Posts by doublelayer

How to run an LLM on your PC, not in the cloud, in less than 10 minutes

Re: curl -fsSL someurl | sh

Not if that exe was basically running its own curl to get the real code. A lot of installers work that way, and although they're not my favorite, they think there is a reason because it lets people install the subset of components without, for example, including the translations and fonts for a bunch of languages the user doesn't want in the initial file. Malware can easily use a basic downloader which won't look dodgy until the specific sample has been reported, and it just downloads the more suspicious code and executes it from memory.

Installing this program on Linux (or anything else) will involve one of two things:

1. Downloading a binary from someone else's server and executing it.

2. Downloading a rather large chunk of code and compiling it.

And running it involves one more:

3. Downloading a model, either from their server or another one, and running it.

That's three methods to run malicious code if they are malicious. Saving the script and reading through it won't help you when this line from the real script

curl --fail --show-error --location --progress-bar -o $TEMP_DIR/ollama "https://ollama.com/download/ollama-linux-${ARCH}${VER_PARAM}"

Can download any binary. If they're doing something malicious, the malicious part would be in that file, not this script which I can read in a minute. If you don't trust them, you can try to build from source instead, but I somehow doubt you're reading every file to make sure you don't think anything in it is malicious.

They listed some specific uses. Most of those uses are not good things, but they are specific things that one could do with them, including more tailored spam campaigns. I think what you're asking is why you would want to use one, and it sounds like the answer is you don't want to and you don't need to. I don't need to, and I probably won't, but if I did, it would be to play around with it, not to do something particularly useful.

That said, I can see some situations where some models could be more useful to the average user. For example, there are text to image models that sort of work. Sometimes, they end up making weird, illogical images and sometimes, they don't even get that far and make images that are a bit more horrifying, but sometimes they also work. I cannot draw. If I decided I wanted a picture of something created, running a bunch of prompts and seeing if I could get the computer to spit out something good enough might be a faster or cheaper way of getting the image than commissioning someone with real art skills to do the job. This is not something I'm going to do any time soon as I have not needed an image created, but it is a case where I could see someone deciding that running a local model sounds like a good idea. I don't know how good the local versions are, so they may be disappointed with the results, but the decision to try it could be logical.

Fresh version of Windows user-friendly Zorin OS arrives to tempt the Linux-wary

Re: I bet in spite of the usability angle, there is little to no fingerprint support

Your distinction between a password and an identifier is firmer than it is in reality. Theoretically, a password is also just an identifier, just one that theoretically only you have access to. A good enough fingerprint is also something that's not trivial to look up and provide, so it's also something that you have uniquely easy access to. The problem comes because it's not as easy to guarantee that nobody else has it. For something that needs harder security, that risk makes a fingerprint a bad choice. There is a reason that my fingerprint reader remains unused. If you're using a system where your password is an easily guessed string, as many people do, that password probably isn't more secure than a fingerprint.

There is nothing intrinsic about a fingerprint that prevents it from being a password, though it is prevented from being a good password. The analogy to a name is flawed. I know the names of many people, so simply asking for the user's name is not a valid password. I do not have the fingerprints of any person other than myself, so they are not useful to me as identifiers.

Re: I bet in spite of the usability angle, there is little to no fingerprint support

If someone is going to the level of cutting off your finger, your password is probably not too secure either. If I'm next to you with a sharp knife and I say "tell me your password or I'll cut off your finger", it's likely you will tell me the password if I sound like I mean it.

What I was thinking of for fingerprint compromise is someone going to the effort to get a good impression of your fingerprint, then applying it to something the reader will recognize as a finger. Depending on the quality of the reader, this can be really basic or it can involve more effort, but it doesn't involve physical harm.

Re: Coincidence...

"On what basis *should* that be much of a priority (ignoring all the twaddle about "year of Linux on the desktop" from a small number of strangely shouty people and lazy headline writers)?"

It depends on what you want to see happen. If you want to see an option that works if you make it, and that's it, then it shouldn't be a priority. If you want to see it win out over Windows, as some clearly do, then it should be one because users won't switch without it. Whether they will switch with it is another question and by no means guaranteed, but it's not going to happen if people are more confused by Linux than they are by Windows or if IT departments decide that they won't distribute Linux machines because there will be something that is prohibitively difficult to teach their users to do. I'm guessing, based on the second clause in your question, that winning over Windows is not your goal here, but it is useful to note it anyway because it is the goal of many Linux adherents who post on this site or in this topic, for example whenever the idea of whether Linux should be installed on corporate desktops comes up.

