Re: Perl didn't like even line numbers
.. but Only Once...
1717 publicly visible posts • joined 10 Jun 2009
That's horrifying to someone who learned programming languages in the last century. It's an approach that a pathologist would take with an organism that shows observable features, but has biological processes that the pathologist does not fully understand. "Let's see, that anomaly is due to this protein, dunno which gene encodes that one, dunno quite why it's expressed right here, umm, maybe the problem is hormonal, or there's another organism involved, or a virus, or..."
Need a (lavatory | rest room) designer from Bethselamin. Award yourself another pint of medication if you don't have to click on that.
Data are only personal if they can be assigned to a living individual identifiable by the Data Controller[1]. I don't have to worry about taking pictures outside Bath Abbey with dozens of faces in them, because although I will process those face pixels, I cannot identify them with a living person. So cameras surveilling the store shelves and aisles are probably not collecting Personal Data[2]. When you stand at the payment till, swipe your loyalty card and Smile For The Camera, then that image is certainly Personal Data. As sergio says, the Privacy Impact Assessment which should have preceded the deployment of these systems, would have to address the privacy impacts and mitigations.
[1] Data Protection Act 2018 UK Public General Acts 2018 c. 12 PART 1 Section 3 para (2)
[[2] It is the very act of bio-measuring the face and collating that with a list of light-fingered ne'er-do-wells that makes some captures into Personal Data when there's a match.
That's a really excellent summary of a saga that had us appalled and fascinated for years. Many of us transcribed documents for PJ at Groklaw, and I remember that I did one in which the Novell executive who had done the deal with TSG (previously SCO) had been deposed. I can accurately recall my summary for the Groklaw post: "Sure, we sold Unixware to SCO, lock, stock and barrel. Except for the lock and the stock, of course, and we made sure that we kept control of the barrel."
That case was never about patents. SCO alleged that it owned the copyright to Unix, and that IBM had violated the copyright. It turned out that SCO didn't own the copyright, only a gig to collect licence fees and keep some commission. It was always a lame attempt to extort some money from IBM (whose lawyers were, in those days, referred to as the Nazgûl, so good luck with that) and possibly to derail Linux, as in the episode where SCO offered to sell licences for Linux.
The sorry tale was documented at Groklaw, and I was about to put a link in, but groklaw[.]net seems to have been taken over by someone punting crypto casinos. My day has just become a bit more dystopian.
It's interesting [1] that the use-cases for these things often mention "elder care" by which I suppose they don't mean removing the elder tree weeds from my garden, or harvesting the elder-berries, but more like calling an ambulance when Grandad falls over again. What they don't seem to mention is the use-case that Jake cites, i.e. looking after young humans that can't look after themselves. Is this because we are inherently more protective of our children than our old folk? Would we sooner hand off the Grandad-wrangling than the care and feeding of babies? Personally, I find both distasteful.
[1] Interesting to me, you may think otherwise!
Yes. Plz to build in safety mechanisms from the beginning. While at the moment SpudCells have to be fed a very specific diet by more complex life-forms, those life-forms don't have to be biochemists, or even humans. If a SpudCell co-opts other life-forms (probably bacteria or fungi) to meet its requirements, then Grey Goo time isn't beyond belief.
PS I'm not sure I understand 'named Potato Cell as a nod to her Polish heritage'. Potatoes are staples in lots of places, but didn't start in Poland.
> the Vatican, which is a nation in its own right
It doesn't have a seat in the United Nations, though, so it may not be a nation by some definitions. See MapMen: There are NOT 195 countries [youtube.com]
Good point, well made. It seems that nothing can be written these days without intensifiers. The spokesbeing did not say that he appreciated the concerns, he said that he "appreciate[d] the hell out of [them]". I play a little game with broadcast news interviews, tallying the times that perfectly ordinary things and events are said to be "incredible". Where do you go when you need a word that means "incapable of being believed"?
...Elon Musk is now worth more than $1,000,000,000,000
My usual definition of 'worth' is 'the amount a buyer would willingly pay'. Now, what would someone pay for a fully-indentured Elon Musk, I wonder?
If by personal worth, however, one uses the definition 'could afford to spend this amount of money', then I still don't believe the headline is supportable.
