Posts by doublelayer

Commodore gets into the phone biz with Sailfish-powered retro 'Callback'

Re: Bah!

The OS is like a smartphone, the hardware is not, which will introduce some problems, and the software they've installed on top of the operating system might make things worse. I don't know as I've not seen it, but this is not the same as Sailfish, an operating system intended on building a smartphone, it's that plus software written by people intending on making a phone for people who don't like being distracted but are unable to disable things on their own. Since I don't have a problem not installing social media apps when I don't want them, I do wonder how many things there might be in that software which would get in the way of something I did want to do. Ideally, they just block Facebook, which since I have no intention of using it, wouldn't get in my way, but I don't know if that's likely.

This isn't just Sailfish in the same way that KaiOS isn't just Firefox OS. The code at the base is the same, but everything around it is intended for something completely different which makes some things less practical and some things impossible.

Re: No

It does support other messaging apps. How well it does that is another question, as it's still designed like a flip phone. There are several AOSP-based things like that around, and you can sideload things onto them either because that's intentional or with a little hacking. I've installed Signal on a flip phone like that, and technically, everything works. In practice, the app isn't designed for a tiny screen, typing and navigating isn't convenient, it's functional if you have to but not as convenient as the SMS app designed for flip phone use, which itself is not as functional as any type of messaging app on a touchscreen.

Firefox 152 understands 'Sssh!'

Re: I run µBlock Origin on FireFox and µBlock origin lite on chrome

And how confident are you that removing Manifest V2 will make any difference to the startup time of Google Docs? If there are no ads being removed, do the slower (claim not verified) operations that don't need to run to remove things still affecting it? Are there other operations that occur nonetheless that still affect it? Or is this just speculation that maybe removing code will speed things up, which is only true if that code was noticeably slow and running at the time?

Re: Confused

I think it's different because it's a mute all button, so you don't have to find the tab. For people who routinely have hundreds of tabs, that could be a big haystack they don't want to search.

Re: They are going about silencing tabs the wrong way

Meetings, for me. If you have an web-only meeting platform like Google Meet or a meeting platform you don't want to install the local binary for, that's in a tab. I often need to be on the meeting, hear people, and look at or take action in another tab at the same time. There's also putting on some audio while you do something in another tab. I would prefer if the default allow or deny was configurable, as I have not found background tabs wanting to start making noise very much so I am content to let them continue by default.

Anthropic reserves right to check ID for Claude subs

They didn't say why they were going to collect that data. With citizenship restrictions, ID verification isn't sufficient, because an immigrant who hasn't been naturalized yet can still have local ID, more problems if dual citizens are also not allowed (I wouldn't put that past the US government right now).

If they're trying to prevent Chinese companies from copying Opus models by repeated prompting, and they think ID will do something about it (I doubt it), then they could easily do that to Opus too. Their statement does not suggest when or why they're going to do this, just that they might, so I wouldn't be so confident that it only applies to US users of Mythos.

That's just two of their models. UK people can use the other ones just fine, and Anthropic still reserve the right to demand ID from people who use them any time they feel like it. Why and what they want to do with that data, well that depends how clear their legal document is and whether you trust them to follow it. Fortunately, UK people at least have the DPA to back up the privacy regulations, and surely that has strong enforcement at the ICO to prevent misuse of all the sensitive identity documents that this and similar legislation from the UK will generate...it's an idea anyway.

The new Siri makes one of Apple's most convenient OS features a cumbersome mess

They have a similar situation to what you seem to have. You've mostly seen negative comments about AI, so you assume that's what's out there for them to see. There is plenty of that, but another thing there is plenty of is people complaining that Apple missed the boat on AI, promised a development they didn't deliver, and really need to fix that in order to catch up. Those people aren't negative on AI, they want more and they specifically want it from Apple and will criticize Apple for not making it. Many of those people are the types who work in tech media for non-tech people. Perhaps, like me, you find that kind of thing annoying because they often swallow any marketing promise from a tech firm without understanding it so avoid it, but it doesn't change that some of them have much more readership than El Reg does. If Apple sees that a few million people hear that Apple must leap forward on AI or be left behind and a few tens of thousands see complaints about all AI, they have two reasons to follow the former. The quantity one is obvious, but the other one is that, if someone hates all AI, and everything has AI, then expect no change in how many of them still choose an Apple product before turning the AI off.

There's no such thing as an agentic CPU

It's a different way, but not that much more different than binary CPUs can be. There are far more things that can distinguish what a processor does and how it does it than just how many states the smallest thing they operate on is. In the end, the difference is speed and efficiency, nothing else, as every CPU, binary, ternary, decimal, or Church numerals*, can do exactly the same set of computations.

