Posts by that one in the corner 4456 publicly visible posts • joined 9 Nov 2021
Even in the Days of Yore, I'd've been thinking about a way to automate the keypress.
Which would mean a few days of fun, hacking something together* but that would be it. The more sensible Hannahs of this world get far more play time.
* yes, yes, these days all you need is a RubberDucky but back then...
Piffling.
The linked-to article gives no useful source for that number and if you do try to search for its references all you end up with marketing guff for and about the usual bit barns and LLM codswalllop. Which includes all the wastage of everyone duplicating webscrapings because businesses "need to guard *their* copy, it is so super-duper special compared to the feeble scraping *those* people did".
Totally ignoring the REAL sources of data that needs to be archived: Big Science.
Places like the LHC already have data storage problems, tending to hang onto only pre-processed data; which is fine IFF we can trust that the data reduction didn't lose anything important that we didn't even know to look for. The LHC is actually in a nice situation: they know precisely when to switch on the recorders[1] and they know what sorts of events they hope to see, so can put some boundaries on the filters, both of which reduce the amount of data needing storage for more than a few minutes or hours.
But Astronomy really, really needs to be able to record and store raw observations, noise and all, from as much continuous all-sky observing as is possible to do. In as many bands as possible. And be able to cross-reference it all, over as long a time-span as we can possibly manage. From the professionals - the data from the Square Kilometre Array, for one has pretty big storage needs! - but also amateurs who are running instruments scattered far and wide (synthetic aperture long base lines FTW - iff we can get accurate timestamps and raw observation storage; timestamps are becoming doable...). And before you pooh-pooh the usefulness of this, as some do, may I remind you of the increasing need to keep an eye on the big fusion reactor in our neighbourhood and our hopes to predict when it might get a dicky tummy.
What we are unfortunately missing still is long-term storage with a decently fast write speed.
Business data really does not need to be stored on any great timescales. But as they have the money, best just to humour them
[1] although they'll miss the resublimating thiotimoline events that occur just before they start the collider running
And they only need to switch
Look around a bit more, there are a fair number of people who do things like floppy recovery on an enthusiast level: if you can't get the cash to use a "proper data recovery service" you can use your (spare!) time instead.
You want to get a flux reader attached to your drive, rather than an "old enough computer" 'cos the computer will (very sensibly) throw a wobbly when it hits a bad patch and fail, whilst the flux reader will just keep on going (reading back the lack of properly organized flux - at that level, it doesn't care enough to throw a wobbly). Then you get to run software on your big, modern bog-standard desktop PC to recover what data is left.
Start by reading up on Greaseweazle.
Oh, and write up what you manage, successful or not, and leave us a link to that blog post :-)
The unfortunate side of this is that the "challenge" is seen to be coming from the notoriously unreliable chatbots, instead of our having a reliable and rational automated system that can ingest all the available data about products[1], accurately take your requirements[2] and selecting the best match for you *this* time[3].
Then we could have a world were "brands" were thankfully utterly irrelevant[4] and instead companies had to actually compete on the genuine ability of the products - and continual updating of how well they were keeping up the quality of the targets that their QC were being told to target!
Ok, this beautiful world free of Global Branding will upset a few people, including those poor souls who have been trained by Big Marketing to believe that there is some personal value they can accrue by wearing brand names on the outside of their clothes - deprogramming camps can be arranged for the worst afflicted. It'll also upset brand marketers, but they can just go and get a socially valuable job instead: we'll always need bin men and loo attendants.
[1] be they airline flights or plastic storage boxes.
[2] this would be the hard part, but wouldn't it be glorious if we'd been spending the time and effort over all these years actually finding out how to get solid requirements out of Users!
[3] including "features" like, where applicable, being able to go to a local supplier to see the product in person and make your final selections based upon "what grabs you", or just be able to pop the item into your pocket (after paying!) immediately. Ie the pleasant parts of going shopping.
[4] and, especially, getting rid of the bleepingly destructive (for the purchaser, being, well, defrauded) trading in brands: buying out a defunct brand name that had goodwill attached to it, then using it to flog tat until the public get wise.
But crossing the road is a useful, even necessary, activity.
Reddit really, really isn't anywhere near as important as being allowed to cross the road. And there are places in Australia where it is not allowed for anyone, let alone children, to cross the road.
> for the crack? (phonetic joke!)
No need for phonetics - "the crack" (spelt just like that) is English in origin; North East, The Shields area. Hence the name of the NE "what's on" listings magazine, simply called "The Crack". What you referenced is a neologism after the word became popular after travelling to the Emerald Isle and "needed" to have an Irish spelling, especially when written outside Irish themed pubs in the south of England!
<nasal_voice>
If you watch the episode, it is made very clear that we saw two distinct sets of faces, those of the younger versions of The Doctor as he was losing the contest and then, after a pause, as The Doctor gains the upper hand, those of Morbius - both sets started with the current incarnation of the character. The only way that "clearing up" works is if you believe that, for some unexplained reason, the images of The Doctor were suddenly replaced by those of an even *younger* Morbius before we get the pause - and Morbius's lives get started again from the current incarnation.
