I thought "The Data Scarcity Problem" from the article is very well known to us engineers?
It's where the pulleys of a very sophisticated statistical machine start to show, and the claims about intelligence start to crumble.
Reason AI is great for boilerplate (because it's been done a million times) and not so great for specifics (even if they're specifics in a popular language).
Steam Deck. An awesome piece of hardware, a leap forward for Linux gaming, but as an early adopter I got the black screen of death (common issue) fairly soon. Tried everything to no avail. Now it's a very shinny and barely used brick.
If I understand the article correctly, any sufficiently capable attacker can:
- Know the global state of your GPU cluster via the client.
- Target the most struggling GPU instances specifically since the client decides which one to hit.
You offer a free tier which means anyone can get an account and try to do it (e.g. you can have one "harmless, mostly inactive" free account with the only purpose of retrieving GPU cluster status, and a bunch of burner accounts to overload struggling instances).
I may be completely wrong, but this sounds like DDoS served on a silver plate to me.
They run these clients themselves and the redis instance isn't publically exposed.
It would indeed be very strange to hope your random users coordinate with your client side load balancer. You wouldn't even have to send real traffic. You could just manipulate redis directly to force all the real traffic to go to a single node. DoSing redis itself is also pretty easy.
I don't think the article implied that the client was for some sort of internal server-to-server communication, or that the Redis instance was directly exposed to the internet.
So no, I don't think they run these clients themselves. If the code runs out there, it's open to inspection.
Either way, you are right to point out that it important to only a try a pattern like this if your clients are highly trusted (or/and have additional compensating controls against DDOS threats). It would be beneficial if the OP made more explicit what their client/server relationships and also flagged the risk you mentioned for general audiences not to go implementing such a solution in the wrong places.
Perl was aimed at intelligent people that considered obfuscated code challenges a sport and had a tendency for masochism.
Its main achievement is being there first, before everyone else, to run server-side code in a scripting/interpreted language.
The rest wasn't neither cultural nor technical. From a purely business perspective, having to fight against an idiosyncratic tool half the time doesn't really make much sense.
Also Redit started as "some BSD boxes" The problem is, when a project scales up, at some moment you will need "commodity" sys admins, so it is easier to just go for linux.
Also as the project gets bigger, at some point somebody will come with the idea to move to linux.
Yes, my best school/university years were when surrounded by people interested in studying. That was the optimal scenario, and everyone's reward was passing grades.
But in work nowadays, the reward is getting paid and promoted. That's not achieved by work, but by socializing, playing politics, creating mutually-beneficial relationships, building empires, and using everyone else.
Which is exactly what happens in a moderately sized workplace today, and one of the main reasons everyone else wants to stay home.
Remote work only sucks if your goals are misaligned with doing the actual job.
Yeah, I understand but it but I wasn't referring to performance only, mostly to "living room PC gaming" in a convenient package, almost like a home appliance. I really hope Steam can pull this off.
You're right, of course, but I don't think blame rests solely on the individual consumer here... I guess it's a bit of a chicken-and-egg problem, wherein Apple makes $200 knitted iPhone scrotes because they know people will line up to buy it, and people will line up to buy $200 knitted iPhone scrotes because Apple made them.
And people have brand loyalty to Apple stuff because quality, or design, or something... but for a product like this, which to me is prima facie a ridiculous, impractical, high-priced, fast-fashion item, you know that the marketers are cashing in on that brand loyalty almost exclusively (in the absence of any intrinsic value).
Half-baked thoughts, I'm sure people have written properly about this. But the conclusion I leap to is that marketing people are the great Satan here. Fuck those guys.
Pretty sure the profit margin for these bags is 10x at least. Way better (and simpler) that dealing with expensive computer/phone hardware and it supply chain, even if their pricing is ridiculously expensive.
Marketing guys just know and exploit a very well known human weakness. It's annoying because it's Apple, but everyone has been doing this forever.
Non-standardized phone chargers? USB-C and its patent hell? HDMI and its licensing? There's plenty of examples for creating wasteful items without them being fashion ones.
The materials themselves probably cost no more than a few tens of cents, so all the cost is going to be the in the manufacturing process. The knitting pattern does look somewhat advanced, so I guess it would require a relatively high spec knitting machine. I suspect what would drive up cost is a combination of throughput and somewhat that you need an expensive knitting machine. Since this is a high volume item that will probably bring down the average cost by quite a bit.
I would guess somewhere in the region of $2 to $5 per pocket to mass produce these? Anyone have a more qualified guess?
The tech industry is basically entirely run on Advertising. Google, Facebook, even Apple owe a huge chunk of their revenue to Ads.
Clearly Ads work. You cannot blame the individual who has been brainwashed, addicted to buying things, by the hyper-capitalist advertising mega-monopolies around us. They are victims too.
I don't disagree and that's factually correct. I'm not sure about calling someone who can spend $200 in an iPhone bag a victim, though.
Plus that kind of wasteful consumerism is only seen in certain developed countries in the world, while the brainwashing happens globally. So corporations are evil but a little individual accountability wouldn't harm.
Pseudoscience like measuring cost to fix a bug in a classroom setting is bad. Specially if it literally does "cost" and "classroom" together. That's just a sad way to grab some more research funding to keep the machine going.
The "garbage pile" of papers is not a new problem. It's been plaguing the science world for quite a long time. And it's only going to get worse because the metric and the target are the same (Goodhart's Law).
From the article itself, each mentioned paper scream of "the author never had to write actual functional code for a living" to me.
> The "garbage pile" of papers is not a new problem. It's been plaguing the science world for quite a long time. And it's only going to get worse because the metric and the target are the same (Goodhart's Law).
I don't think this observation is valid. Papers are not expected to be infallible truth-makers. They are literally a somewhat standardized way for anyone to address a community of their peers and say "hey guys, check out this neat thing I noticed".
Papers are subject to a review because of editorial bars from each publication (that is, you can't just write a few sentences with a crayon and expect it to be published) and the paper should not just clone whatever people already wrote before. Other than this, unless you are committing academic fraud, you can still post something you found your conclusions were off or you missed something.
In such trajectory science is meant to cross the information overload / false equivalence threshold, where the "hey, check this out" scenario won't scale and the cost of validating all other people's papers outweigh the (theoretical) gains.
Not sure if you think that threshold has been crossed already or not.
It's where the pulleys of a very sophisticated statistical machine start to show, and the claims about intelligence start to crumble.
Reason AI is great for boilerplate (because it's been done a million times) and not so great for specifics (even if they're specifics in a popular language).
reply