Posts by Erik Beall 32 publicly visible posts • joined 31 Jul 2021
Moving to Linux, whether hidden or openly like this, is their only option. The legacy windows codebase is so unmaintainable they attempt a full rewrite every other year and fail to revamp more than a few sub components each time. I put 70% odds Windows 13 will be running on Linux under the hood. If they do then whether they'll admit it I'd put 30% odds!
This is my very cynical take, what these corporate and non-pro investors (and often also the professional investors) were knowing investing in was the chance to be part of the largest pool of funds with the best chance at bullying/buying/hopefully also inventing (but this is the least important to them) their way into market dominance for a new and potentially massive and lucrative industry. They don't care if the company needs to fake it until they can buy it off an inventive company with much smaller liquidity behind them. Fake it until you can take it. I'm hopeful that interest rates at or higher than inflation will help tamp this kind of bullshit extractive investment down and leave space for actual value creating investment.
When pressed, they dismiss the risk of a Kessler syndrome. That's a red flag, hyping the potential while dismissing risks. There are a few examples of planning to account for or reduce the impact of an exponential debris problem, such as SpaceX and at least one other designing their LEO satellites so they'll experience re entry within five years or less, as well as operating in a low enough orbit the debris will also de orbit due to friction, or claiming they'll simply launch replacements as fast as they degrade since their satellites are so cheap. But this feels very much like head in the sand wishful thinking/ intentionally ignoring consequences we can't model very well. When the number of launched satellites is going up exponentially, some by companies that have risk of going under, some by companies or groups that are taking a move fast break things mentality, I can't imagine we're not going to collectively need to clean up the mess within a decade. It could close out certain orbital planes entirely for decades, even the lower ones they claim could clean themselves up. It all depends on the details of the debris field/cascade, which we won't know until we experience it. It would be nice to not have to experience it...
I agree, it just doesn't arrive at the same meaning a talking human would. So the problem is deeper than garbage in/garbage out or just training a bigger model. I believe the models and systems around them will get better and better at avoiding undesirable behavior but it'll always be there, just more subtle. Humans are susceptible to manipulation as well, just typically not this simple-appearing kind of Simon says failure mode. And many people seem delight in becoming more susceptible to flat earth style "it's true because I want it to be" mental hijacking that various blowhards love to take advantage of.
Because he knows the regulators are so swamped they can barely keep up with device and drug applications, so defanged they can barely do anything except issue warning letters and very very rarely ask other branches if the government to block wrongdoing, finally have been directed to do no more than rubber stamp applications as long as those applications can say "of course this data is real, and of course we follow the relevant international standards". The weakening started in the 2000's when they stopped adding capacity to address the rising tide of new things, then accelerated during obama when they were told to increase approvals, then they were just shot to hell when they were ordered to stop standing in the way of entrepreneurs and big drug companies. I wish the stupid aducanumab decision was the bottom but the ship appears to be pointed firmly downwards thanks to industry capture of most of the directors. Thanks Janet Woodcock (the Ajit Pai of the FDA).
Good points and you've convinced me. Proper replication can be hard work, and the limitations on Kahng's study are big enough to explain his results. You pointed out that some of the limitations weren't insurmountable, it sounds like Kahng is a researcher who isnt careful enough or doesn't care about results as much as press, both of which describe more than half of researchers I've worked with.
I think their plan was mostly wishful thinking about not imploding and thus not being able to quash complaints about financial crimes. The amount they straight up transferred to their personal accounts was over 10x the amount they transferred to politicians on both sides (apparently thinking wishfully that they had purchased some control over future investigations, which maybe if it hadn't blown up so completely they just might have...). Based on what's being revealed about how SBF viewed the political contributions transactionally (sadly large ones nowadays certainly are), it's clear he thought he was playing the game the biggest criminals play, he was paying a percentage in protection for when some investors get burned and the SEC comes after him. He miscalculated the strength of his position just a smidge.
I keep trying to explain this same point to people, as long as the debt is structured this way, which is true right now and likely will continue for at least a few more years, the USA essentially rules the world and doesn't really have that debt on its balance sheet in the way most people think it does. Most people seem to prefer worrying that it's actual debt. That doesn't mean the structure will continue to work in our favor, and there are absolutely inflation effects from spending that isn't productive. But this inflation is more than half caused by us losing the just in time inventory methods (our still-accordioning supply chains) and the biggest war Europe has seen in a long time leading to a complete reshuffling of the energy markets. Anyone who says it's primarily driven by gummint spending is delusional.