* Useful for fun mathematical papers, useless for everything else, the computer science theorist's equivalent of the one-instruction ISA.

France's digital sovereignty push is struggling to escape the Microsoft gravity well

I propose an alternative method of switching people to non-Microsoft Office: pick a standard you think most people can use, actively switch some people to it and require them to use it, have a group whose job is to provide support and document what happens, use that experiment to identify what goes wrong most frequently and whether you need to train users or fix the software to get around it, use those lessons to do a bigger migration, repeat until everyone or almost everyone is successfully switched, probably some corner cases for a few people or unusual use cases.

Doing it in pieces leads to friction. Group A is busy and doesn't want to switch right now. Group B has some people willing to switch, but they've got an Excel guy who refuses and they don't want to have multiple pieces in active use at the same time. Group C actually did try, but they ran into some problems, there was no support, so they went back. Group D has no objection to trying but rely on things produced by group A so are only willing to when group A does. If they all switch together, there's more chance that they figure out how to solve group C's problems because someone in group B had the same thing and found a solution, group D easily switches to using group A's new things, and the Excel guy in group B is dragged along because all the Excel guys in the other groups proved that there's only a minor change required for the macro pyramid to keep working. If this is a big, centralized priority, they can afford a big, centralized assistance plan that makes sure it's implemented without breaking things.

Britain plots digital bedtime after kicking under-16s off social media

What you suggest mostly already exists. Domain allowlisting isn't in the default restrictions, but filter lists with sites either added by organizations or voluntarily adding themselves are plentiful and easily enabled per device or per connection. Plenty of existing software can add the stricter method of only allowing domains you've approved, though I expect parents actually doing that are going to get annoyed quickly by having to approve things so often, still, their choice. This means that any parent who has bothered to read about and enable that can do what you're suggesting already. This leaves us in the situation that some parents didn't do that because they don't care, either at all or at least enough to do the work, and what if anything do we do about that?

Legislators in the countries that have done this probably don't know about what exists because they, like so many others, don't care, but part of the problem is that some aren't turning on these restrictions because they don't agree there's a problem or they don't want to argue with their children, or they didn't bother to find out how. If those happen, are those problems we need to solve or valid choices? Existing laws seem to decide that these are not valid and everyone must follow these, and that's the remaining part of the vitriol. I'm most concerned about identification of every activity, but I also don't like that politicians have decided they should decide what children may or may not do and enforce it on every individual, parent, child, or otherwise.

Like it or not, and I generally don't, countries' citizenry at large tend to get blamed for actions from that country, especially if they're longlasting ones. To demonstrate this, haven't you seen or made some statements in these forums suggesting that the US bears responsibility for the actions of its current government and suggesting that, even if that government falls, the country would still be untrustworthy, even though you know there are US citizens who oppose those actions? That's what is being said about the UK here. I'd prefer if they were more specific, but it's far from unusual and they have plenty of UK-sourced privacy erosion to complain about.

In addition, they're not necessarily wrong if they round the results. The article quotes polls on these regulations. Polls specifically used to engender support for it, so probably not great ones, but nonetheless they show rather large figures in support of the changes. There's a sizeable chunk of the public which doesn't understand or disagrees with the concerns we have against a policy like this, and they vote too.

The same point was made by the support being used to bolster the argument for this law: "Three-quarters of respondents said restrictions would lead to fewer arguments at home, while 77 percent said schools and teachers would find it easier to manage children's digital behavior."

I don't know how that's being received by others, but to me, that's one of the worst arguments I've heard. Parents have lots of control. While I usually oppose the "do this because I command you and I get to do that" style of parenting, parents can do that if more logical arguments fail. If a parent can't explain to their children why they shouldn't use social media and won't forbid them from doing it unless a law requires them, do they really care? There are far more important arguments parents need to have with their children. Those shouldn't be legislated either.

What makes you think that? There are plenty of things that would still cover the message from Reform or any other party on which young people can still freely view them. Also, if Reform can make the point that they're the ones who weren't in power when your devices started recording you every time you want to access a site, that could boost their electoral support among the young. This is also only relevant to the youngest section of voters, because after their first election, they'll have had plenty of time to see everything social media has to offer because the regulations only block and identify children but do nothing about abuses of the platforms that affect adults. This is not the problem the regulation is trying to solve, it will not be solved by it anyway, and if you want to solve it, you will have a very hard time doing it at all, let alone without serious consequences on civil liberties.

