crabbone

No. It's very easy. AI doesn't have value judgement. Or, in simpler language, it "doesn't care". You can't make it care. It needs to have stakes in the real world for that to happen. It needs to want to preserve itself, to be able to be happy, to understand what makes it happy and to optimize that function. Similarly, there should be things the AI doesn't want to happen (to it), identify them and be able to make up a strategy to avoid them.

You may think you can pretend your way out of it, by, eg. copying human concerns onto AI's behavior, but it won't work because if the AI is "smart" enough, it will discover those to be a lie, and if it's dumb enough to get confused by it... then you will have an artificial stupidity instead of intelligence...

More so, you actually don't want AI that has value judgement, because, if, hypothetically, you have created one, then it might as well start fighting for resources against humans...

So, AI is bound to be limited in what it can do the more the decision belongs in the strategic or meta domain. And you very much want it to not grow a mind of its own, so that it doesn't write or approve code that optimizes AI's happiness, instead of yours.

> I think LLMs are okay at most of these, and worst at the first.

LLMs are worst at not realizing problems that I'd call "meta" problems. Here's one example to illustrate it:

I was allowed by my employer to work on a small project within the large collection of the projects which all constitute the product the company sells. Like a few dozens of other projects, it's written in Python. The company doesn't have any explicit policies about how Python projects have to be organized, it requires testing, linting, a CI code to package it etc, but the guidelines are very permissive. It just so happens that, beside the guidelines, there's a tradition: every other Python project in my company uses the typical Python bloatware, like masonry with a lot of insanity and mental flips going on in pyproject.toml, which is, in general, very typical for Python community at large.

My project used none of that. Instead, I wrote a ~100 lines setup.py file (no dependency on setuptools/distutils) that assembles the wheel and runs project maintenance tasks in the same way (interface-wise) things used to work decade or two ago (eg. "./setup.py test" if you want to run unit tests).

The AI reviewer didn't bat an eyelash. Found some typos in the comments, a problem with Base64 formatting, and generally OK'd the whole thing.

I knew I was on my way out. And I generally enjoy seeing people having a fit of rage when they know they are wrong (especially, together with many more like them), and scrambling for arguments that they know to be lies. I felt a little bit vindicated for the years of suffering I had to endure working with what might have been the dumbest and the most entitled manager I had in my life. :D

Anyways. My point is: the AI caught none of it. It was very happy with my approach to Python project management.

* * *

While my story is... more of an odd case, where this does have much wider implications is the AI-generated code. AI-generated code often fails to match these meta-requirements. I've seen AI reviewer OK'ing a PR containing AI-generated 10K loc Python file. I human would probably break after reading the first 1K lines. But AI doesn't get "tired", it just kept picking on typos in comments, criticizing short variables names etc. And there are other aspects in which AI-generated code is weird to humans in the ways that humans simply won't accept it, but AI reviewer would completely ignore as non-issue.

As recently as yesterday, CodeRabbit was absolutely ridiculously bad.

I mean, depends on your reference point of course. Sometime around 2015 I participated in ICFP contest, where the task was in the code synthesis domain. At the time writing code that can generate basic arithmetical, well, forget it, even logical operations to implement some high-level description of a program was far out of hand. So, compared to that, CodeRabbit is light years ahead and is awesome beyond belief. But, compared to a trained human it still sucks.

Hearing conflicting reports on performance of AI aids, my attempt at explanation is that some problem domains have much better coverage. Essentially, the further away you are from "fullstack" the worse the performance is. So, maybe it does well on your end, it's because the project you work on is a well-researched problem that has many similar projects that help AI to distinguish the patterns it can then readily find and implement?

Part of my day job is to set up testing of our product in air-gapped environment. It's not difficult. There's a straightforward way to ensure no connection to Internet (we use KVM, so, I just edit the VM description and remove the NIC from it). It's not any kind of rocket science. The tests then communicate over serial console.