However, if it is not your goal, there is another reason to try to make it happen anyway: more availability of the tools you want. If more people use Linux, more stuff will be developed for Linux, and at least some of that will be something you're interested in running. More Linux success also means more compatible hardware. Consider, for example, something I'd like to see more of which is open mobile devices, devices where I not only have the freedom to replace the operating system if I want to but where there is a realistic chance that my replacement will boot up because I'm not relying on some component that the manufacturer is hiding the driver for. I'd like this so I have more control over the software I run, so I can keep devices for longer than typical ones get updates, and sometimes because they let me experiment. If you don't want that, you can substitute something similar that interests you, but the parallel is probably still valid. If the operating system I choose to boot on this theoretical open device is something that works for me because I wrote it but doesn't work for anyone else, then it's not that surprising that people don't build devices compatible with it. If it's something that people can adopt without having precisely the needs and skills that I have, then more people will use it, and the more of those there are, the more of them may have the ability to build hardware that makes it better. The more people that the software pleases, the healthier the system of related technology, and as a user of the software myself, I benefit from that.

Re: Coincidence...

At least some of this looks to be sites online that list instructions for installing packages that could be installed from the package manager, but the instructions only give the CLI commands to do it. The author of that assumed that a Linux user would be familiar with those, as many of us are, so they just gave instructions for that. I typically install things from repositories from the command line because I have seven terminal windows open already, and so I may fail to consider how many people aren't using that method. The user, in turn, isn't aware that they can read those commands and use a GUI tool to do that.

That is assuming that everything is in the repositories. If it's something that has its own repository, they're almost certain to describe the installation procedure to add that repository to the package manager using CLI commands. I'm sure the GUI applications are capable of doing it, but it's still going to involve finding keys from a site and putting them in a box, and that's going to require more explanation to the user. The set of software that comes as downloaded loose binaries or packages, Snaps/Flatpaks/AppImages, or tarballs with scripts in them, is also large enough that it can cause problems for the unfamiliar user. I confess to being at least partially responsible, because my installation instructions are usually somewhat terse and assume the Linux user knows and is comfortable with the CLI. I don't do a step-by-step instruction on the various GUI options. Maybe I should consider it, but most of my tools are useful to other programmers so it's likely to require action from developers of more user-focused tools.

Re: Coincidence...

Not to mention that Liam makes it quite clear that he does exactly this. In another comment, he explains that he uses a Mac but not the app store on it, which means that to install tools other than the defaults for the OS, he installs them by... downloading binaries from websites. Alternatively, he downloads the code from websites, then compiles them with the compiler he probably got from Homebrew, which he installed by pasting a command from a website. Somehow, this method of installing software, the way that almost everything works at some level, is Microsoft's fault and should never be done. Linux may make it easier by having a repository built in to most distros, but it doesn't prevent it from at times being the only available method to install something.

Re: I bet in spite of the usability angle, there is little to no fingerprint support

It depends on your security requirements for passwords. There are things where the password is not very long or complex because the user believes, correctly or not, that more complexity is not required, where a fingerprint is as secure. This is valid for systems you control, because if your fingerprint is compromised in some way, you have the freedom to turn off the fingerprint access and use something else instead. There are certainly situations where a biometric is considered more secure than it is and therefore the system using it is improperly configured, but there is nothing in it which prevents it from being used as a password for low-security systems.

Re: Coincidence...

"no site that can be trusted will tell you to type commands."

About that: yes, they do. There are still things that aren't found in a package manager's default repositories or that the default version isn't going to work for. There are a few ways around this, but they usually involve downloading a file of some description or running some commands which will do it for you. The file might not be a binary itself, but if it's a shell script, it's no less dangerous. For many users, the distro's default repositories will contain what they want, and I admit that having to use things that are outside it are more often encountered by me when I'm doing more technical things, but there are times when a user might want something that is not in those repositories.

For example, consider tools like youtube-dl or it's modern equivalent, yt-dlp. These things change very frequently because the old versions stop working very frequently. There's a chance that it was never added to the repositories at all, but for many distros, it's probably in there. By it, I mean a version from six months ago that might or might not work. If you want to use that, or a GUI program that wraps it, you will want something more updated. You can download a binary, you can retrieve it from its Git repo, you can get the Python source from Pip, you can find a custom repository that updates the packages more frequently, but in all of those cases, you will be getting some file or command from the internet and running it on your host and if you don't, you won't have the tool. Your statement might represent what should be the case*, but it doesn't represent what is the case.

* Maybe you think that everything should be in the repositories, but you also know that they can never contain everything that someone might want to use. This is especially true when the concept of non-open source software for Linux comes into play. There are some proprietary programs made available for Linux. I have one that I purchased and sometimes install on my machines, and you can bet that it's not in repositories. Their GUI installer is pretty good for less technical users, although the default installation method is still the CLI installation script, but it does involve downloading a binary from a site and using it.

Re: I myself am not so averse to using the command line, but I know I am in the minority.