A modest proposal: Reformat everything to make documents more palatable to AI

I know people want to input documents. I don't know why those people are incapable of building a parser for formats instead of shoving it into the LLM directly, and why their solution to this is not to build that parser but ask me to use another format. From the description so far, there's no advantage to anyone except the LLM user of using this format. It seems like it would take longer for me to write than Markdown, which is efficient for me to format without thinking much, and it doesn't seem to do most of the useful things that HTML can do, so I'm expending more of my energy for LLM users' benefit. If I wrote in any other format, then it would likely require automatic exporting to DocLang anyway so they might as well take the same code for automatic importing as DocLang. If you're asking me to change my methods, I generally need a reason, and "so you run the export code instead of me" doesn't seem like a very good reason. Is there a better one?

Re: PDF and AI

Exactly. They'd get better input by implementing PDF processors which look at the files, extract as much information as they contain, then using more expensive visual processing only when necessary. The same goes for most other input formats, where instead of shoving an entire HTML page into an LLM and then complaining that you're having the LLM do the parsing, shove it into a tokenizer which strips out errant spans or whatever else is considered useless verbosity and then put the output from that into the expensive LLM. Bonus for the LLM users, that can eliminate all sorts of tricks like text with font size 0 or white on white using a few if statements. If the people writing these systems can't understand how or why to do that, why should we assume their suggested format has any utility at all?

Chinese e-tailer claimed 14-inch box stretched the size of a 9-inch tablet

The Surface's one does have the problem they specified partially, or even mostly, because it connects to the screen part instead of the keyboard part, which means it has a short bend because the bottom of the connector is very close to the table. Had they put the connector higher on the screen so it didn't have to bend as far, that would likely be less prone to damage.

They don't say if the software simply wouldn't work without one or if the software just had enough stuff on the interface that a small tablet would be painful to use. Most software can't tell the size of the display, only the resolution, so it would run equally happily on a 1080P 14-inch tablet or a 1080P 9-inch one, but the users would still notice a difference.

Apple version of Office 2019 becomes useless in a month

Re: "plenty of alternatives"

I think I agree with your attitude to regulation, but it's not fair to say that every company does the same thing. Even Microsoft didn't do the same thing with other versions of their software. For some reason, they decided to do it with this one, and we should stop them if there's a way to do it, but that doesn't mean everyone would try that and if we act like they would, we can punish other companies based on the fear that they would do something they haven't which really doesn't help prevent the problem from growing.

Re: What does the EULA say?

That's not the case. Windows also uses a certificate, just not the same one. I don't know whether they'll do anything to protect Windows users from this, but they could decide to let them fail at some point too.

Re: What does the EULA say?

It seems that Apple licensing was implemented in a different way, with a different certificate. From the way this went, there's no reason they couldn't do exactly the same things to Windows and Android too, and perhaps they will whenever the certificate they used for those builds expires. In both cases, fixing it would involve getting another certificate, quite easy as they're the ones who issue and approve it, and shipping an update that does nothing other than contain that new certificate with the existing, unsupported code. I don't know if they have any legal validity in not doing that*, but it would be easy for them to respect the purchase their customers made and they're nonetheless choosing not to.

* Any other software that relied on external licensing servers which eventually shut down or stopped accepting license validation requests has the same situation. Microsoft's a glaring example because they're still shipping these updates for other operating systems and for other versions and because they're so huge that they obviously could do the work to release the certificate maintenance updates, but the wider problem isn't new which suggests the legal complaint method might not be as feasible as we want it to be.

AI is code – and can't be prompted into being smarter

Re: That means what, again?

Property-based testing is a way of getting test cases for unit tests without having to manually come up with interesting values. For example, if I create a type in a library and you're going to use it, you probably want to test that you can handle corner cases coming from the type. The non-property way to test that is that you think of cases and test them, but since I wrote the definition, I probably know about far more possibilities for weirdness. I can define the properties of this and their possible values, and you can use those to automatically generate a bunch of test objects to run through, providing you more opportunities to identify things you hadn't thought about in the unit testing phase. There may be more to this or the library, but that's the short version of property-based testing.

Re: "this software is not intended for use outside the United States"

No, the Cloud Act is neither a contractual term nor Microsoft-caused sabotage, it's a conflict between US laws and places they shouldn't apply but the US will act as though they do. That doesn't work as a parallel, but if you disagree, you can consider the original question with that: does that make it totally fine? We can consider the ethics of this approach without having to be LLM fans. I have done so above, mostly focusing on the legalities and whether the license needed to be different to protect the creator (no).