</nasal_voice>
OR we accept that that episode was made well before the "12 regenerations" idea came about - and the Canon is exploding*
* Plus, of course, well after that attempt at "clearing up", Canon is now that The Doctor was never limited to 12 in the first place, so all those pre-Hartnell Doctors are quite ok and the "clearing up" has been "cleared away". It also raises the issue that someone on Gallifrey still nows all about The Child and staged some special effects to cover up that Matt Smith's's Doctor would regenerate anyway. Which may have been The Spy Master, as he supposedly was the only one to find out...
OTOH the members of the Landsraad were not shy of facing off against each other, despite all being shareholders of and incorporated into CHOAM.
After all, we've had the idea of Battlin' Business Units for a while now, as each one tries to keep a larger portion of next year's budget; it is only a matter of time before someone gets around to scheduling the live fire tests on the same day as the company-wide morale boosting paint-ball outing. He may only be a Junior VP now, but that Harkonnen lad has his eye on the big prize.
> Captain: What The FUCK? I'm in a WAR and the canon is fucking misfiring!
Oh come now. John Hurt's Doctor wasn't *that* bad. Okay, if you count all the faces in Brain of Morbius then the whole "12 regenerations" thing was already shot by the time you got to him, putting the *entirety* of the New Who into question even before that Northerner with the big lugs appeared...
That is how the ill-equipped always tackle a steep descent: just start walking down without seriously examining the problem (and finding out what has happened to other people trying the same thing), instead of crawling and crabbing carefully across the landscape.
Soon you are running faster and faster, managing to stay upright and congratulating yourself on how well you are doing, you are a natural at this, fie to all the naysayers.
Now you are out of control, the momentum is too great to let you go in any other direction than straight down in the grip of a force that keeps you accelerating. Your feet are hitting the ground so hard the scree is flying out in all directions; over there a flying rock slashes an innocent bystander and in the other direction it has created a landslide, undermining the plateau from where you started, those same inevitable forces causing the previously solid ground to collapse beneath the onlookers above.
The bottom is a sudden jarring moment of realisation that you've just run out of luck. The best you can hope for is a quick end; otherwise you will lie there immobile, nothing to do but watch the lives of those you've affected come crashing alongside.
PS
In the above, for "you" please read "LLM tech bros"; putting that in every time would have ruined the flow.
> I doubt very much that military troopers need software to know where they're supposed to sit and store their equipment.
Because the load balancing calculations were all done prior to loading (where do think the checklists come from?) and the troopers were all trained to put their kit *here* and the you sit *there*. Anything that is out of the ordinary or totally unplanned - well, any military that goes out with its materiel unplanned or improperly stowed is due for a right bollocking at the absolute minimum.
I'd be more concerned about small plane operators & staff who couldn't do the basic balancing themselves: stories abound of island hoppers, even in the dreary old English Channel, not just exotic locales, where the passengers get quietly moved around (commercial flights being too polite to just say "all the fatties to the front on the left hand side, we're carrying a refrigerator strapped to the right of the hold" - the "hold" being the last two rows with the seats unbolted today).
PS
Can you give a reference for their use of "Windwows" software? We're sure it wasn't running on an iPad?
> Your correspondent is a member of an Australian scheme that charges royalties for reprints of news and other content and disburses them to creators who register their works.
The LLM scrapers are indiscriminate and, as they don't track how much each of their billions of nadans was affected by each bit of input, can not possibly provide fair attribution for their final output*. So are all the humans going to be compensated for the scraping done *before* this scheme comes into place or is the compensation done by, ooh, percentage of the total material registered by this scheme that has your name on it?
NOTE: the Australian scheme mentioned is clearly aimed at explicit REPRINTS - that is, the item was found by some means (e.g. a search) and then explicitly taken as something to reprint, with the attribution clearly attached. Trying to apply that style of scheme to chat bot results has NOTHING whatsoever to do with LLM scraping of copyright works: it could be applied to the proactive bots that go out and pull another copy of the work, in which case it not an "AI" issue but instead we are back to the old "Google is showing too much of my text in their search results" problem.
I'm all for compensating the creators, btw; it is just that schemes like this all seem to be working from the outside in and don't seem to match reality. But they'll make *some* payouts, especially to all the big players who are already rich enough to pay the lawyers, and because of that will allow the LLM operators to be able to say "no, no, we are paying, everything is fair now" whilst all the poor young artists are still left starving in their garrets, producing genuinely great stuff that won't even be recognised after their deaths because it has been hoovered up by the spiders before anyone else and the serial numbers filed off before you can blink.
PS
If the answer is "yes, of course you have to register with all these schemes to get your due compensation" how much is THAT going to cost all the aspiring new authors? Even if these databases are free to enter (so how are they funded? By taking too much of your compensation?) there are going to lots of them - and nice people who'll be sure to get your work into all of them for a "reasonable" fee.
