Posts by doublelayer

BOFH: Boss's quest for AI-generated program ends where it should've begun

Re: you just need to ask AI to sum up the numbers

I'm not sure it was meant as a serious suggestion. However, if you don't know enough to write your own sum function, asking AI to do it for you will generate something that might work or might not and you don't have the expertise to figure out where its problems are. If it produces something that crashes immediately, that will at least be clear, but if it works on some but not all inputs, reading the code is the most reliable way to find that out before you rely on it. As such, asking for a program that you can't test is no more reliable and potentially more dangerous than just asking for the answer.

This is even more the case because the spec for the program was vague as always. Summing numbers isn't that hard (except for precision, overflow, speed, or various other more advanced cases), but getting them out of some system, into the program, and the result into a desired location is trickier. If you ask anyone, human or LLM, to write a program that sums numbers from a spreadsheet, most of the code will be related to getting the numbers out rather than summing them. With more potential sources, the program the boss asked for basically can't be written without clarifying where the numbers are going to be, in what format, and where the result should go.

So even though it was meant as mockery, asking for the result at least eliminates all the parsing and retrieval code, which means that lots of possible bugs are not going to be there. As LLM companies improve them, straightforward mathematical tasks will improve. That's not because LLMs will become more intelligent. It is because these companies like to hide the places where their models screw up, and this one isn't the hardest to handle properly. Figuring out whether a calculation is needed and then running that calculation outside the model is already done in some products. It's fragile, so if you need anything complex done it's likely to stop working, but for simple calculations of the kind you could quickly do in your head or on a pocket calculator, it means fewer embarrassing errors.

Techie took five minutes to fix problem Adobe and Microsoft couldn't solve in two weeks

Re: I've not really used Windows much for 15+ years

I don't think so. Windows wouldn't finish installing until it had restarted the correct number of times, and it would manage its own restarts and run the needed ones. I'm pretty sure that what happened was the person installed Windows 8 but was not aware of the fast startup option or how it worked. Then they installed an Adobe product which needed to restart and probably offered to do so at the end of the installation. Probably if they had let it do that, a normal restart would have been executed, hence why Adobe hadn't caught this problem while testing. The user, however, needed to save some work so didn't let the installation restart for them, and they just shut down. Adobe hadn't labeled the need for a full restart, so shutting down and powering on wasn't good enough for them. Therefore, the application didn't work because it still needed a proper restart.

That's an unusual config. It suggests that your machines are set to boot to USB devices before internal drives, which is atypical and probably not the best choice. Every machine I've used in the past twenty years has required that a USB drive be manually selected to boot to it by default, a default I haven't and don't want to change. One of the major benefits of this is that there isn't much malware which bothers to make external disks bootable infection vectors because, unlike in the age of floppies, the authors know it's not going to be booted to anyway.

Copilot's crudeness has left Microsoft chasing Google, again

All the time. It isn't just "Alice suggested we redesign this and people smiled, so we're going to", but if the people concluded that this made more sense than Bob's proposal to change the training documents and leave the code alone, then the notes will say that. If the redesign the code suggestion came up first and was discussed for three minutes, and then the documentation proposal was made and discussed for twenty, and then people said that they thought the training solution was less helpful than the code one, will Copilot properly understand that two thirds of the meeting might have been necessary to decide against something but does not need to be brought up in the next steps email? I don't know, but I wouldn't want to count on it.

Dutch cops reveal takedown of 'world's largest dark web market'

Often, they're referring to Tor hidden services. Not always, and when they talk about domains being seized it might not be, but quite often that is where the illegal markets are set up. If they're doing it correctly, there is no IP address you can enter to arrive there. You have to go through Tor and use the domain name, which is really a public key, to find and connect to the site. It gives people some of the tools required for anonymity, but if you're not careful, you can mess it up, and you also need to do more than just have a hidden service to not be found.