Re: Dune

I agree. It's even gotten to the extent that I don't read large series even if I've read single works by an author and enjoyed them. In my experience, a few are good, and many will have about as much good stuff as would fit in two books, but you have to read ten to see it. There must be more exceptions to this rule, but I don't know them until I've already read two of them*, so it can take a while.

* I think series can at times justify having not as much in their first book, which introduces a bunch of things. If neither of the first two were good by themselves rather than what you thought they might lead to, abandon it. Maybe some day I'll find a way of finding out this answer without reading that far into them first, and thus I can stop avoiding anything with too many titles in the list.

Re: In support of Plain English

I don't think the "clearly" is as clear as you think. It's certainly arguable, and a risk-averse person would probably be well advised not to put that behavior into their software because, whether they win the case or not, it's not fun to have to deal with one. However, the factors of open source code and manual installation weaken many CMA claims. You end up debating what counts as permission, since the owner is the one who chose to obtain the code and run it in the environment they picked and because the behavior wasn't hidden. Most CMA claims concern either malware which someone else installed or software which disguised the damage it was going to cause, but neither apply here.

Re: In support of Plain English

I don't think a legal complaint would work, and I think his choice is sufficient for that purpose. There's already the disclaimer of liability from the original license, and I think that's also why he said "is not meant" and not "is not to be". The latter would either involve making a new license which tries to prohibit it, a license term that definitely makes the license incompatible with definitions for free software and open source, or it is somewhere else and conflicts with the existing license. "Is not meant", on the other hand, is a warning that stuff may not work, meaning that a legal complaint about poor quality or negligence, which already has a steep climb to get over the disclaimer, is weaker because the user was warned about consequences.

The only avenue left for someone making a complaint is that the instructions were deliberate attempts to sabotage their systems, and that works whether or not the license permitted or forbade use with something else. Even if I violate a license to use a piece of software, if they have embedded criminal code in it, I can still have them prosecuted for that. It is just hard to demonstrate that when the software was openly readable and installed solely by the complainant.

Re: Hilarious article leads to question

The way they're built now, there is no separation and no way to make them separate. I could envision ways to try that, weighting instructions from one field highly but explicitly de-weighting them from another (and several other required stages) but there are three problems with those I'd have tried if I was working at one of the LLM companies:

1. They would require training new models, at least the refinement stages, and that's expensive. What I hypothesize could be partially useful cannot work from system prompts around existing models.

2. Users would have to be more careful about what they put in each part, and doing so incorrectly would likely produce poorer quality output. Having seen LLM users, I'm not sure many would notice independently, but it only takes a couple people to get angry about degradation to generate a bunch of negative feedback, as the example of the protests around GPT5's lack of praising everything no matter how stupid demonstrates.

3. It would not work perfectly and might not work at all. Like most things using the machine learning algorithms on which LLMs are based, trial and error is the only way of testing what happens and how often it happens. Having someone knowledgeable do the trying helps, but it is not enough to guarantee outcomes. Having me personally work on this wouldn't be ideal either as I have never written an LLM, only other types of ML models, and even then not for some time. LLM companies have far more experienced people they could assign to this, meaning either they don't care about fixing this or they tried and it didn't work.

Scientists pour cold water on claims phones are rewiring kids' brains

Re: One decent bit of common sense

Maybe I'd be more in favor of mandatory age standards if parents were trying hard to prevent children from doing things and failing. When parents don't try at all, then I fail to be convinced that the many tools that exist aren't good enough. So far, the only theories I've heard for why parenting, or at least the part where you keep things you don't want your children to see from them, is harder nowadays are three unconvincing points about tech:

1. They've got a device on which they can look at anything and I can't physically watch them all the time. That's why you control what devices you give to them, whether that's by putting control software on it, only giving them restricted devices where doing things you wouldn't like is difficult or impossible, or not letting them have it all the time.

2. If I don't give them a phone, someone else will. This is not true. Children don't have a magic phone tree handing out relatively expensive hardware. They can, maybe, circumvent your control at times to read something you disapprove of, but that's not new, so parents have to handle that just as yours did, with perfect control impossible for you just as it was for them.

3. Existing control software can be disabled by someone with sufficient tech skills. Leaving aside the problem that people with those skills can probably disarm the software you're trying to get mandated as well, most of these "sophisticated attacks" involve things like guessing a reused password. I have yet to see any of these claimed easy bypasses that justify forcing everyone to live with your parenting preferences.