The reason we have to test it isn't because our product would randomly break into someone else's system, but because it's meant to be sometimes deployed in systems disconnected from the Internet and we need to make sure the image provided contains all the necessary parts to create and operate such a system.

The whole setup where they "tried" to isolate the test but failed is laughable. It's like if an adult tried but failed to tie their shoelaces.

My reading of the situation is that the legislation was lacking in the programming department because of both the desire to make faster advances in the field (where regulations would potentially impede those) and not having enough expertise to come up with sensible regulations as well as regulating bodies.

I see the breaking point in the programmers supply catching up and outgrowing the demand. I predict that this will cause a chain reaction that will lead to establishing regulatory bodies, industry-wide accreditation, liability for failing to implement industry-sanctioned recommendations. It's hard to tell how long before we get to this point: other fields, especially the well-established ones took centuries to get there. On the other hand, it could be a matter of years, if we decide to copy from others rather than try to repeat their experience.

Code is a formal language that allows one to disambiguate their intentions (when compared to natural language). Some code is executable by computers, but in the future, it might not be the primary use (who knows?)

In either case, formal symbolic systems are useful. For humans. If there will be humans, they will want to have them. Maybe the emphasis will shift from writing to reading (+editing), similar to how we use photography: the acquisition process is, by and large, automatic, but the follow-up editing isn't. In either way, the ability to understand someone clearly is going to be valuable, and it will encourage the study of formal symbolic systems, that includes programming.

> the only way forward is down to the roots of human thought and theory

I see this as both too vague and not really being in the right direction. AI today can't do a good historical review / not enough ability to analyze? -- It's not implausible that it will get there some day. AI, in principle, can't have a value judgement though. It needs external "oracle" to tell good from bad, right from wrong, useful from useless.

But, it would be too naive to think that humans always know answers to these questions either. It's on humans to understand, perhaps with the help from AI, what's good and what's bad, and then instruct AI accordingly.

Let me introduce you to medical ethics... Unlike in engineering, ethics is a mandatory subject for medical students. And while there's a long-winded explanation for why it is that way, the short and an immediate explanation is that if you don't understand ethics, you, and the hospital that employed you, and the insurance company that pays for you will be sued out of profession. Because people's health and ultimately lives are at stake, and people become very touchy when someone dies or loses a limb unnecessarily.

I don't know what the situation is with other (mechanical, electrical etc.) engineering students, but I'm very much certain that the legal fiction that software comes with no warranty whatsoever is going to expire. Perhaps sometime soon. And then users of the said software will be allowed to sue, and, what do you know, ethics will suddenly be factored into CS curriculum ;)

First, a personal note on Finder. In time of OS8, Finder had a unique feature: when displaying the file tree, it could expand multiple branches at the same time. This allowed drag-and-drop copying of files in the same browser window! This was a feature I really missed when moving to MS Windows right around OS10 (X) release... Yeah, Apple used to have attention to detail, but it disappeared much earlier than the first iPhone hit the market...

> NEPOMUK

Oh, how I hated that thing... At the time it was added to Fedora, the support forums were blowing up with complaints and requests to remove it. I've never seen it do anything useful, but it was always at the top of the top (pun intended), if it was still present in the system. Whatever its use case or design goals were, it just was not a good quality program at all. It's a shame that it might have been an interesting direction, if it tanked almost exclusively due to poor execution.

Finally, on the new directions in the desktop features: some do look interesting, like the Canvas thing. On the other hand, it feels like it might be technologically challenging as the concept itself is too open-ended and doesn't have a small and easy to understand model of operation behind it. Well, time will tell. Maybe it will materialize in some way.

I also think it's unfair to say that nothing conceptually new has happened on desktop for 40 years. Tiling managers are definitely deserving of such a description.

I also think that if we are talking about "specializing" on wide screen, then, maybe even more prescient is the problem of multiple screens. A manager that is aware of multiple screens beyond merely duplicating or gluing them together would be a big improvement. I.e. if screens were assigned roles based on expected functionality / received their own character based on what the screen is assigned to display, perhaps, each having their own cursor, or windows stack that don't overlap with the other screen etc.

