
firmly of the belief that guitars are real
Fuck it, let’s all just start winding our own magnetic core memory arrays.
Open source is just another commons, and companies have a way of uncontrollably exploiting common resources until they collapse.
In the case of open source, it’s healthy in the sense that money is flowing, we have companies sponsoring projects, tons of code is available for inspection and reuse, etc. Very nice. But if you go back to the original concepts of free software, in many cases we struggle with actually exercising the four freedoms. Red Hat has engineered an EULA that basically lets them ban practices that had been thought protected by the GPL for at least a generation, and so on and so forth. So is the open source community healthy or dying? Doesn’t the answer to that depend on your priorities?
I think it would make a lot of sense to try to create an economic model that can fund open source software development without relying on corporate injections of cash. It’s not that they don’t pay for it ever, they just pay for it to the bare minimum extent. IE, the heartbleed fiasco – tons of companies were freeloading off one guy and like half the Internet’s security got fucked for it. Imagine if OpenSSL had had some kind of economic support structure in place to allow for, uh, more than one guy to manage the encryption library for like half the Internet before something insanely stupid and predictable like that happened. Well, we can never have that with corporate-controlled open source.
Since CRF settings aren’t that useful for hitting a specified filesize, you can use the following equation to calculate the bitrate needed to encode a video of a given runtime at a given file size
b = (t*8*10^6)/s
ie, my copy of Serenity is 01:58:55 long, which is 7135 seconds (see https://www.calculateme.com/time/hours-minutes-seconds/to-seconds), I want it to be 2.5gb, my equation is
b = (2.5*8*10^6)/7135 = 2803 kbit/s
You can use any tool, handbrake, ffmpeg, whatever, any codec, and this equation will tell you the average bitrate needed to hit that file size. You would use “vbr” encoding mode instead of crf. I’d recommend enabling 2-pass for x264, not sure if this would be needed/is available for x265 as I’m a bit of a stick in the mud re: video codecs.
Couple notes, I’m using SI units (powers of 10 instead of powers of 2) for the conversion, and am converting from bytes to bits as this is a more common unit to represent bitrates. If your software uses different units for the bitrate for some reason, or you prefer representing file sizes using gibibytes/etc then you’ll need to rewrite the equation accordingly
With the SSD’s I can afford, there are what you might call “net negative savings” when I save maybe a couple dollars in power a month but have to replace them every few months. We can’t all afford EVO’s.
deleted by creator
I mean, with stuff like ZFS, it’s a little hard to justify the outlay for all solid-state disk storage when I can build out a large storage array using HDD’s and use one mid-size SSD for ZIL and then L2ARC to provide read/write speedups. Who actually cares what the underlying storage mechanism is as long as the dataset is backed up and the performance is good?
Hi, sorry I just saw this. “SFF” is short for “small form factor.” It’s just industry jargon for “a small PC.” They tend to be designed to use less power which makes them a good fit for home servers. Pretty much any line of PC sold to businesses, like Dell Optiplex or HP EliteDesk, will have small form factor variants.
Everyone likes to trash machine learning because the power requirements are high, but what they don’t realize is that we’re in the very first days of this technology (well, first couple decades of the technology being around, first few years of it being advanced enough to have anything to show off). Every technology that got bundled together into your phone was equally as useless when it was first invented. Honestly, compared to the development of most other technologies I’ve looked at, the pace of development in AI has been shocking.
Literally once a week, I see some news story about AI researchers delivering an order of magnitude speedup in some aspect of AI inference. The technique described here apparently allows for a 20x speedup on GPU’s.
Shout out for ODROID, their product revision cycles take too long (lmao why are they still selling a 32-bit chip that was an iffy investment back in 2013), but when they drop new stuff, it tends to be great.
deleted by creator
Bonus: there is a literally endless supply of used x86 SFF hardware from large institutions, so unlike SBC’s, there’s no special, weird supply chain managed by an English educational nonprofit that could just suddenly decide to not sell to the public for years at a time.
Would you also agree it’s rude to imply that this group of researchers, who actually advanced the state of the art in machine learning, are just a bunch of ChatGPT jockeys who don’t deserve credit for their work?
You don’t get to the point of solving scaling problems without having something to scale first.
Yeah, so the actual law is that if you didn’t do any work and just gave ChatGPT or Midjourney a prompt and it shat out a picture and then brag to the copyright office in your application that you didn’t do diddly squat, the work effectively had no human authors. If, instead, you build a new machine learning model, tune it for your specific problem, analyze the results, and furthermore, break new ground understanding how it solved your problem, and then you write the paper, in fact, you have tons of ownership over the work.
The fact people can’t tell the difference between the two and are actually upvoting you kind of says a lot about how little most people understand this stuff.
deleted by creator
Education.
All of the hacked systems in this article are home based systems.
[citation needed] because that’s not in the article. According to the article, attackers used automated scanning software, which strongly implies they brute-forced cameras connected to the Internet with default or weak credentials. That has nothing to do with whether or not the service is based in the cloud.
In general, cloud services have far better security than DIY systems
As a matter of fact, it’s known that the leading cloud-based surveillance system, Ring, has been subject to employee abuse and user accounts have been widely compromised via credential stuffing. In fact, Amazon is currently facing a proposed order from the FTC over the fact that they allowed abuse by employees and more or less knew for years that their lax security practices were placing their customers in danger from cybercriminals. Hell, it’s 2023 and all you have to do to pre-empt most credential stuffing attacks is enforce 2FA, and this was optional in a HOME SECURITY PRODUCT from a LEADING cloud provider. “In general cloud providers have better security” my ass.
Cloud based security only gets better when regulators force cloud providers to improve security, after cloud providers allow hackers to harm thousands to millions of customers.
I’m just gonna say it again: the cloud is just someone else’s computer.
I also gagged when I saw the name, but it’s a team of researchers at Western Sydney University in southern Australia, who are probably simply too high up in their ivory tower to realize/care what Deep South means in America. According to them, it’s an homage to a couple other systems:
The supercomputer is aptly named DeepSouth, paying homage to IBM’s TrueNorth system, which initiated efforts to build machines simulating large networks of spiking neurons, and Deep Blue, which was the first computer to become a world chess champion. The name is also a nod to its geographical location.
deleted by creator