That is not what they said. You're doing the same thing the MPs were doing, insisting that the data says what you want it to say despite repeatedly being told otherwise. Partially, this was the scientists' fault by being polite and assuming people who ask them questions are smarter than they are. When they say "correlation", they think it clearly communicates the lack of strength in the data because "correlation does not necessarily imply causation" has been drilled into anyone working with data for a long time, though Professor Mareschal did attempt to explain this. I'm guessing you know that too, but because you think that damage is plausible, you too are willing to interpret this data differently than you would otherwise. Nonetheless, that's not the right approach.

Specifically, your "there's lots of evidence" parallel, no, not at all. The professors were asked to comment on the quantity of usable research, and they were consistent:

Blakemore: "What evidence do we have? Almost nothing. There are a few small studies, but they haven't been replicated, and they're purely correlational." (That one was in the article. Did you not read the article? Or maybe you don't understand why unreplicated research is a problem even bigger than correlational data? I have a feeling you know both those things.)

Dorjee: "As my colleagues have said, there isn't a lot of direct evidence that would involve brain scanning and could track changes in brain structure and function" (She goes on to describe things you can use, but since they don't involve saying "this is bad and I can show you on this chart", they kept asking for that instead).

Mareschal: "I know this is a mantra that there's very little information"

The MPs wanted, and your inaccurate summaries are insisting, that simple data exists. You're acknowledging that it's observational and not strong, but that's not what the professors were saying. They were saying that the data is inclusive and sparse even when you consider that observational information. They give plenty of advice about things you can measure which would suggest harms and ways to address them, but since they aren't as simple as at what age you stop blocking social media use, for which they were explicitly asked and refused, the MPs kept asking for that. Plenty of the harms they mentioned didn't work with the predecided plan because they indicated ongoing harm to adults as well, others because they suggested situational factors which a parent can manage but a law cannot, and still others because they suggested that harms weren't the same, sometimes less and sometimes greater, than the random guesses from armchair neuroscientists with no experience.

Re: Forget the adolescent brain

That's a fair point, but I'm not sure a single activity is necessarily a sign that that is an addiction. If, as certainly applies to me, someone has an aversion for whatever reason to stopping halfway through something, then an activity that can be done while continuing to watch feels easier than abandoning it to do something else. I suggest adding another question very similar to yours: do you [someone noticing this, ds999 in this case] feel like getting the phone similarly during other activities or just when watching television, and do you feel it the same way no matter what's on the television? I think that might be more correlated with the answer we're looking for.

I think you have it wrong. When someone makes a claim, pointing out, as the professors repeatedly did, that the claimant have no evidence for their claim, no way to obtain the evidence they claim would exist, and the research that comes closest doesn't even always correlate with that claim is pouring plenty of cold water. It's true that, if they could have said "we did research this, got exactly the evidence you're asking for, and it says you're wrong", that would be much stronger opposition, but the burden to defend a claim as true, especially when proposing legislating as a result, is on those making the claim.

Consider a parallel:

Claim: There are aliens on Saturn.

Researcher: There is no proof of that. We have only a few tools for looking at what is on Saturn, and there are parts of it we can't see. Everything we have seen can be explained by understood natural phenomena and all observations from probes and telescopes works together.

UFO people will often decide that statement doesn't weaken their point, since the researchers admit that aliens could easily hide from observations and they can't prove they're absent. In reality though, it does weaken the claim severely unless any reason to think it's true through valid science arises, independent of someone thinking it's likely or wanting it to be true.

Re: Forget the adolescent brain

I sometimes have that too. When it happens to me, what it often means is that the thing I'm watching is not that interesting, but I'm making myself watch it anyway because I started so I'll finish*. I don't know if that applies to you as well, but it could be useful to consider whether it does because wasting time not really enjoying the television is not any better than wasting time not enjoying the content on a phone.

* I have this reaction to many things I probably shouldn't, a combination of sunk cost and persistence being its own reward (it isn't). If a book's not good, then there's no reason I should feel the need to finish reading it anyway just because I've already read 30% of it.

EU sovereignty push gives tech buyers a new alphabet soup to swallow

Re: Possible fly sighted in ointment.

This is where "sovereignty" needs a very specific definition, one that few people have tried to state and those who have don't agree. If the concern is things like the Cloud Act, then no, Linux is not a risk, because the Linux Foundation doesn't run any of the systems on which it is installed and therefore they don't have the requirement to help law enforcement get access to them. If the concern is that the US government will force anyone inside the country to become saboteurs, then perhaps you do need to worry about Linux but you'd also have to worry about a lot of things where the people writing them will be less motivated to refuse any such demands (which would incidentally be illegal). And if the concern is running anything that wasn't written entirely within the EU, then Linux, along with every other operating system in common use, is unacceptable whether or not the main people stay there because it's got a lot of global contributors.