Of course devs like the CLI, as I know from experience as one of them. We like it for two reasons:

1. We spend a lot of our time in it to do our work, so we're pretty familiar with using it.

2. It's a lot easier to design an interface on the command line than to build the graphical UI components. To build a GUI, you have to write so much code and link it together, none of which is the interesting thing you're trying to solve. If you want it to run on multiple operating systems, your tools for doing so are restricted to a smaller number of frameworks, and each of those comes with some restrictions. Yes, we know how to do it, and when it is necessary, we do, but the first version of something to see if it works is usually tested as a CLI program. This is also one of the reasons why you see some apps using web frontends; it's a lot easier to write something that will work well on multiple platforms if you do it that way.

Of course, good programmers recognize when the CLI is a valid option for the program they're writing and when a GUI is needed, and when the program is to be run by nontechnical people, the CLI is usually not a valid option. Even when it's mostly run by technical people, whenever there are too many elements to fit well on the console, it is time to build a good GUI around it.

Ahead of IPO, Reddit blends advertising into user posts

Re: Blending

I don't post or frequently browse Reddit myself, but I do find myself going there and occasionally finding some useful data. From my brief experiences, I have a few reasons why I don't stay there longer, but first, the positives. It seems like a site where people interested in a somewhat niche topic can find others who are similarly interested. This is kind of why I post on El Reg. A comment section where a lot of people don't understand technology gets kind of old when nobody is bringing up anything interesting about news related to technology, but most of the people here are knowledgeable enough to have interesting conversations on a topic I'm also interested in discussing. I have found small forums and mailing lists related to topics of particular interest to be similarly engaging, and the theory of Reddit appears to be that all those forums can exist on one site, making them easier to find and less likely to disappear because the moderator of the forum got bored, stopped paying the server bill, or even more likely just stopped doing anything including allowing any new user registrations. I have found some areas of Reddit that do appear to work like this.

The problems that have made me a rare reader instead of a frequent contributor sound like they're getting worse. I have never been impressed by the interface of the site, and although there were other frontends, I didn't use the site enough to use them. Now, there aren't other frontends because Reddit planned to send massive bills to all developers, and the main interface is still pretty bad. Maybe there is a method to make it less annoying, so I can actually see a discussion thread in an order that lets me follow it, but I haven't used it enough to find out. The other problem is that there appears to be weird criteria about what is on topic, meaning that I've seen both things that appear irrelevant and arguments about something that looks relevant in the same place. The comments above this one make me conclude that this isn't going away. Therefore, I tend to visit Reddit only when it appears on search results where I'm looking for relatively uncommon information. If a Reddit post appears in a search where I'm looking for someone's comparison between two products, I tend to read that because it's likely to be what I asked for. If it appears in a more straightforward search, I skip it.

Developers beware, Microsoft's domain shakeup is coming soon

If Microsoft ceases to exist, their TLD will probably lapse. But you're going to have to wait a while. Microsoft isn't likely to fail in the short term, and even if it changes its name for some reason, the old one will still have importance. Facebook may call itself Meta now, but the name and domain connected to Facebook still has a lot of meaning. The same applies to Google/Alphabet. When they changed the name of their corporation, they didn't change the name of the service. Even when a name is supposed to be retired, the change does not happen quickly. Elon Musk may want us to call Twitter X, but the x.com domain just redirects to various names without X in them, but with twitter in them. People still recognize the Twitter name, and if someone else eventually takes the husk that Musk left, I can guarantee that they'll want the Twitter brand and will probably not care too much about the X one, which is good because Musk is oddly obsessed with it and undoubtedly will want to keep it for something.

Re: Why break things with gratuitous change ?

It's not a .cloud TLD, it's a .microsoft TLD. The second-level domain is cloud, hence cloud.microsoft. You can bet that Microsoft runs every part of the infrastructure related to the .microsoft TLD.

Re: Single point of failure

It sounds like they are diversifying. teams.microsoft.com is one domain, whereas multiple things *.cloud.microsoft are multiple ones, which can have multiple name servers involved. Not a lot of diversification, but there is some. A major DNS failure probably won't be much better with that, but it won't make it worse.

Raspberry Pi OS 5.2 is here, with pleasant tweaks to Wayland-based desktop

Re: Flash Storage Contortions

Since most of the devices we're talking about are using standard hardware like SD cards instead of a custom flash storage integrated directly with the board, that makes a lot of sense. I wouldn't want to have to deal with SD cards, some of which will work with all devices because they have wear leveling in firmware, and some of which can only be used with a filesystem and operating system built to do it for them. I have enough trouble with identical-looking cards with vastly different speed classes, although this is made easier by none of the devices that I see supporting the faster ones anyway. It's a recipe for people buying a card and not understanding why it fails to work in many devices, including Linux machines because they didn't apply such a filesystem when they formatted it. If there's a basic compatibility layer that exposes it anyway, then they'll just fail fast for anyone who didn't format them that way. I question whether the wear leveling software is that much better at extending life compared with the firmware on the cards to begin with.

Securing open source software: Whose job is it, anyway?