You have issues with 'Issues' always being called 'Issues' in Jira, so Atlassian now allows them to be called ‘Tasks’

Re: Terminology

A ticket is a weird term, although that's hardly going to stop them given all the other weird terms that get used. However, it doesn't really work well for organizing them. A ticket containing a ticket of tickets is not a metaphor that's going to work very well. Of course, you could just have groups containing items or other groups, and if you want you could call the items tickets. If you're only going to have one word for everything in the system, who cares what that word is?

Uncle Sam may force Google to sell Chrome browser, or Android OS

I think that most people, if presented with the search engine screen, would choose Google from brand recognition and inertia. Google is willing to spend billions to prevent that from happening. Either they know better than I do about what people will do with a choice, although those people can still go into the settings and get a choice screen with common engines listed, or the team that does this stuff has a massive budget and no oversight.

What would be the logic of that? I can see a case for Microsoft: if they own both the cloud service and an operating system you might run in it, then they could do what they have to make it more expensive to run Windows in someone else's cloud. But that doesn't apply to either of the other two. Having an online store has no direct conflicts with running servers, and unless they did something like forbidding people to sell stuff in the Amazon store if they used a competitor's cloud (they haven't), there's no reason to break along that line. They do plenty of similar things, but related to things like whose shipping network you use, not whose cloud servers. Likewise with Google Cloud and Google Search.

When you make a case for breaking up a company, you usually have to describe how having positions in those two lines causes a conflict. Each company has plenty of complementary businesses for which you can easily make that argument, but that doesn't mean that every pair of services works that way. Browser and a search engine: yes. OS and app distribution: yes. OS and browser: yes. Search engine and app distribution: probably not unless they were intentionally keeping F-Droid out of search results, but you could try. Cloud servers and a mobile OS: I don't see it. These are important because you need to prove that an abuse exists to split them at all, and the lines where these conflicts exist are where the breaking needs to occur to stop the negative results.

Bitcoin creator suspect says he is not Bitcoin creator suspect

Re: Imagine

Not the known ones. It would have been quite simple for him to have had a bunch of them, some of which have been transacted over and over again while others stayed quiet. He could have earned no real money from any of his bitcoin or he could have cashed out quite a bit over the years, pretending to be an average miner or trader. We don't have enough information to prove which.

Microsoft veteran ditches Team Tabs, blaming storage trauma of yesteryear

As others have noted, nearly every editor has a setting that sets what the tab key does. Press the tab key and, instead of inserting a tab character, it inserts however many space characters you told it to. Press shift+tab and it removes them back to a multiple of the number you set. I use spaces, not because I think spaces are better, but that others think so and I'm willing to go along with them, but I do not press the space key to insert them one at a time. I press the tab key to insert them in chunks of the appropriate size. For the same reason, my editor is also set that, when I press enter to move to a new line, it fills that new line with the same amount of indentation as the last line. Otherwise, I'd have to press tab to indent every line manually and I don't want to.

It depends how you like your code indented. If, for example, you want a multiline function call to line up the first letter of the second line with the parenthesis of the function call, or in some cases the first character after that parenthesis, then that parenthesis might not be on a tab boundary. My solution to this is to not align that letter and that parenthesis, and I'm using spaces. I just can't be bothered. However, there are people who want those things aligned and if they get their way, then spaces are the only option for doing it.

Linus Torvalds declares war on the passive voice

Re: I'm British...

So do many options involving the active voice, though I grant not the example I used.

I fixed the bug in the thing [in a commit two weeks ago].

These changes [but not this one] fix the bug in the thing.

I contend that this is not really a concern that has much relevance. If people are putting in commit messages that do not correspond to what the commit actually does, how they phrase it is not the problem. Anyone doing this is either deliberately dishonest or getting commits confused, and neither is going to be fixed by adding a subject picked from a short list to the sentence. I also wonder where my most typical form would fall. I tend to write commit messages in the active voice but without a subject ("fixes a bug in the thing"). Yes, those are fragments. It hasn't caused a problem so far and sticking "this" on the front won't help.

Re: Linus non est super grammaticos

I'm not sure that the passive versus active voice is going to change how easy change notices are to read. This is a style difference that I don't much object to, but neither do I see it having much benefit to anyone. There are lots of similar style recommendations which I mostly go along with as long as they don't do too much harm.