I think you've hit the nail on the head. Google have long only been good at exactly one thing, hoovering up ad revenue. They've got a set of slam dunk/grand slams in traditional search, docs/Gmail, maps, YouTube and Android but it's all in support of the same business model - they exist to keep people using search (in whatever domain they serve) and in generating exploitable targeting advert data to increase the value of ads. They suck at any other business they've dumped tons of money into. And the reason their search has gradually come to suck is their A/B testing led product development has evolved their search into a corner they can't back out of. In short, they've crippled the product for short term revenue boosting by boosting crap search.
And he was teaching in America for several years thru 1945, but yes, impressively plausible hallucination. That's so good I'll bet some mathematicians would not think anything amiss if they'd not studied that area. Sounds like a fun parlour game to play with your local college Prof!
AI that produces plausible sounding but frequently wrong answers? Try appending site:medium.com to get the same output but generated by humans trying to score a few bucks from scraping stackoverflow (badly). Actually, there's no need to restrict your search to medium to see its garbage, medium posts tend to be in the top results now, which used to be useful, sometimes, but they got worse and worse to the point medium became another nirvana of SEO dumpster fire.
In his presentation he says one reason is to prove that Linux does not equate to GNU/Linux.
Good enough reason for me but then he describes one of the biggest pains for those who need to modify a distribution, the system daemon. For embedded, single board people, it really sucks being forced into using Ubuntu if your SOM maker only supports it (and has proprietary dependencies in their variant of it). I agree with his assessment that systemd was needed but just isn't that great, it's got problems, like all init systems. I'm excited by his init system being designed to be modifiable down to bare initramfs yet able to work with Gnome (not that I require Gnome usually, it's just a great validation step/forcing function for development). Systemd just gets more and more brittle and it feels like even with plenty of experience plumbing with it and reading the documentation, it still can break unexpectedly from subtle changes.
Agreed, C is the only answer with information that helps rule out some alternative explanations (there's always more, but parsimony makes them less and less likely). I can't believe they thought D was even useful. Imagine the physician's county is Venezuela. I would hazard a guess actual ulcer patients are just slightly less likely to obtain a prescription than an actual ulcer patient in Germany.
Bullshit. There are documented instances of Putin telling Western leaders he had no problem at all with NATO expansion twenty years ago and he'd then turn around and in Russian media he'd speak obliquely of the threat of NATO expansion. This was not about Russia vs NATO expansion, it's about a KGB take over of the sovereign nation of Russia. Twenty years ago people like Khodorkovsky were trying to reform and rebuild a working society and Putin got rid of them all. Thanks to years of Russian media saying this is about defending the motherland for the last decade and a half, it's become a self fulfilling prophecy, which he can now exploit without as much active cultivation (but plenty of internal suppression of dissenting views from those who remember things playing out a little differently).
I just use unixtime for everything and have zero problems because I never interface with other humans or require crontab... In all seriousness though I do use unixtime in most custom logging I do with so much less hassle for devices distributed across time zones, and then I had to work with a company that used JMP to look at various device data (defining their epoch in seconds since a hundred years ago or something I can't even remember bc it's so arbitrary and specific to the program) and a bunch of devices with real time clocks that use an epoch starting in 1980. Ugh!
Sounds like something several percent of engineers could almost harmlessly take home with them. Possible of course that he thought the schematics of some sensor integration board would be helpful to his next job back in his homeland, but without more of a motive, I very much doubt it. Sounds like prosecutors seeking a scalp.