My opinion leans to worrying most about surveillance that already exists, so Linux's US foundation doesn't concern me. This is helpful, because software like that which can be run and maintained by others if something changed makes it possible to switch from software that can't. If you have to build everything yourself, it becomes much more expensive and difficult, and I think people are more likely to abandon efforts like that after years of little progress.

BOFH: For one ambitious security type, chaos is a ladder

But why would he need to set that up when the lack of video is an advantage in this particular situation because it demonstrates exactly the things he's framing the head of security with?

So this security chap says he didn't steal any laptops and they must have been planted? Sure, let's check the video for this area.

[...]

Would you look at that? Ransomware on the system with all the surveillance files. We've lost everything before the latest off-site backup. This is what happens when you don't scan computers when they've been incompetently managed.

[...]

My fault? If you'll remember, our friend here in security required their network segmented off and forbade us to scan machines for malware. In fact, I have a video of him refusing to let us disinfect a machine because I was documenting the asset labels of the ones I was removing, and I have this recording of a call where my boss accepted that. You know, if you think there was planning, it certainly seems like it was this security fellow, possibly with my boss. Malice or incompetence, it goes the same way, but I think it's both.

KPMG's AI report becomes an accidental demo of AI hallucinations

Re: Lazy report writers

That depends whether you see adherence to an LLM as an automatic expulsion from the intelligence club. Unfortunately, I have seen people who are otherwise intelligent who don't understand what LLMs do and why they shouldn't trust them the way they do. That doesn't prevent them from continuing to be able to do the things that required intelligence before. I would consider it simplistic and a bit arrogant to declare their intelligence gone because they didn't agree with me on one thing while continuing to be able to do things that few others can manage, the same way I would consider it simplistic and arrogant if they declared me unintelligent in all respects for failing to agree with their theories about LLM brilliance.

Intelligence isn't straightforward. There was a time in my childhood when I assumed that intelligence was one-dimensional and total. When I wanted to know the right answer to a question, from "what are protons made of" to "what is the most ethical way to balance people's needs or wants with limited resources", I assumed that I just had to find a person smart enough to answer it, and if people disagreed, I just needed to find a smarter person to make the decision, and I was learning how to be smarter so I could make some of those instead of needing to find smarter people than me to do it. Of course, that isn't how it works. Though it's unpleasant, it is possible for someone to be very intelligent about one thing and stupid about another. The generic openness and interest in improving helps, but it doesn't necessarily work on everything.

I think this was partly a test of the tools, hence their question about reliability. But also, humans aren't great at detecting that either, and the shorter the text, the less reliable they are, especially when there are more options than raw LLM output or human-written all the way. Most of the ways to improve reliability involve things like comparing to a corpus generated by the same person to see if their style changed (not always available) or looking for a couple basic indicators and jumping to conclusions (if you're searching for em dashes or "not just ... but" and calling it, you might be better off with the software). That's especially true if the short writing is also unusual; style of forum posts, exam question answers, sales copy, personal messages, open source readme, and formal messages to a customer are all very different even when the same person is writing them.

Those are also hideously unreliable. You can get plenty more data to test with, and you'll start finding correlations, and those correlations will often be correct, because otherwise the companies couldn't honestly promise they'd trained their detector. But you'll never have anything reliable enough to punish people who use LLMs if that's what you want to do with the result without punishing plenty of innocent people. If that's not what you want to do with it, there's almost nothing else you can do.

Fired IT worker jailed for 21 months after sabotaging old school district

Re: The elephant in the room

It's much easier to declare security poor when you invent facts about it. The spreadsheet was not, as you imply, one created by someone else which he copied. It is one he created himself. Collecting passwords is not that hard when you work in IT with people who don't. For those in support, ones who have proper admin rights so don't need to ask for anyone's password, do you think you could find a user who would give you their password if you asked them for it anyway? I think I could social engineer them from a number of people with the helpful asset of them knowing for sure that I'm an IT employee.

There are practices they could have used which would have helped with this, for example it doesn't sound like they had SSO access to external services and didn't reset all the shared credentials. Doing that likely wouldn't have kept him out of everything, but at least a few things like the social media accounts would have been fine. That was an actual failing. But since you made up one they didn't have, you wrecked a point you could quite easily have made successfully.

Re: U wot?