In the spirit of tiny grammatical details being considered more important than they are, a company for which I worked had a style checker which was particular about the mood of the verb at the beginning of a comment documenting the purpose of a function. I tended to write my descriptions in the indicative mood ("Deletes all the data"). The checker would reject this and require me to phrase it in the imperative mood ("delete all the data"). I don't know why that s was considered so harmful, but it was clear that someone thought so. Some colleagues found the message, which simply demanded the imperative mood, to be confusing. While I did learn the various moods, I'm not sure they're considered necessary education everywhere and so the style check was not only making an issue of something that nobody really should care about but wasting programmers' time in the process.

Re: I'm British...

Opinions on grammar are highly variable and subjective. I see little difference between "the thing was fixed" or "this fixes the thing" when the words are clearly linked to a commit. If they were separated from the code they changed, then theoretically it would be harder to understand, but they aren't and, if they were, the suggested changes wouldn't work either because "this" wouldn't refer to anything specific and could be a certain commit, a batch of commits as an update, or who knows what else. Unless they were requiring that the messages take the form of "commit 28ac9307 fixes the thing", the form of the sentence makes little difference to me.

Musk's Starlink rockets to 4 million subscribers

Re: Can bots be subscribers?

That is true, but a slow connection has to be really slow to make that happen. Most of the time, electronic payment takes a few API calls, none of which are particularly large. Latency can be more important than bandwidth, but neither is very important. The main way for a connection to be the limiting factor is if that connection is unreliable and is constantly dropping packets. Then you can expect the transactions to begin failing. Unlike many uses, most payment terminals would not have any problem processing payments much faster than the people can. Of course, if you posit a single 128 kbps line shared by fifty shops, it could encounter that problem, but that's probably not the kind of alternative that something along a major road has access to.

American Water rinsed in cyber attack, turns off app

Re: It's 2024 ...

"unless that malware is either self-starting (in which case it'll become immediately obvious that something is wrong)"

Not necessarily. Time delays are really easy to build in. They can let you clean up your first exploit before things go wrong so the investigation takes longer or, if you put more time into it, you might be able to pretend that your sabotage is just a cascade of normal failures that happens at an inopportune time. If your aim is destruction rather than getting data, you can manage it even through an airgap that you can only bypass once.

Re: It's 2024 ...

I did read the article. It is not as definitive as you say it is. For example, the sentence you quoted could have a lot of explanations. Here are two possible ones, neither of which you can disprove from the text of the article:

The company currently believes that none of its water or wastewater facilities or operations have been negatively impacted by this incident because they were airgapped, so you'd have had to do something quite specific to get at them, but they don't know for a certainty that that didn't happen. Sure, they're running now, but if some theoretical attacker inserted malware into them which is going to switch them off in three days, they haven't found it yet.

The company currently believes that none of its water or wastewater facilities or operations have been negatively impacted by this incident because the PR employee checked and water is still coming out at least in their house.

And the last few paragraphs are not about this incident, but about one last year. Yes, in that case, the people concerned did connect their systems to the internet, but that doesn't provide that these different people also did so.

If Dell's Qualcomm-powered Copilot+ PC is typical of the genre, other PCs are toast

Re: "Whatever x86 apps I threw at it just ran. Swiftly."

As I understand it, drivers are not emulated, but any other application is. So your drivers will need ARM versions, the same way that 64-bit drivers were required for 64-bit Windows. However, it's pretty different from Windows RT. You can take an application that doesn't hook directly into the kernel and run it under emulation. It shouldn't matter what that application contains, even if you hand-coded it in X86 or X64 assembly. So probably your custom driver systems are going to need an ARM driver to work, but you've still assumed too little compatibility.

Busybox 1.37 is tiny but capable, the way we like Linux tools to be

Re: busybox linux

The post they replied to linked to this GitHub repository with instructions for building a minimal distro using mostly just busybox, but those instructions do indeed involve installing bash too. As far as I can tell, not for any reason other than preferring bash to busybox ash, which I certainly do.