Re: Contributing Back

"remember he is an academic who has never had to earn a living from being a software developer, this combined with a hippie mindset has poisoned open source and our ability to charge anything for it, hence why we have all the open source funding problems we currently have."

I see the problem, but not the solution. Most of the ways that you can mandate payment have an unavoidable effect on some of the other freedoms that I value highly. If, for example, how much you have to pay depends on what kind of user you are, then it's no longer free to use by anyone, anywhere, the way that open source tends to require. It also restricts the freedom to modify and distribute, since the version I changed and distributed still contains most of the work that went into it before, so presumably I have to collect payment and redirect most of it to those authors.

A lot of the suggestions I've seen are, to the user, little different than proprietary except that they can modify the code on their own computers, and there are some proprietary licenses that also permit that. If that's the intent, then, from my perspective, you might as well be proprietary. At least the proprietary authors aren't pretending to be something they're not. I'm open to hearing ideas about how this could be done differently, but so far, I haven't seen one that works, preserves freedoms, and fixes any of the problems with funding.

Re: Contributing Back

"GPLis satisfied by open source being freely available to customers/service subscribers, you want a secure supply chain then pay the subscription to be a user of that service."

This sounds a lot like the Red Hat case, which you're probably aware isn't making them or IBM very popular these days. If it's only the supply chain that you're paying for, that will easily work with the specifics of open source licenses, but I don't think you'll get enough people. If you also intend to keep the source available only to those who are paying, you'll need extra measures, like those Red Hat used, to keep that happening. I don't like that. You are free to disagree.

"I think we need to get away from the free beer which many are acustomed to…."

I think most of the methods used to try to get away from it are harmful to the free speech aspect as well. There is a reason why a lot of licenses are not open source, even though I can see the source. If we end up splitting into open source and commercial source-available branches, I will stay on the open source one.

Re: Contributing Back

If you intend commercial to mean that it has the right to require users to pay, it will run into all the problems that other licenses that intend to require payment have. I understand the desire to get revenue, but it does make a large, and in my mind important, change to what open source and free software have meant which has its downsides as well as upsides.

LockBit ransomware kingpin gets 4 years behind bars

Re: Mikhail Vasiliev the “cyber-terrorist”

I agree on the terrorism bit, but not on the cause damage to protected computers bit. I can certainly damage your computer by working only with the software. I can do that in two ways. I can damage the hardware by running intentionally destructive routines repeatedly. However, what they meant, and what we all understand, is that they damaged the computer as a system, a system of which the software is a crucial component. That damage can be repaired by wiping it and rebuilding, but nothing said that damage has to be irreparable to be called damage.

As for protected, of course it was protected, the protection just failed. Your house is protected; it's got walls to protect it from flying projectiles (blown by winds mostly), it probably has fire suppression of some type designed into it including materials that are hard to burn and alarms, and it has locks on doors and windows to make entry harder. If I walk up with a big hammer and a flamethrower, your house's failure to prevent those from doing damage doesn't prevent it from having had protection, just not enough to withstand what happened to it.

US House goes bang, bang on TikTok sale-or-ban plan

If TikTok were to die, I'd be perfectly happy with that outcome, but I don't think this is an appropriate or even legal way for that to be arranged. I oppose it, not because I want TikTok to stick around, but because I don't think adopting the unfair legal regime in question is a good thing, especially if it expands to more countries and more companies. China already has restrictions like this, and I'd much rather see them drop it than everyone else adopt it.

Oh look, cracking down on Big Tech works. Brave, Firefox, Vivaldi surge on iOS

It might work if you set up your account with an EU address and make sure your network traffic always comes from there. Depending on what you do, you may not appreciate the transatlantic latency for all communication. However, if Apple doesn't want you to use it badly enough, they can collect information about your connection (American mobile provider) and GPS location to determine that you're not actually over there, so don't count on it working now, and if it does work now, it can break at any time. I'm guessing that your best bet will be to wait until the next jailbreak comes out and see if you can unlock it from there.

Re: Brexit Bonus?

I'm not sure if people in NI get this feature, but I do know that buying it there and bringing it back means you don't get it. Your current location, not purchase location, is used to see if you have the right to use those features, so much that if you leave the EU for long enough, you lose them. I'm also not really sure that people in NI would get it. I understand that there are trade regulations in common with the EU, but it doesn't make every EU law apply, and the DMA would have to be applicable law for NI to be included.

Microsoft decides it's done with Azure egress ransoms

"Q. How would MS know the difference between an "egress" and running a bunch of queries on your datatables that happen to involve exfiltrating all the contents to somewhere else?"

They wouldn't, which is why both are billed and why you have to shut down your account or meet some other requirements to qualify for that bill to be canceled.

"2nd Q. Why on earth did you go cloud or Azure in the first place?"