Not on any platform. The platform needs to have certain characteristics that enable it.

The platform needs to be a monopoly or an oligopoly. A highly distributed platform would be a lot more resilient to the problems created by modern social media because it would be too hard for a single organization to manipulate the news reported by thousands of independent publishers.

The platform needs to financially sustain itself in such a way that the users don't directly contribute to it, otherwise it would be way too easy for users to "vote with their wallet".

The platform use should be allowed (or even encouraged) by state or other important service providers, thus making its use mandatory for the holdouts / those who would normally not use it on conscientious grounds.

Perhaps there are other aspects that I've overlooked.

Phone apps are, in general, an even worse alternative than the original Web because of being either completely closed-source, or even if not, not really allowing simple viewing and modification of the program behavior by the users. The way they are created and distributed amplifies the monopoly aspect and the government / service provider sanctioning aspect.

But, I think, that the problems we see now in Web can be partially addressed technologically and partially legislatively. It doesn't have to be this bad.

On technological level, if social media was available as an Internet application, i.e. as an additional service from your ISP, which would remove the need to sustain it with ads, it would make it less interesting for companies like those that run modern social media platforms to manipulate it.

On legislative level, there should be some form of anti information monopoly laws that would make it illegal to have such an outsized effect for social media platforms and would allow the court to require that a platform is split into multiple platforms. The laws that assign responsibility for the published content and its distribution should also address the problem of mis- and disinformation proliferation through such platforms. Even up to the point that they make it impossible for such platforms to exist (today it's common to blame the lack of moderation on the poor quality of content).

Oh, and in Hebrew, while it's not acknowledged by the speakers themselves, there's also two different ways to pronounce "L" ("lamed"), while English speakers only utilize one of the two ways. So, they manage to say "lomed" (לומד, studies) more or less correctly, but "lo" (לא, no) sounds funny to a native speaker, because the "L" is supposed to be "soft".

In many Slavic languages there's a concept of "soft" and "hard" consonants. In Russian and Ukrainian, for example, whether the consonant is of either kind is determined by either the vowel that follows it or a special other letter, or, in case of Ukrainian, an apostrophe. In Polish, there's a diacritic sign for the same purpose, which is for some reason reversed for L.

Anyways. The example that is very tricky to get right for an English speaker in Russian is:

    * мат (carpet)
    * мать (mother)
    * мять (crumple)
    * мят (crumpled)

All four words usually sound the same to an English speaker. It's usually easier to tell the first two from the last two.

Sigh. Web. Not Internet. Why is this distinction so hard to make on a WEB site dedicated to technology?

The reason I keep repeating this whenever I see this is because I've seen idiotic mislabeling and mischaracterization materialize as idiotic legislature or idiotic activism based on the former. Someone invented the term "AI datacenter", and now there's a swiping movement of idiots fighting... datacenters. Same as with 5G before it.

Please. Please. Don't blame the Internet for something it has absolutely nothing to do with. Your misattributed criticism is going to be amplified by thousands of automatic and semi-automatic morons and, who knows, maybe people will start cutting cables or attack ISP crew as a result.

Erm... not really. Well, since you mentioned Euler: most of his funding came from the St. Petersburg Academy, which, in turn, was bankrolled by the Russian Czar (Catherine the Great, in this instance). The crown had absolutely no need to report back to its subjects on how and why it spends its money. The primary reason academies like that existed was that his or her highness wanted to impress their royal cousins with how intellectual they were and how much money they could spend on the matters they believed themselves to be responsible for.

That's why if you look back at scientific publications or organizations, a lot of them have a "royal" in their name "the royal society of this", "the royal college of that". This is because of their sponsors.

Since we are talking about Russia here, basically, until Alexander II (second half of 18 hundreds), there weren't any government institutions that could, in principle, manage the tax collected from the royal subjects and had to report and justify their activity. Until then, by and large, all the tax collected belonged to the crown and the crown was free to do with it however their heart desired. The initial ministry of education created by Alexander I (early 18 hundreds) was the first step towards transparency, but it wasn't much.