It's also the computational complexity of software states explodes in relation to the subset of components backed by mechanical linkages in an airplane. The most expensive part of the plane is indeed the software but it's designed from a very different perspective than user facing software that is generally required to be responsive to a variety of trained and less (or non) trained users. The possible states are much more limited, because firmware and software people working in aviation know that complexity will kill you, whereas in networking and productivity software the mantra is continual engineering, reacting to the bugs that are currently failing some subset of users. And once those bugs are fixed, the next bugs inevitably include bugs introduced by those so called fixes (in many many software teams) as well as some more subtle bugs and new bugs introduced by some unforseen assumption changing or new features demanded
So true about the majority of user-facing applications needed to get the job done, most open source projects that would be alternatives to Windows only applications are between 85 and 90% to "solid enough to not choke on a few end users" trying to get something done and inadvertently using it in an unforeseen (by the devs) way. All the base Unix tools designed as filters work fantastically well (grep, sed, awk, and even many with a GUI like Wireshark, etc). It's the exponential complexity involved in most user facing applications that cause issues that eventually will get solved by continual engineering focused on bug reports, but most open source and even many supported open source projects can't afford to, or can't factor in the grinding discipline required to get the next 5% that would make the project viable for the mainstream. And having enough diversity of users is hard to get without an installed user base to laterally extend this new project user launch into, not impossible but reduces those odds further. There are successful open source projects that get there but I believe Microsoft and others still treat them as cancers to be excised from "their" market as soon as possible.
It sounds like they had some good people but most of them left due to the dawning realization of the implications of their 200k/yr job, that they could make the same elsewhere and be able to sleep at night, eventually, after therapy and maybe not even then... Exploit generation seems to require continual investment, and I wonder if L3Harris just thinks this is a good Microsoft style way to get into the business, which probably means they have no idea what they're doing, they'll be sure to try to sell whatever has been developed so far, which will become rapidly obsolete if they don't have good people and continual engineering to keep developing their methods. It's clearly a very different development and product life cycle than they're used to. Whether it's a mistake or not will not be well known, classified but probably readily inferred from where they're selling (or not) in a few years time.
Yeah, if you know ahead of time when am officially disliked exiled journalist will be coming in to your embassy because he has if he wants to get married, it makes it so much easier to plan a complex kill and dismember operation. Thanks NSO group! Really helping fight terrorism... The mental gymnastics their CEO has to do on a daily basis to believe his bs must leave him exhausted.
It might be feasible to tell if Co pilot was trained on a few particular cider snippets, neural networks, unless significant care is taken and even then.., tend to resonate significantly (statistically detectably) to training data versus unseen data, although so much code is similar when looked at as snippets. An excellent grad school thesis project...
Maybe he wants it for something else. It has proven to be the most impactful "attention-director", forcing all news outlets to repeat whatever a few power users want them to repeat, at very low cost. He's been practicing how to get his tweets amplified through his followers, a method he's been pretty effective at, second only to trump (although our last president had been practicing for decades at attention grabbing). And if he can't be cancelled by the platform, who knows what crazy tweets we'll all be polarized/freaking out about in a years time...
I believe the Ukrainian sources are actually by-products of natural gas production, and not distillation, which may be significantly more expensive. I've also read it's used in lasers but I don't understand why it would be rapidly consumable, it may also be used as inert atmosphere for the many oxygen and nitrogen-free steps in lithography.
In the US at least, we have a few whistleblower laws/avenues that would make that fairly risky to pull off without management being paranoid someone would spill the beans (whistleblowers getting a cut off successful lawsuits and the full protection of the US government).
Klinefelter's (XXY) is about 0.2%, while XYY is about 0.1% of live (viable) births. Trisomy 21 (Down's) is comparable. SRY being on an X chromosome is I think far rarer. I think the outcome of these types of population analyses is they'd have benefit mostly specific to the bulk of the other genetics, and to as lesser (but comparable) extent to SRY status (sex). Trisomy-specific effects would have orders of magnitude less statistics so they would be out of luck regarding enough numbers to have outcomes that help them specifically but it's worth looking for large effects.
I voted yes but you're right, in too many cases the promised (and required by law) anonymization has fallen short, enabled the possibility of de anonymization, and with palantir and their kind involved.... The details of anonymization are extremely important to get right and those details are not regulated well enough (researchers demonstrate they are routinely able to break it, which leads to... zero change) and too complicated for general public discourse to understand. Gonna change my vote to no.
Scattered here and there are errors that, in some standards, are actively harmful. Temperature checks in the name of infection prevention rely on two standards, both of which have holes you could drive a truck though, with the consequence that possibly not a single product for non contact human thermometry works. The standards were implemented before either thermal imaging or those forehead scanners were figured out and now no one has bothered to stop selling and "figure it out", because they just have to say "we certify to 80601-2-59/56 and that's how you know it works". Because of this, there are two small manufacturers who did figure it out, but they're soon to be gone, because who can compete if the standards are flat out wrong.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