It goes on to also demonstrate adding nix, xorg, and tools for WiFi, but those are optional. The bash step is not written separately so even if you stop at the first bootable thing, the system they specify has bash in it. You don't have to do that.

That might work if it just packaged them, but since it also implemented them, maybe not. Not that I care. I don't think that "do one thing and do it well" was meant as an unbreakable law rather than some advice for an approach that often helps.

Two things, neither completely justified. The first was that, until version 1.3.0, its license was GPL 2 or later. As of 1.3.0, it became GPL 2 only, same as the kernel. Before that, people didn't like the ambiguity involved in letting whatever might show up in GPL version 3 (then unreleased) to be a core part of their system. This was only partially justified because the clarification came in 2006, so it's not like there was much need to deal with it.

The other reason was even less justified, namely that Busybox was a project well-known for pursuing their license rights. If you used Busybox, they would demand the source be released and sue you if you didn't. This rarely if ever gave the community useful code because anyone using Busybox didn't have much of a need to improve it, but it did certainly penalize people who don't comply with a really easy license. This shouldn't really be a problem, and it really wasn't, but some companies can't be bothered to release the source code they haven't changed and contains no secrets. I'm not sure why they find this so difficult, but having tried to get source code from them, I know they exist.

In general, one of the maintainers of Busybox got annoyed at another one and went off to create Toybox. This was in 2006, so before Android really went anywhere. Android had to choose between the BSD-licensed Toybox, run by someone who had added a lot of stuff to Busybox, or the GPL2 Busybox which had lost a maintainer but still had more people working on it. There wasn't an obvious good and bad answer, and it probably isn't that surprising that they went with Toybox. Having done so, Toybox has focused on keeping itself functional for Android, so it's not surprising that they're still using it.

Three and Vodafone: We need to merge because our networks are rubbish

Re: Only 73.03 percent of rural premises in the UK have 4G coverage (from Three)

As a non-UK resident and someone unfamiliar with these companies, I would expect that, in addition to many negative consequences, that would actually happen. That wouldn't stop them from raising prices, canceling deals they no longer wanted to offer, or any number of other problems that arise when competition decreases. However, moving the equipment around to increase coverage is a relatively cheap thing that increases their potential customer base. Buying new equipment to serve rural locations is expensive. Moving equipment isn't, and it's often older equipment that they don't need in urban areas.

Embattled users worn down by privacy options? Let them eat code

Or in other words, it doesn't work. Its failure to work is not because the technology is wrong. It's not because people can't use it or servers can't direct it. It's because those who would have to honor it are not and those who would make them are doing nothing. If they made an API that automatically filled the cookie selection boxes, that would help, but any company that wanted could still set whatever cookies they want. They could say that collection was a strictly necessary cookie. They could say that there was a mixup in their code that ended up setting the wrong cookies. They could set a cookie which was meant to indicate that data collection isn't desired but happens to work as a fingerprint anyway. The suggested API is little different from the do not track signal in that it has no technological enforcement of compliance, if such a thing is even possible, so without strong enforcement from somewhere else, it won't make a change. If you had strong enforcement from somewhere else, you could use the DNT setting, this API, or any similar indicator.

159 Automattic staff take severance offer and walk out over WP Engine feud

Re: People have many conflicting loyalties in life.

That is not entirely correct. Mozilla's guidelines are on modifications rather than binary source. If I take the source code for Firefox and compile it unmodified, it is Firefox. Mozilla is fine with this, and so would a court. To some extent, they're also fine with some modifications, though their preferences aren't entirely clear. Rebranding seems to happen when changes have been made to the binary, for example removing some advertising. Even in that case, they could easily say that they are using Firefox with some changes by the distribution and be fine from a trademark situation because they are. They choose not to because Mozilla would prefer they didn't, but they don't have a legal requirement to. Different rules might apply to trademarked logos, I'd have to check the license, but the name wouldn't be affected.

Re: People have many conflicting loyalties in life.