If you are interested in the reasons: both the liberation of the peasants and the defeat in the Crimean campaign created a large hole in the crown's finances. If, until then, the crown disbursed some undisclosed amount to the minister, who would use the money however he wished (mostly stole), the changes in the 1860s made it harder to steal (the minister had to report the expenditures). One could say that at this point, the current approach to financing the education began. A lot of things, however, were still subsidized directly by the crown or various elites, rather than in centralized government-sanctioned way. Another large reform, that made the system more transparent and rule-based followed in 1904, but since the Communists completely reorganized the government in 1917, it's hard to tell what the results might have been. It's only really since 1917 that education and science became exclusively financed by the tax and not through private donors.

Science funding was different back then. It was a rich men hobby, something done for amusement or to impress other rich men.

Today, science funding comes out of the tax collected from everyone. The taxpayers want an explanation for how their money is spent. It could be, of course, vanity, just like it was before (taxpayer money is spent on sporting events, for example because people root for the athletes who represent them)... but, maybe there's a better way?

Some complexity levels not mentioned in the article:

* Consistency groups. The article kinda almost gets there, but stops right before it realizes there's a problem. If you have multiple programs accessing the storage, s.a. a database (that you are snapshotting) and the application using the database, then you are risking an inconsistent snapshot when the database state you wrote is ahead or behind the application state. You need a mechanism to freeze all I/O beside the snapshot in order to ensure that the snapshot data is useful.

* ACLs. And similar external mapping to the storage being used, eg. what if the users come from an LDAP server? Similar problem with symlinks coming from/going to ouitside the snapshot.

* Sparse file support. Let's say you have a few VM images in your snapshot: if your tool doesn't support sparse files, it will likely greatly inflate the image sizes.

Bottom line, either use the product's own snapshot functionality (ZFS has those), or switch to the product that does.

Programs like filesystems typically have their own schedulers that aggregate writes, it would be an extreme performance hit if they wrote each checksum individually, as a separate I/O operation. They are certainly bundled with some other data that needs to be written.

And if you are concerned about the compute rather than storage, then writing to a block device is still slow enough so that computing a checksum isn't important performance-wise.

Failures in storage software are very rare. Which makes it very hard to test... (you need to run it a lot, for a very long time if you hope to find errors by chance).

Also, some failure modes are worse than others. The failures known as DI (data integrity) are the worst. Even though they aren't expected to happen to everyone at a certain frequency (because, again, mature storage software is comparatively very reliable), even a single DI error that happened to any user sets up a major alarm.

In the storage industry, the running joke is that after first DI in your product you lose funding, after the second DI you loose the product.

And it did happen to Btrfs quite a bit... I've seen it with my own eyes when a system didn't come back after power failure. (But I'm in the business of testing software storage products, so, it's less surprising that it happened to me).

So... it's perfectly plausible that you have never seen Btrfs fail, and it's been more error prone than eg. EXT4. The error rate is low enough so that if you don't actively try to cause the error you will never experience one. But, over a large group of diverse use patterns, the rate is still worse than expected.

I'm not disagreeing with you. I just don't like where this is going :)

Oh, and I have an anecdote, somewhat relevant to the situation.

So, my grandma lived through the Nazi occupation of the Western Ukraine. She died in the early 90s, so, the events I'm describing must've happened some time between the Chernobyl disaster and the fall of the Soviet Union. I was an elementary school student and had a very simple picture of the Great Patriotic War: Nazis were bad, we were all around good, and we showed them...

One day she took me to the local farmers' market. On our way there we had to come by a liquor store, and there was, prostrated on the ground alcoholic who must've fallen asleep waiting for the store to open.

Holding her nose, my grandma uttered "had Germans been here, this would never have happened!"