They were relocated from the prosecution, now "prosection". I found that a very weird typo, because while there are a couple more ones in this article so I think it probably slipped through editing entirely, the U thing is not anything a typo should have accomplished.

LibreOffice brands Euro-Office a 'de facto ally' of Microsoft's lock-in strategy

I hope you're right. I certainly don't know enough to say that Euro-Office won't continue to develop it. I have seen far more people happily announce that they've forked something who then didn't do much to maintain it that I am wary of new forks until they've got enough history that I can see they're going somewhere.

Re: or

Which is only useful if you can define what dragging each type of thing into each type of destination is supposed to do. If I select a chunk of audio from audio editor A and drag it into email client B, what should happen? Should A cut the chunk, put it in a file, and attach it to B? Should A attach the whole file to B? Should A be compressing either file since it's about to be emailed and the raw audio that you're storing in an editor is going to be quite large unless it's a tiny clip? But if B is something local, you probably don't need to compress it. If you assume that this is necessary for sending things*, which one needs to happen: A must know that B is an email client and implement compression, or B must know that the object from A is uncompressed audio and implement compression?

* That assumption isn't as arbitrary as the example seems. You can email uncompressed audio since it's just a file size limitation as long as your mail server doesn't have a limit. There are plenty of cases where conversion is less optional. If you want to embed an image file in a document format which doesn't support the format it's in now, either something has to convert it to a supported format, or raw pixel data is being sent and it's being separately encoded every time (a recipe for degradation.

Re: or

Not everyone wants to outsource mail delivery to another commercial service, a service that itself has been used by spammers, for something a mail server of their choice can manage. Mail merge is probably not great if you're sending a thousand messages, but a lot of people need to send messages to smaller lists, easily do so without hitting limits*, and don't see the need or value in a commercial service to do that.

* It's not the ISP anymore. The ISP has no idea how many emails I sent because the traffic over SMTP is encrypted (was that twenty emails or one with a big attachment). It's the mail system which wants to protect themselves against spammers so they don't have to wrestle with blocklists. Except if it's yours, it doesn't care, and if it's commercial, you can often prove to their satisfaction that the mail is not spam and get a relatively high limit. Not necessarily high enough that you can run a large discussion list off it, but large enough that you can send the occasional mass-mail without exceeding it.

Re: or

The principle is correct, but I don't think you need integration or common authors to make that happen. What you need is standard user interfaces, and we mostly do have them. We used to have them even more, but designers like changing things, and when they run out of good reasons to change things, not enough of them stop. Still, the email writing window in Thunderbird still looks like most text editing controls, including the document window in LibreOffice Writer, with the same buttons and keyboard shortcuts. Options are generally in the same places, and even the menu order and organization is recognizable from those in other programs. Both of those differ more from OS to OS than they do from each other since they both follow the interface design of whichever OS/DE they're running on (some DEs don't apply because there are a lot of them). I wonder if we actually need more than that.

Re: or

Tables should be attachments by default because why exactly? People have computers that can write in multiple colors, with emphasis, with tables, with headers with different font sizes, and somehow we're supposed to deny them the ability to use any of these when the documents are messages sent to someone else? Should we tell websites not to support those either? If not, why should email be different? Users want to do that, there is no good reason to tell them they can't, so it was inevitable one way or another. The only thing you would have gotten if people had only permitted plain text emails is a successful Microsoft email alternative in the 1990s that could do those things which would have replaced it for most people. Ideally an open standard, but the history of Internet Explorer suggests that, though it probably would have been, it wouldn't be a nice one.

Microsoft's worst 'Nightmare' unleashes BitLocker bypass 0-day

Re: Dog bites man?

"How about keeping anything of value off the device in the first place"

Sure, we can do that with the thin client approach, but there are reasons why people don't like and advise against that, with requirement on a stable connection (add "broadband without low usage caps" to that if the user does anything with big files on their local CPU) being the biggest. You can mostly remove that bandwidth requirement if you use cloud desktops, and Microsoft will be happy, but most others, from IT to finance to the users will not like it as much.

"How about making sure that in the unlikely event that a device goes missing you can block and wipe the device?"

That's kind of what Bitlocker or other drive encryption is there for. For wiping devices, you have three not great options:

1. If it gets connected to a network, you can wipe it, but to get it connected, you have to leave enough open that someone who got a machine can do that. That gives them a bigger attack surface for bypassing your protections before they connect it.

2. You have something which will automatically wipe machines that haven't checked in for long enough. You don't have to give as much access for that to run since nobody has to log in, but you have to boot to the OS or have custom firmware to manage it. Also, people will be unhappy if it goes off by mistake.