Well in the case of some of my employers, because they're relatively small, so they don't need enough servers to set up their own server room, and renting them across continents instead of collocating them was considered either cheaper or more reliable (I write code that runs on the servers, not choosing where to put them). I think they could have done better by running the compute-intensive internal stuff somewhere local, but running the public-facing stuff on cloud servers makes sense the way they've arranged it.

Whizkids jimmy OpenAI, Google's closed models

Re: My, what drama

Of course they are not hallucinating; they are emitting information that is not of use in the situation, a situation they don't understand or try to fit. Hallucination is the word we apply so we don't have to write that sentence, or the more detailed paragraph, that actually explains what happened. We use such verbal shorthand all the time. After all, we say that a program "writes" to a file rather than explaining how the program transfers data to an operating system buffer and the operating system obtains an available location on a physical or virtual disk, transfers the data from its buffer to that location, and when necessary, links that location to other locations that are related to the file, and if I want, I can expand "file" into several sentences of technical correctness too.

This quibbling about the term is a problem for your argument. We are not talking about whether an LLM "hallucination" is similar to a human hallucination, but how useful it is. Whether it is similar to a brain (no) is not what we're discussing. Many of your philosophical points about whether facts are facts or the result of our perceptions are wholly irrelevant to most of this, because we all have a practical understanding of factually correct or incorrect statements, which is what we have an interest in here. I've had similar discussions before, for example a recent one about what conditions are necessary for a statement to be accurately termed a lie. When you're discussing it on that basis, there are a lot of gray areas and it's difficult to come up with a firm definition, but in real life, there are some obvious lies and trying to divert into the philosophical version to distract from them is not germane and does not convince anyone.

Trying out Microsoft's pre-release OS/2 2.0

Re: Touchscreen

I'm not sure why you said any of this, because if you read my comment again, you'll realize that the "that" that Apple made wasn't referring to multitouch technology, but to the IOS SDK and app store, since that's what I contend was necessary for the success of the iPhone. They did actually make those things.

However, it also doesn't really matter much that Apple bought in a lot of their multitouch technology if what we're comparing is who released multitouch devices. The credit can go to FingerWorks as much as you want, but a lot more people had iPhones than iGestures. By that time, the FingerWorks engineers and technology were part of Apple and thought of as such.

Re: curated app store

The iPhone came first by a few months at least. I'm not convinced that Apple's victory over Microsoft at the MP3 player was that big for the Microsoft side of it. Without the iPod, I think Apple would be in a much worse position, but Microsoft didn't, and still doesn't, have too much in the small electronics business. I think that, had they not tried making a Zune at all, they'd basically be in the same place they are now.

Re: curated app store

"No. When the iPhone was introduced there was no app store (curated or otherwise) and no SDK for third party developers."

That is true, but they did have both of them in a year. While the first iPhone got a lot of discussion and people did buy them, I think its success came after, and because of, those additions in IOS 2. The touch screen, in turn, probably drew the attention of developers when the SDK became available because it enabled them to create interfaces that were harder to accomplish using much smaller screens and keypads instead. Had Apple not made that, the iPhone would probably have done much worse.

Re: Very Different

The problem with your summary is that you stated a lot of supposition as fact and didn't really try to back it up with anything. For example, you describe a plausible OS/2 instead of NT scenario, and that sounds like it could happen. However, you provide no reason why the people who, when Windows was the alternative, chose to buy Macs, wouldn't have done so when OS/2 is the alternative. In my summary, I simplified my guess into "Windows is just bad and OS/2 is just great", but I'm not sure that's objectively true and I also have no reason to think that would be the deciding factor. When you make predictions like that and state them as if they're certain or very likely, we try to understand why you said it. If the answer seems to be that you just made it up, it makes you sound less knowledgeable than I think you are, having seen other articles of yours.

Re: Very Different

I thought the same when I read the paragraph. The really bad part is that Liam didn't even explain why he wrote that, so I just have to guess. My best guess is that his theory goes something like OS/2 is great -> Windows is unnecessary -> everyone buys OS/2 computers, and they like it more than Windows -> nobody buys a Mac, not even those people who in our world did -> Apple actually goes out of business instead of coming close -> Apple isn't around to make other things. Do I believe any of that would happen? No, but that's the most logical way I could go from an OS/2 release to no such thing as an iPhone. I don't think whatever was intended there is accurate, but I also just don't know what was intended there at all.

No App Store needed: Apple caves, will allow sideloading in EU

Re: Your freedom weakens my security.

Fortunately for you, there is. There is a default set of locks that make it difficult because Apple would rather that you didn't go through this in the first place, additional locks in screen time settings (yes, I think that's a weird place for them to be) that make it harder, and a method for you to completely prevent someone else from using it unless they enter a separate code, mostly intended for children but if your elderly friends trust you, that can work. A search will give you full instructions.

I still find it a bit surprising that you're worried about this more than the kind of scams that are more common, the kind where the scammer asks them to hand over financial details, something an iPhone doesn't at all prevent.