It was a point of pride amongst a certain class of APL programmer to get all the code for a simple-ish programme onto one line. The result, though was what came to be known as write-only code: even the author couldn't easily tell why it worked after but a few months from writing it.
See, this is where it's easy for the legal and regulatory frameworks to make a positive contribution. It is an old, old feature of common law that Masters are responsible for the actions of their Agents. Even HUMAN agents cannot have the blame heaped on them exclusively, much less any given AI model, which has not got any initiative of its own (due to being a product of artifice).
Many lawyers are learning at the Courts that having AI draft submissions and correspondence and then not checking before submitting or corresponding is a Bad Idea, because the Courts are in a position to make sanctions. What is the ordinary user of AI going to do when misled? They ought to be immune from being told "Sorry, the AI did it".
Just another example [lawgazette.co.uk]
> find a big enough USB key, check it has space, copy the file onto it, eject it, go to the other machine, insert it, and copy the file off again
I'd put a sha256sum -b the_file | tee checkit before the copy, send checkit over with the_file and do a sha256sum -c checkit at the end. Just to be sure!
Looks like a backdoor: as in, a conveniently wide gap in the wall, disguised as brickwork, but with a hinged panel that opens to allow anyone who knows about it to get in.
If it's an accident, it's a remarkably well-formed one. Exploitation (or invocation!) needs no infrastructure, and depends on the RE boot sequence paying attention to the System Volume of a device on the USB hub, and keyboard signals. What's it doing that for?
> drop the storage device into another machine and exploit, profit
This doesn't work for most 'casual' implementations of BitLocker disk encryption in Win11. In those myriad cases, the Key Encryption Key is stored in the non-volatile memory of the TPM chip on the motherboard of the original host device. The encrypted disk is not readable on another machine, without that KEK.
> They're not. They're just bugs.
But in other coverage of YellowKey there are reports of bizarre behaviour - the exploit needs one to plug in a USB stick with the malware, and then to hold down Ctrl while rebooting into the Windows Recovery Environment. This is already a very convenient mechanism, dropping the user straight into a highly-privileged command line environment with the disk decrypted. Also, the exploit files disappear from the USB stick after a single use. The case for it being a deliberate backdoor is made more fully in the README on GitHub.
> ... no current AI can ever lie. They have no way to be aware.
True but not useful. The AI appears to its users to be aware, to understand. We are so horribly bad at assessing another's [1] motivation and credibility that huge numbers of users just assume that the stuff coming out of their laptop (or whatever) is the product of intelligence. The Turing Test sounds terribly quaint and old-fashioned nowadays.
[1] In 'another' I include all sentient species and the electronic simulacra.
I thought I'd reply with an explanation for those terminally deficient in the sarcasm-detection department.
$ chmod +s weights.bin
will set the SETUID bit, which means the file will execute with the permissions of the file owner. If that's root, well, it would open up a big attack surface. If as Jou suggests the owner:group is nobody:nobody then it would be much safer. That user and group is present in /etc/passwd specifically to be the least privileged.
I also have been visiting El Reg on an almost daily basis, for the last 18 years, via the /week page. This accounts for a late response, because I thought the 404 message indicated a problem that would be fixed!
In a minute I shall go looking for the post I made the last time that the El Reg team mucked about with a perfectly adequate user interface. I mean, who cares about the number of articles that can be above "the fold"?
My key user requirement is a page which quickly lists El Reg articles, youngest first. I've learned to live with the slightly greyed-out text to indicate 'visited' instead of the red that The Week In Summary page used to have (once upon a time).
The list does NOT need a silly picture to accompany the headline, thank you, as featured on the Archive page. Who thought that an earthworm picture (12K bytes) was a useful addition to "Worm rubs out competitor's malware, then takes control"?
> What is concerning is the amount of trust people put in it.
On this forum, everyone knows at least something about how LLMs work, and the extent to which they can be trusted. In Real Life I have come across multiple people who have swallowed the hype, and are entrusting the weightiest life and health decisions to chat bots.
One man told me that he didn't need professional legal advice for his issue, because "Grok is PhD level now".
Someone is going to get badly hurt.
An interesting tale of two different AIs leading (or driving) two different users into full-blown psychosis: Under the Influence of AI: The Documentary [bbc.co.uk]