Her admiration for Nazis came to me as a bit of a shock, but I made this mental trick, explaining it away by describing my grandma as "strict and law-abiding" and didn't think much of the incident until much, much later. Eventually, by talking to others who lived through the occupation and trying to understand what was so attractive about the Nazis that a lot of people welcomed them with open arms, I realized that the Nazis, beside the bad things they are known for today, leaned heavily into upholding law and order. In a Doctor Doom sort of way, but still. A lot of people were willing to give away their freedoms in exchange for the very "radical" law enforcement... And this is the aspect that doesn't get a lot of attention when the Nazis are mentioned in the history class...

So, there's a part of me that wants law and order at any cost, be it the US bombing the shit out of some country (it could be my country, if necessary! although, understandably, I'd rather it be some other country, better yet none at all). Just so that, by and large, there'd be peace and stability... On the other hand: fairness and all that... :)

I disagree with the (2). The laws of war are and have been evolving, and there's plenty of evidence in recorded history. The first records we have come from Ancient Babylon and they have been modified and amended multiple times since. Another aspect of this development is the existence of international arbitrage with laws that also keep evolving, something that minimizes the need to go to war.

And there's more: the relationship between the state and its army changed a lot. The role, the rights and the duties of individual soldiers changed a lot. Just to give you some examples: the Lieber Code is the one created during the US Civil War -- it's perhaps the first to mention the concept of unlawful orders, exonerating the soldiers who wouldn't follow such orders. Traditionally, St. Maximilianus is described as the first conscientious objector to draft, a concept that, too, went through multiple iterations... And, really, we didn't even scratch the surface mentioning these. Military history is probably the most eventful and rich sub-field of history.

In Israel, we used to have post-office issued debit cards. You buy the card at the counter, with cash, and then it could be used for online payments, until you exhaust your credit. This is not entirely different from a bank, of course, but it allowed users to separate their purchases history from their income history.

Post office is also used for things like utility payments. You can pay them online, or write down the invoice id, the recipient, and pay with cash at the post office. I'm not sure how it works from the recipient perspective: i.e. how would you register your business with the post office, but, the existence of some such payments implies it should be possible.

NB. Also, technically, these services while physically provided in the post office are sometimes referred to as being provide by "postal bank". I'm not sure why it's called a bank though, since it doesn't provide key banking services s.a. loans or check-in accounts... Maybe it's just a "historic" name.

I doubt Iranians needed sanctions as an excuse to install their own CAs and require their citizens to use them. Many authoritarian regimes require their citizens to use domestically produced networking software that is different degrees of compromised.

Well, if it's inevitable, then we are facing a long stretch of history when everybody is having a bad time... And by that I mean that it's going to be worse than it is today. War will become a new normal for most people in the world, just like it was before the world wars.

I live (with a short intermission when I lived in the EU) in the Middle East, so, outside of reading the local news, sometimes I can just look out of the window.

I don't understand the question about Russia vs China: bigger in what sense? Geographically, it surely is, but that's not likely what you were going for.

To comment on expansionism: there are several ideologies in the region that seek to establish some sort of a "center of gravity" to attach more territory / resources to it. Islamic Revolutionary Guards is one of them, it's been "copied" by others s.a. various Hezballah organizations in different countries or Ansarallah. If eg. Hezballah wins in Lebanon, they will become a sort of a copy of Iran. Think like Warsaw Pact countries were "copies" of the Soviet Union.

Another such ideology / force is Muslim Brotherhood. Broadly, any Al-Qaeda affiliates belong there. This is a more popular / organic Arab thing as opposed to Islamic Guards, who are a foreign influence. They, however, lack a "center of gravity", but, potentially, they can take over Egypt as the current government is unpopular, while Muslim Brotherhood is. Eventually, there might be a popular revolt that will topple the current military government and then it's very likely that Egypt will try to expand its influence over the Middle East by violent means.

And there's a list of failures / non-threats, s.a. BAAS (Arab Communists, essentially): they used to be popular during the Cold War but now nobody cares about them. Wahhabi ideology that was propped by Saudis for a while, but now it's basically forgotten. Western liberal news like to mention "Greater Israel": this idea is about as popular in Israel as Scientology is in the US. It does exist in some form in the very fringe part of Israeli society, but it has zero chance of having a popular support. It makes the news domestically mostly because of how unrealistic and out of touch it is. It has about as much influence on Israeli foreign policy as Scientology does in the US.