AI models show racial bias based on written dialect, researchers find

I agree that the research isn't really telling us anything we couldn't have assumed, but unfortunately, I think having that answer stated unambiguously with some proof behind it is necessary. The reason is that there are stupid people out there who still think that LLMs are appropriate tools for all sorts of things they can't do properly. I'm not talking about some people here who find that LLMs are useful to help them write things. I have my doubts, but at least I can hope that they're checking and further processing the output. I'm referring to companies that want to take one of them, wrap some other code around it to handle input and output, and unleash its results automatically. Whether that's LLM job application filtering, LLM content moderation, LLM schoolwork assessment, or perhaps the most worrying one, LLM criminal justice, there are people who think that that sounds like a great idea. Mostly, they think that because they think they'll get paid a lot to make it and they'll never have to face it themselves. Something like this is another useful tool in proving why they're wrong so I can spend less time hearing why I'm against it because I'm just a Luddite who fears new technology and losing my job.

Re: Yo! dahs here dahalect hatahn' a' dem racahst stereotypes :|

The distinction between the dialects they're using is a bit more complicated than s/i/ah/g. You may be able to determine this from the examples containing I, TH, and various other things your example has excluded, but having several changes in word usage that your example didn't include at all.

Re: Bias that matches society?

I agree with you that the causes are society's responsibility to fix, but the focus on LLMs has a point. If anyone is stupid enough to try to use an LLM to decide on criminal sentences, and although I would like to think that nobody could really want to do that, I am not optimistic, then it is important to know that LLMs will not only fail to help reduce this bias, but will probably make it worse.

The bias reported by an LLM is not necessarily the same degree as that found in general society or the subset who would otherwise be making decisions about criminal justice matters. If the training data contains more input from racists, the result is likely to be more racist, and its input data is checked so little and hidden so well that we would find it difficult to estimate whether that has happened. The other side of it is that society can change and sometimes quickly, but an LLM doesn't adopt that until it's retrained, and possibly not even then. Each individual decision in society can be reviewed, analyzed, and modified, but an LLM does not explain its reasoning and won't change its mind unless it's told to in which case it will simply do what its prompts tell it to. The point of this study isn't that LLMs are particularly biased, but that they are a crap tool for anything where biased output would be harmful.

Re: Broadcaster Presentation / Linguistic Standards

"The problem with this sort of "progressivism" is that it fails to recognize or address an important social function of radio and TV broadcasters: to provide a linguistic standard, and fight against the entropy which fragments language and destroys effective communication between people."

Who said that's what those were supposed to do? My view of media is that it's to provide information, not to show me the rules of language. I have grammar textbooks for that, and I already studied those in school.

Also, I think that having many local dialects will reduce the entropy you speak of. If I have only heard people speaking one dialect of English, it is more difficult for me to understand someone who is using a different one. If I'm exposed to many dialects, it helps me to recognize patterns that are common. I have never lived in the UK, but I understand UK media. When someone came to where I lived and spoke English with a Scottish accent (a somewhat light and generic one), some people I knew had trouble understanding her but I did not. I partially ascribe this to having listened to other Scottish accents on media, something I would not have done if they were suppressed and told to speak received pronunciation or get off the air. Similarly, I had trouble understanding some more isolated UK accents when I was first exposed to them, but having heard them for longer, I am better at following what they're saying. This means that I understand them, and if they are listening to my and other accents, so are they. That reduces entropy, not increasing it, because even if we speak differently, we understand one another. If you want a recipe for linguistic divergence, put up a barrier to people hearing what other people sound like, because then it doesn't matter when they differ so much that they can't understand one another.

Airbnb warns hosts who use indoor security cameras they may face eviction

Re: Easy fix for Register readers

"to be sure you didn't miss one a Flipper Zero can be used to flood all the wifi channels with garbage so his cameras can't communicate"

That might work with 2.4 GHz only WiFi, but trying to flood all the 5 GHz bands is not going to be possible with a small board. You'll need more antennas and more power for that. It also won't stop a camera that records locally and transmits when possible, or in fact one that only records locally to save on power and someone physically collects the card to see what they got. Oh, and people are going to be pretty annoyed when you do it. If you're going to try it, it's useful to know why it won't work.

How do you lot feel about Pay or say OK to ads model, asks ICO

Re: Never

That is absolutely your right, as is trying to block their adverts from your sight. When, for example, advertisers try to ban ad blockers legally, I have a similar problem with the request.

My problem is with those who think that there should be some legal requirement to let them have any service without ads, just because they don't like ads. It's unrealistic, would not work if they got it, and it risks painting real privacy concerns as no more important than wanting something for free. When I'm explaining privacy, I already have to make it clear why my concerns are realistic. It is a similar problem I have had when trying to explain why technologies like DRM are harmful. I raise objections to compatibility or the ability for a customer to use what they've purchased, but if I have people arguing for the same thing whose opinion appears to be that anything digital should be free at all times, it makes it harder for me to make those points and less likely that either group will get what they want.