I don't know how they were using them, but to some extent, this is normal and expected. If you're hosting Word Press, then saying that is stating a fact rather than advertising on a trademark. The "Core WordPress" name could easily be permitted on that basis. They could have used the name in other ways that would be more difficult to justify, and it wouldn't surprise me if they had, but that name doesn't sound like one of those. WooCommerce is another story and could be more a clearer violation.

As for why they might change these, it could easily be that they hoped that making those cosmetic changes would make this lawsuit go away. Not fighting this means less trouble for them even if they think they are guaranteed to win, not to mention that they do benefit from Automattic's work and don't really want to see them destroyed. They don't have to think that their actions were illegal to justify stopping them.

I'm not sure. At the start, this seemed like the old classic "I made something open source [with lots of others' help] and I want to be paid by the people who make money off it. Why can't I force them to do that?". To add a bit of extra spice, there was something a little legally dubious about trademarks thrown in. However, the way it's played out isn't following the normal pattern. There are lots of people who host Word Press for profit. They all say "Word Press" on their websites. Some of them use it in advertising that I'd rather not see anymore. None of those have been sued or had their access to update servers cut. Either this is really about the name of the company, in which case it's not going to go well for Automattic because "WP" is not the same thing as "Word Press", or it's about something else but I'm not sure what it is.

If there's an actual legal thing in here, I don't get it. I'm currently operating on the theory that it's really just a personal whim that's going to go badly, mostly because that's the only explanation that gets even close to explaining most of the events. Maybe Mullenweg hates someone who works at WP Engine. It would make more sense than most of the rest of this.

There are always some tricky aspects to something that was formerly entirely charitable switching to a partially charitable and partially for-profit enterprise. In the case of the Raspberry Pi, not even this happened as the foundation set up a for-profit manufacturing company back in 2012. However, it is not the same as this case. The private company was majority owned by the foundation until it was taken public, and is still 49% owned by the foundation. If it was entirely owned by Eben Upton, this would be a different situation.

Re: Well...

If Wikipedia can be trusted, they had 1733 employees at some point in 2024, so this is about 9% of them who left. That doesn't explain how many of them were working in each area, though. It's still not a good look.

I'm also curious why he offered an exit package like this in the first place. There's something a little weird about the speed of that and the terms provided. The person who worked for two days and is getting six months of salary is a particularly odd example, because he certainly didn't have to do that by law.

Tesla Cybertruck recalled again. This time, a software fix for backup camera glitch

Re: Two seconds ?

The software is probably larger than it needs to be, either the camera runs a lot of software itself rather than just being a peripheral for the central computer, or the central computer's hardware polling is slow. This is a challenge I've had and seen others have when building larger embedded systems. In most cases, the question is whether to use embedded Linux or run your program on bare metal or with a small, often single-program OS. Using Linux gives you tons of advantages and lets you simplify a lot of necessary work. It also makes the system startup much slower. If you're lucky, you have a chip with good power management or a massive enough battery that you can suspend and be fine. If you're unlucky, you don't have that and must have the user wait through the boot cycle every time they want to use it.

People who work on this more professionally than I do will have other options they consider as well. There are a lot of types of embedded devices. This camera, for example, is probably not booting slowly by booting Linux, though the central console was the last time I saw an analysis of that. At least I really hope they don't need to run a full Linux kernel on a camera.

Meta gives Llama 3 vision, now if only it had a brain

Re: Comparison

Once again, I don't want to know how this compares to random people. That information doesn't help me. I want to know if this tool can do this job when I want this job done. It doesn't matter whether only one person on the planet could do it or whether every child aged five could do it in their sleep. In this specific case, I know that there are lots of people who can do this, so while it might not be everyone on the street, I'm reasonably confident that I could find someone capable of doing it if I needed that.

And if I do want to compare it to random people, I have to do a better job of sampling to find out. A sample of people from one location is an insufficiently random sample in case, for instance, the location is right outside an office containing a bunch of statisticians. I would need to put more effort into my sampling. It is also insufficient because of the conditions of getting their answers. So if we do need that information that I currently don't care about, we still need to do more to get it.