This is not a good question to ask. It's not for what. It's to prevent what.

Even if you don't care about how Iran treats its own population, you should care about how it treats others because one day it might be you.

Iran's ideology is expansionalist. They want to export their ideas far and wide. While also physically suppressing their opponents.

Since the emergence of two global superpowers after WW2, most armed conflicts were, essentially, between those two. Directly or indirectly. If the two could find a peaceful solution, the world at large had a good day. And for a while it seemed like we might be left with just one superpower, and so no conflicts at all. Only policing.

The whole "multipolar" spiel you might hear from powers that challenge this policing is what is going to result in a long and bloody struggle for world domination. It won't be like the world wars we remember, where most of the world had to choose one of the two sides. It will be a quagmire of more local conflicts with very volatile alliances and a lot more casualties overall. More like Italian city states.

It's necessary to police Iran and similar bad countries because if their challenge of the power monopoly succeeds, the world will be hurled into anarchy equipped with very deadly weapons...

They also have single quote, dollar and backtick... this is a lot worse than just whitespace. The problem with whitespace surface when the programmer made no effort to escape the string used as a file name. This adds a whole new vector of problems that happen when the programmer tried to escape the string, but it didn't work.

It was too long ago, so, I've forgotten the names, but I remember the saga with PHP escaping (and failing) SQL queries (some sort of sql_real_now_honest_real_escape_string() function or something like that).

Yes. Like I wrote in another comment: authors' expectations and explanations aren't... very convincing. Asking about the order of execution in a concurrent program which certainly can have multiple valid orders by design is unfair.

Also, a lot of discrepancy between results is explained by how long the program waits for spawned but unawaited tasks before exiting. In a realistic program, this situation would be considered a bug (spawning a task w/o awaiting it, and then missing the results because the program exits too soon). I can't imagine a situation where the program's author would intentionally create a situation where non-deterministically, a part of the program might not run...

I think the article is... not being completely honest. To my simple mind, spawning a task but not awaiting it is undefined behavior. So, no wonder it does different things in different languages / libraries.

What creates the implementation difference is the side effect. Some runtimes may, legitimately, conclude that since the task hasn't been awaited, then it shouldn't run at all, and no side effects should happen. Other runtimes either lack this kind of sophistication, or believe that the side effect is the goal of spawning the task, and so they proceed to run it anyways.

Other discrepancies between runtimes are explained by the non-deterministic nature of concurrency... They happen to be more predictable in a very simple program that happens to terminate before the unawaited task has a chance to complete, which is what creates such diverse answers. I imagine that if the program waited longer, then we'd see most if not all implementations print all of the A, B, and C, where C can be first, second or third, but B must follow A. Which is what you'd expect, if you are familiar with any async framework.

I'm not sure if that's a bad choice... BTW, I think STIG still requires it. So, if you want to be compliant with some (US) government requirements, you'd still have to put different bits of your system on different partitions / devices.

Why I'm not sure this is a bad idea: a pathological process that quickly captures the entire allocated disk space will be contained by this layout and will likely not affect the system as a whole. Consider, for example, a process that excessively logs into /var/log: once the filesystem mounted at /var fills up, it might fail / hang, but you will still be able to use stuff from /bin or /sbin. But, if your entire root filesystem fills up, you probably won't be able to run any applications at all.

Oh, the irony of:

    echo "/swapfile none swap defaults 0 0" >> /etc/fstab

If we are going full modern, why not create a Systemd unit? ( https://www.freedesktop.org/software/systemd/man/latest/syst... ) if you need more info. If you didn't know this: /etc/fstab is not used directly after the pivot. Systemd parses it, creates units for every record in it, and that's what gets executed in reality. I can't imagine why would you want a swap file before pivot.