Re: Never

"My screen, my rules."

That is not how anything works. Their site, their rules. The one who is taking the step to unite them is you: you are visiting their site and displaying its content on your screen, not the other way around. I fully believe you should have the right to block anything you want to not appear on your screen, but not that you should have the legal right for that to always be easy. There is no privacy reason why you should never see an advertisement; how much you see them has no connection to how much tracking the site did to pick which ones to show you. Your desire not to be annoyed is not something that is or should be codified in law with the same vehemence that the right to privacy should be.

You got legal trouble? Better call SauLM-7B

Re: Untestable-Quality LLM-Based Legal Help

And also relevant is that hallucinations that take the form of a completely made up case can be detected, but hallucinations that take the form of saying that a case decided something that it didn't, something irrelevant to this case, or with conditions or in a location that make it invalid can't be algorithmically detected. It's not just missing something that could have helped; the potential to give out incorrect information is still very much there.

Job interview descended into sweary shouting match, candidate got the gig anyway

"what if he hadn't been offered the job? What if the CTO had gone and looked at the code and found the problem and fixed it himself?"

Then he wouldn't have done much work, would he. If I tell you that "looks like someone modified that and broke it", I haven't done a lot of work. It doesn't make it unethical for someone to check on that, and finding that I was correct, do the work to fix it themselves.

If I was asked to look at their code, understand it, and fix it, all as an interview question, that would be a different story. That's a serious amount of work to do and would have taken time and effort they haven't paid for. To suggest what appears to be the most likely cause in broad terms from a quick summary is a very different scale of effort and one that I don't mind being asked for. There is a major difference between describing how you could do something and actually doing it.

"What if they had only offered him the junior position in the end?"

Then he would have the job he was hoping to get when he started the process. His opinion could easily have changed. I know mine would have been if an interviewer had yelled at me, because I would probably right off the company right after that, so there's several reasons he wouldn't have taken it if offered. Still, he entered the interview in order to and with the hope of getting the junior position, and that would have been a completely valid outcome of the process.

I do know this from experience. Early in my career, I was asked to complete a grueling interview process where the company asked for far more than was reasonable and I, knowing little about it, gave them what they asked for. I made sure that, if they wanted to use the code sample they asked for, the license wouldn't let them, but other than that, they got plenty out of me. I don't think that any and all tests are acceptable, but there are some that would be. This type of high-level test seems justifiable to me.

This is where we disagree, I guess. I don't think that spending over a few minutes is screwing myself; I'm at this interview and they're going to ask me questions to try to assess my skills and abilities, and I expect that I'll be tested in a variety of ways. That's why I would do it in an hour-long interview, but not a multi-day project. Nor do I think that talking about a theoretical problem which turns out to be real is that unethical. We all have our limits for what we think acceptable, and I would seem to have a higher tolerance for that than you do. I don't conduct many interviews, but I've taken quite a few, and I rarely felt disrespected because of the technical questions they asked.

My problems with interviews have mostly been in other areas. Usually, my negative reactions were more "that question is stupid" than "that question is trying to take advantage of me". Maybe I simply haven't experienced the questions that you find unethical to ask, so I'm considering something different than you are.

This really depends what the problem is. If it's "our massive codebase which you haven't seen and won't see now is broken, how do we fix it", then there's no way for the interviewee to answer it and no useful information to be gathered. If it is a bug that can be described, and they ask how the candidate would go about debugging it when the cause is unknown, it allows them to give an idea of their thought processes. In fact, in such a situation, the bug's cause being unknown is sometimes an asset. If I describe a bug, and I know that the cause turns out to be a communication problem between the application and the database, I might dismiss any other suggestion they give because I know it's not what happened, even if I tried the same thing back when I didn't know. If I don't know the answer, then I'm probably thinking about what did happen. However, the number of situations where you have a problem that can be a viable interview question are low.

In my experience, interview questions that end up being relevant to work they're doing often involve the design, at least for the primarily programming interviews that I have done. For example, I've been asked how I would design a system that they are also deciding how to design. Maybe the suggestions I make there will be something they didn't consider themselves and get adopted, but I haven't really done much work for them by describing it. They learn about my ability to design something practical, not just theoretical, and any productivity they get out of that is likely to be small. I don't begrudge them using it.

Re: Hypothetical Example

I'm not sure "all are equal" is always correct there. Not that "director always pulls rank" should be either. There are real cases where there is a higher priority involved and a justification for leaving a user's problem unfixed so that something worse can be prevented, but management often thinks there is when there is not.