3. You put in place option 1, but you mostly rely on users not being able to get into the device in the first place by putting pre-boot blocks in place. It's true that a stolen device won't be wiped if they can't boot it, but it also won't be read.

Signal says UK plan to scan devices for nude images 'endangers us all'

Ah, that's reassuring, thank you. It's good to know that Signal, the very people whose warnings about the UK's law you used to support your "worse than 1984" case at the start of this thread, don't have a clue about Canada's law, and their theory that it is worse (they're threatening they'd have to leave Canada while not doing so in the UK) is based on bad information.

I'm glad we have someone so knowledgeable about the details of this law. Not knowledgeable enough that you knew what I was talking about with specific laws mentioned or could find out before deciding it was something unrelated, but you had a couple hours to learn all about it to come to your confident conclusion. A shame you couldn't have shared any of that knowledge; it could have given us all peace of mind that Canadian surveillance is nicer than anyone else's, but I'll assume that you definitely have as much basis for that as you did the rest of your posts.

"Interesting. The legislation in question has been leaked, and says absolutely nothing like what you are accusing Canada of. Rather, they're proposing age restrictions and national legislation regulating social media."

Do you have an allergy to reading? The law in question is C-22, it hasn't been leaked because it's already public, and it has nothing to do with age verification. The Canadian government calls it "timely access to information" (in their description on why this is definitely not warrantless surveillance which it is), and it's about access to data by police. What law are you talking about, and you know you would have identified what law I was talking about by reading either the article I quoted, my comment, or the post we both linked to which specified the bill by number, all of which specified that we were commenting on surveillance and not age verification? The law everyone else knew I was talking about does, in fact, do what I accused it of.

But in addition to once again ignoring things that were right in front of you, you did decide to bring some of that nationalistic smugness. While this isn't about age restrictions, you had opinions about those. When the US was doing that, you said "I have absolutely no intent of installing or using anything that helps the American attempt to build a global fascist surveillance network of "Recall" type intrusions on the operating systems produced on American soil.". When Canada does it, "You do not have a "right" to the Canadian market without following our legislation and laws.", which while legally true, applies equally well to the age verification laws from US states and the one you criticized from the UK in the start of this thread that I also don't like; pointing out that laws are binding doesn't make a point about whether they're good or even acceptable. You inaccurately changed the topic of this discussion to another law, maybe, I'm still not sure what law you were talking about, and you didn't even do that consistently.

What's your point? That through significant effort, you think you could circumvent this? Unless your opinion is that this makes the law unimportant, your confidence about avoiding it doesn't change whether it's a problem. Since I never argued that Canada's law would be perfect, unavoidable surveillance, pointing out that it isn't does not change anything.

It blocked us at 'hello!' Anthropic Fable 5 refusing innocuous prompts

Re: Boiling frog

"state censorship is rightly not considered acceptable": I'm not sure that's true, but I'll concur that, in democratic countries, you at least can't be honest about that and have to call it something else.

"at the same time, a bit like the boiling frog, we are slowly eased into accepting that private, unelected and increasingly unchecked authorities, can and do apply censorship."

That depends what we're talking about, but I mostly don't agree. Let's start with the article's topic, refusal to execute queries sent to a model. That's not censorship. A program rejecting an input isn't limiting your freedom to say things, it's limiting what that program does. I don't see any censorship risk in writers of software deciding what they'll not allow the software to do, and there are competitors who will be willing to execute more queries if you find Anthropics limits unacceptable and still want LLM execution.

There are companies with more power to prevent you from saying or doing things. The two largest examples are social media, which can prevent you from posting if they don't like what you post, and software distribution monopolies such as Apple's app store or Google's announced mandatory developer approval system (details mostly unconfirmed). Of the twoo, I care more about the latter because they are monopolies; if Apple refuses to let me distribute software to IOS devices, I cannot do anything else and expect owners of iPhones to have access, but if El Reg refuses to let me post on their forums, I can go to any other forum, including hosting my own, and anyone who wants to hear my opinions can follow me there without losing access to the previous forums. They probably wouldn't, but it would be their free choice.

We can't define any restriction as censorship. It depends a lot on what happens if the restriction does affect you. If Anthropic doesn't execute your prompt, you can run it on any other LLM, including plenty of smaller local ones, and in fact you're not restricted from modifying it and trying it again, which might work as their safeguards aren't very strong. There could be harm in some restrictions, sometimes enough that we need to regulate to prevent it, but calling them censorship when they're not will not help demonstrate or address that harm.