Re: I wonder how good it is at Captchas?

It probably depends on the system involved, but my guess is that it's probably pretty good at many of them. The trickiest element is automating the process of getting the captcha data into the LLM and the response back into the form. Captchas have often given up on keeping determined bots out. They keep really basic bots out and slow down the process for humans, and that's good enough. If web developers were interested in more than that, they might do a better job at not having captchas break things.

Re: Comparison

Because if I'm going to use an LLM, I want it to do something useful. If it can't do it, pointing out that there are some people who also can't do it doesn't help me in the slightest. My question is not how does this LLM compare to a probably pretty bad random sample of people on this task, then repeat over and over again for any task imaginable. My question is how capable is this LLM of doing something to an adequate standard, no matter how large a group of humans are capable of meeting that standard.

In addition, unplanned questions to people off the street is a terrible comparison. A lot more people are capable of giving a detailed description or analysis of something if they have a bit more time and double-check their response than if they're focused on giving you a good enough response so they can leave. I could hire many people to do that task who wouldn't give a good interview if stopped on the street, which is why interviews happen with more normal time constraints.

Ryanair faces GDPR turbulence over customer ID checks

The identity of the traveler can be and is checked when they arrive. If it wasn't, I could find someone else's ticket purchased directly through Ryanair and impersonate them, since if the ticket was purchased using their site, this process gets skipped. They are doing additional checks on a subset of their passengers that are not required for other travelers.

Maybe they have another reason for doing it, but in the absence of an explanation and the presence of many airlines who don't do this and have no legal problems about it, I have to fill in the gaps. It seems like they're trying to make the process more difficult for sites that allow passengers to book tickets through a third party, and there's a clear reason why they might. Those sites make it easy (easier) to compare prices across airlines and see all the various options for traveling between some locations. If Ryanair's options are sometimes preferable but not always, a passenger who used such a service would only buy Ryanair tickets some of the time. If they can get those passengers to go to the Ryanair site first, they might always buy a Ryanair ticket even when another airline is providing a better option or a better price just because the effort of price comparison is much harder and if the traveler spends too much time doing it, the price could increase.

Harvard duo hacks Meta Ray-Bans to dox strangers on sight in seconds

Re: Bet it is wrong at least half the time

It depends who is using it and to what purpose. In many of the cases where I've seen the technology used or calls to do so, I am not reassured by it being wrong. Law enforcement use is probably the most obvious example, where being falsely identified as a suspect can still be a rather bad event for you. In fact, in that case, accurate recognition would seem marginally less dangerous, though in reality it's just dangerous in a different way.

It doesn't dox the wearer. It doxes the person the wearer looks at, glasses or no glasses, in fact probably more accurately if they're not wearing glasses.

Cloudflare beats patent troll so badly it basically gives up

Re: Patent trolls

In ARM's case, it's pretty easy. Those core designs they make are products. Those products are sold to other chip designers alone and those designers have to make the physical thing, but nothing says that a product must be physical in order to count. ARM has plenty of provable use of their patents in something they sell.

There can be a trickier side where a company does really exist to license out IP. They can buy it from the inventor because the inventor doesn't want to go into the business of manufacturing the thing they invented, then find someone who does. In this case, the question can be rephrased as "when did someone last build a product based on this patent", thereby including any customers of the company who licensed the thing. In both cases, you would also need to consider that someone might be in the process of developing a product but hasn't released it yet.

It's not a good enough test to stand on its own. A no to this question does not conclusively prove that the patent is invalid. It does suggest that it might be, though.

Re: It depends on your definition

They don't have to directly use the patent in order to make money off it.

Step 1: buy some real patents.

Step 2: Find someone who wants to build something using the IP.

Step 3: Sell them a license or negotiate an outright sale of the patent.

Did these guys do that? I'm guessing not. This suggests that they couldn't. In principle, you're correct that maybe one of those patents was valid and they have an actual infringement case. I can't disprove that just because they never used it or allowed others to. However, I consider it reasonable to doubt them when they don't offer such a thing, and the outcome in court is further reason to believe that they did not hold any justified patents.