I have less of a problem with interviewers using real problems as a part of the interview, though it shouldn't be the only part. There's a limit to how much work you would expect an interviewee to do, but hearing how they would solve a problem with an unknown solution is useful because there will be plenty of those for the actual job. I don't particularly mind that the ideas I come up with might be used to solve an actual problem. If I can tell them how to fix their problem in an hour and they can actually use what I said to get the job done, I haven't done something inordinate and they haven't gained a massive advantage. Had it been a multi-day project, I would feel differently.

That said, my typical rules of courtesy don't allow that to get into the shouting match area at all. There is no situation where, as an interviewee or interviewer, I would shout at the other person. If I'm disappointed with their answers, I will either show disappointment, or most likely stay neutral and note it for making the final decision.

Apple's had it with Epic's app store shenanigans, terminates dev account

If there are basically only two sources of flats, and they use that to make conditions that are basically the same, yes that would be a problem. This usually isn't the case because there is more choice in providers of flats, and when there isn't, landlords usually use it to increase the prices rather than add pointless conditions. When that happens, it is something that gets complaints, investigations, and regulations.

Re: the average Apple user spends more than four times as much

True, but what I meant was that they do not have the option to develop that as an alternative if they're planning to keep a Google option at all. It is an all or nothing venture, and that means OEMs are unlikely to try it. Developing a consortium wouldn't work very well unless they all did it in secret and unveiled a mass exodus overnight, since as each manufacturer indicated an interest, Google would call in their contract terms and, since their version wasn't ready for release, their competitors would end up getting business. The chance that a company would take that risk when their competitors have not is very low.

Re: the average Apple user spends more than four times as much

"What about Android? Google gives it away for free,"

No, they don't.

"and if they couldn't make a penny from in app purchases then what's the incentive for them to keep developing Android?"

All the OEM fees and mostly the massive amount of data they steal, and I want to do something about that as well.

"Should they charge OEMs to license the OS"

Already doing it.

"(and see them say "bye Google we'll do like China and use AOSP to create a free version")"

Their contracts forbid that.

"or should they embed ads into the OS itself to fund their development efforts?"

Already doing that too, or at least data collection and a lot of their apps which are required to be installed by OEMs have places for the resulting ads to appear.

Of all these choices, the one I'm happiest with is OEM fees. I give them money to buy the device, they license the software to run on it, that works for me. Admittedly, it's probably one of the reasons why Google talks about but never actually does much to increase software update timelines, because if my phone stops getting updates, there's a higher chance that I buy a new one, but I don't see their Play Store tax or their data collection causing them to release any security fixes.

Trump, who tried kicking TikTok out of the US, says boo to latest ban effort

Re: Reference

What gave you that idea? A company can be involved in politics in most ways, and the bigger they are, the more likely they are to be involved. They are certainly allowed and able to lobby politicians. There are sometimes restrictions on a few matters to make it harder for companies to bribe politicians, but otherwise, they're allowed to do pretty much what any person does. For that matter, you don't have to be a citizen either. I am not a citizen of the UK, but if I want to start writing up election posters campaigning for the general election that's coming this year, I can do it, put them online, send them over to the UK, and many other political activities, perfectly legally. I just can't vote in it.

The S in IoT stands for security. You'll never secure all the Things

Re: Some smart devices have strong security

Admittedly, some of the most successful locked-in devices are also quite secure as a result because the manufacturers go to a lot of effort to make sure that I can't break in if I bought it which also keeps out most others. It just means there's a different reason that I'm not buying it.

Tesla Berlin gigafactory to take week-long nap after suspected arson

Re: Typical Elon argument

You can't use that logic. Just because a comment doesn't criticize something doesn't mean that the writer supports that. If that were the case, you have, by omission, announced support for murdering anyone on the planet, simply because you didn't specifically say you opposed doing so.

The comment to which you replied was about two things: the comparison of pollution output from a factory and the actions to recover from the damage caused. While the former is negative toward Tesla, it doesn't say anything at all about the arson being justified. Nor does stating a negative opinion imply that. Someone can say that they don't like Tesla and still have an even larger problem and greater opposition to arson. There are clearly a large group of people who think Tesla's factory is causing pollution problems, hence many protesters who didn't burn it down, but that doesn't mean they all looked on as the sabotage occurred, totally supporting it.

US politicians want ByteDance to sell off TikTok or face ban

Re: Choice

It's just taking a leaf out of Apple supporters' books. It's a choice to not have a choice, and that is more important than anyone who wanted a choice to have a choice. Those politicians should try the relatively basic tactic of imagining what they would say if someone did the same thing to them:

The government of [some country, I don't know, let's just say it's the entire EU and they now have any authority necessary to make this hypothetical work] has announced that Google must sell its platform to a local subsidiary or be banned. The president of the EU has the ability to, without even passing another law, apply the same rules to any other company, effectively at will.

There are some legitimate worries about ByteDance's operations and control, but this law is not a valid way to deal with it. Laws intended to counter concerning actions should be directed at the actions concerned, not a specific company. Politicians shouldn't need this pointed out.