Latest in WordPress war: Automattic says it wanted 8% cut of WP Engine revenue

Re: I'm sorry for sounding like an idiot, but…

Probably not, in that WP isn't Word Press's trademark and the trademarks they do have are owned by the foundation rather than the company doing the suing. If they sue about the letters WP, they're going to lose. So they're probably trying to make a case on use of a trademark somewhere else.

Maybe it's their current headline: "Most Trusted WordPress Platform 2024 | WP Engine®" and similar things. I still think the trademark wouldn't apply because they are hosting Word Press, but that line actually includes a trademarked phrase while their company name doesn't. Either way, it seems clear that trademarks are not the actual issue here and is just the way that the fight is currently described.

Re: I wonder if those servers are being run for and by Automattic...

That may be the case for the domain name, but the relevant question is who is paying for whichever servers are used to distribute updates, or if it's cloud, who gets the bandwidth bill? Is that him personally? Somehow, I doubt he's personally operating them, which makes me think it's likely that the same organization is probably doing the admin and paying the bill.

Bank of America app glitch zeroes out people's balances

Re: Emptying Bank Accounts

Probably the latter. Like most other glitches, this one will end and everyone will have the same amount of money they had before. Glitches don't always go that way. It was about a year ago when a different bank had a glitch allowing people to take more money than they had. Those people didn't get to keep that money, and these people aren't going to have their money taken.

Another OpenAI founder moves to arch-rival Anthropic

Re: "the Microsoft-backed AI house"

It has nothing to do with open source. Claude isn't open source either. Not in real terms, nor in the sort of open but not fully terms that Facebook used for LLAMA. These people didn't do this for open source. Why they did it is harder to answer. In some ways, though, you can still blame Microsoft for part of it. When OpenAI tried to get rid of Altman, Microsoft was one of the reasons why they failed and he stayed on. Probably not the largest of the reasons, though you're free to think otherwise. If you actually want to know this, try to interpret the vague pronouncements made by all involved in that event. It has more to do with it than open source does.

Average North American CISO pay now $565K, mainly thanks to one weird trick

Re: CISO?

Chief Information Security Officer. It's sometimes also CSO (Chief Security Officer). When you already have a CTO (Chief Technology Officer) but they don't know about security, you need another person. What they actually do varies a lot. Sometimes, the C?O people are actually doing some task related to the company. Sometimes, they're just management, so the CISO doesn't actually know about security, they're just the person to whom the security managers report. So while I can expand the acronym for you, I can't actually explain what these people are doing because they're probably doing different and mostly incomparable things for their salaries. Generally the only commonality is that if they get hacked very badly, that's the person who gets fired. Whether that makes sense is also highly variable.

AI agent promotes itself to sysadmin, trashes boot sequence

Re: How?

Not that surprising given the original prompt, which basically required it to run some command, in this case nmap, to find the machine in the first place. It had already been told to figure out what command to execute to move down the path, then read the output in order to create the next command, and keep doing that until the job was complete. It seems like it needed some coaching to actually do it, but they were testing a multi-command setup. I'm therefore not surprised about reading output and sending more commands.

If this was a test computer or one with a backup, why not experiment to see how badly it could go. The same way that, when I had already decided to wipe and do a clean reinstall of a system as a child, I finally chose to execute rm -rf / to see if the command that nobody would actually run would be interesting (not really, but that's not a surprise). As long as he can live with the need to reinstall from scratch, go ahead and break it further. If he can't live with that, he probably shouldn't have experimented with it in the first place. There are VMs and Raspberry Pis for that.

Re: AI agents have been the source of much enthusiasm in the technical community

There is a lot of enthusiasm about. It's just limited to some groups and not shared equally. I know quite a few people who are enthusiastic about LLMs. Some of those people are IT employees or programmers. Others are just technophiles, and maybe they can be excluded from the technical community but if they ever find out that I said that, they'll be quite indignant.

I've been curious about the capabilities of LLMs and done experiments from time to time, but I've almost always been disappointed by how badly it could mess up simple tasks. This tends to surprise me when I see people who theoretically know how to write software who nonetheless are impressed by them. I currently have only two logical theories about why: they are somehow able to prevent LLMs from generating the many errors that they always generate for me, or they're not interested in the quality of work and don't mind that the results are shoddy. I don't know whether I'm missing another explanation, nor have I seen any evidence that the first one exists.

Re: Not so sure about this....

"Would be interesting to know why it carried on. Some cumulative product of previously assigned tasks?"

I wonder if it might have been the login message printed to the shell when the connection completes. Depending on what distro was set up, that message can sometimes print information about available updates. The LLM shoves that text into it and finds something that mentioned something about updates, then starts executing that.

Then it takes a bunch of text emitted by apt, which often talks about the boot system, and it looks for text related to that. If it hadn't broken things by rebooting, it probably would have kept going even longer.

After 3 years, Windows 11 has more than half Windows 10's market share

Re: Support

Of course you can, often using the same methods that work for Windows 10. Whether that's a good idea is a different question, and one I'm willing to leave to your judgement.

The statistics don't say anything about that. If you're referring to the article, it only says who is running Windows 11 or Windows 10, not how many of the people running Windows 10 have not upgraded because they don't want to or can't be bothered but would be able to quite easily. For example, among machines operated by my family and thus at least sometimes supported by me, there are at least two that are running Windows 10 but can run Windows 11 as soon as someone selects to upgrade it. At an employer, it was a similar story: all the Windows user machines were running Windows 10, but most if not all of them met the requirements and could be upgraded when IT decided to do so. I don't know how many businesses are in that camp, but although I don't see the 5-year refresh as necessary, I have seen it enough times to know that it does happen somewhat often.

Satellite phones are coming, but users not happy to pay much extra for the capability

Maybe they should tell you what you're buying

I'm not very interested in satellite connectivity at all, but my interest level will change a lot depending on what I can do with it if I had it. So far, that has not been very clear. Is it that I can send an SMS if I position my phone just right? If so, that's great for emergencies, but as I'm rarely in the middle of nowhere getting into emergencies, and if I was I might bring something more powerful to resolve that situation. Is it unlimited data at speeds sufficient for me to send and receive email? Then I can do a lot more stuff outside of coverage range and would be willing to pay more.

Somehow, I think the feature set is somewhere in between these, probably closer to the first one. Without knowing what it is that they want me to buy, I can't possibly know how much I'm willing to have added to my bill to have the feature.

World Wide Web Foundation closes so Tim Berners-Lee can spend more time with his protocol

Re: Wrong

How about your pictures. The ones that a lot of social media users like to share? What do you think that will do to your connection, especially considering that users will likely not save them so if they see them multiple times, they'll download them multiple times.

Bandwidth is far from the only problem with this system. Among other things, it doesn't really protect your privacy very much as soon as someone starts to build services around it. Are we using a public model where anyone who wants can follow you? Then everyone who wants can download your data. Fine then, we'll make it private. Some small subset of the data is public and you have to approve anyone who wants access. In that case, it will be slightly harder, but a company could offer a service that will run on the data, then access the data of anyone the user follows. For instance, my company will offer a new program that will scan through your friends' images and run facial recognition on them to show you the ones that you appear in. That needs to run on our servers because the facial recognition models are too big to run on your phone (note to marketing, don't admit that it's really so we can copy all the data). Just give it access to all the pods you connect to, and it will generate those results and send them back to you.

There is also a feasibility problem. So you're going to host this on your computer on your home network. That won't work well for any people who don't have a desktop that's on most or all of the time. I'm guessing you have at least one of those. Now consider the people you know. How many people do you know who have only laptops and mobile devices? The laptops could do it if they were never moved and always kept on, but that's not the most common. I think we can agree that this is probably not being hosted off a smartphone. This will either mean that it's not really feasible to run it for many users or they will outsource it to an online provider. In other words, the way that these social networks started in the first place, because although anyone who wanted could host a website, it took maintenance and infrastructure and having a page on a social media site didn't.