* even when paraphrasing, or just plain quoting, a recognisable chunk of text: if it spits out "Nobody expects the Spanish Inquisition" did it get that by reading Monty Python scripts or from ingesting El Reg comments over the years? What if the LLM is found to always mention vultures whenever it produces that quote? BTW, the use here of a trivially small quote is simply to avoid (1) having to make this comment stupidly long and (2) to try to get the reader to ponder how often even quite large chunks of material are found outside of their original piece without attribution "because we all know where it came from" even out of context; imagine how much of a work you can ingest just by reading a fan site** that assumes you know the context... So no responses like "that is too short to be worth paying for".
** now you've got the idea, replace "fan site" with "professional platform for discussing serious subject".
> 10 year of email addresses? I doubt I could remember all of mine. I certainly don't keep a record of them all.
Guaranteed I can't remember mine.
Since beginning from my first personal 'Net connection with Demon, I've always used catch-all emails with a domain, so every registration, online shop, random interaction has used a different email. But as I didn't get around to buying "my own" domain until very recently, I've lost a few domains over the decade (and others before that) and therefore all of those email addresses. A few of the more useful registrations would be updated for continuity, but on the whole didn't bother (this is at least my fourth El Reg "identity" IIRC). And as there was no point in keeping track of email addresses that were no longer usable...
> Yes, that is what the tariffs are intended to address in the medium-to-long term
There was no medium- or long-term strategy evident when the tariffs were created, especially with that idiotic board and the whole "tariff the penguins" nonsense. It was all just knee-jerk, innumerate "punishment for not treating the US fairly".
Any claims now of there having been any "strategy" really need to be backed up, as they otherwise look like post-hic rationalisation and wishful thinking.
As the list of 18 books banned in Utah includes a recent Margaret Atwood novel, I shall take it as read that the other 17 titles are equally worthy.
Does anyone have an email for the Utah school board, so I can send them my thanks for making it so easy to flesh out this year's Christmas Wish List?
If X pays this (picayune) fine without complaint, under the illusion it is "cost of doing business", but doesn't change what it is doing, then we have the start of a winning streak against them: the EU can, as TFA points out, keep issuing "periodic penalty payments." Which can gently increase in both size and frequency, putting the "cost of doing business" on an upwards slope.
If the EU directly tries to "fine X out of existence" in one go, that'll get tangled up immediately and only the lawyers win. If the can get X to just pay up to get the EU out their hair then there is precedent set for being told to pay up the next time. And the next. And the next. And ... (as I can't see X ever understanding that they could actually start being transparent, accurate and/or honest, that is no longer in their playbook).
> you want the least possible interception risk.
Shining a laser through water is "least possible interception risk"?
You did see the photo, with the clearly visible line of displaced light showing where the laser is - and therefore whatever modulation it is carrying? And that is a nice clear test tank! Add a bit of real-world turbidity and everyone in the nearby spherical neighbourhood will be able to read it pretty much as clearly as the intended recipient.
Now, lasers in space are another matter, of course.
> people break the law by reporting on the activities of law enforcement
Are you saying you believe it is illegal in the US to report on any activity of law enforcement?
So newspapers, TV, bloggers etc etc, who report that police have raided a house, or been involved in a car chase, or stopped cars for no apparent reason, or solved a murder case, or failed to identify a body dragged from the river, or opened a new precinct house, or were shot at, or shot someone, or provided an honor guard to the President, or gave testimony in court, or been the subject of a court hearing, or clashed with rioters, or accompanied an entirely peaceful march through town, or sung "Galway Bay" - all those reporters have been and are still breaking the law?
If true, that would would make the evening news a lot shorter.
> Wiw, nowadays it is news when MS bothers to fix a bug?
When you compare it to the average story about Microsoft, yes, yes it is news when they *fix* a bug.
Like, oooh, making the User have an individual Xero account they log into.
Putting the charges for API usage onto the End User who initiated the the transaction, rather than the front end they decided to use. You know, like paying for the electric but not caring if the demand comes from an oil-filled rad, GPU rack or any other way of turning current into heat.
Treat APIs as commodities, *especially* when they are for access to something like financials, where (one hopes) the processing isn't aggregating data across multiple Users. Then the third party developers could create front ends that make Xero more useful & appealing (ref complaints, above), Xero gets more use and more income, the devs can use whatever payment model they are happy with.
Too simple? Too workable? Won't allow for Marketing's Big Idea Presentation to the Board at Xero?
That goes along with believing that copying an example diagram verbatim from a SIL manual will magically make it a 1:1 fit with your project.
Think of it as trying to make a 1:1 fit between TCP/IP and the 7 layer model, only instead of arguments over a pint about whether all the fields in this packet belong in this layer or that, there are real safety concerns if you can't 100% fit the two together.
Here is my foul and evil IP infringement story generator:
print "Tell me your character names"
let a$=input
let b$=input
print "What do the like to do?"
let c$=input
print "One day, " a$ " asked his friend " b$ "'Would you like to go out and " c$ "?'"
PS Okay, the pictures are a bit dull, but it is a start.
The Smithsonian should demand that the loaned items be returned and get at least one loaded onto a flatbed.
Drive it away, park out of sight - then offer to let Houston have it back, charging the full $85 million for the tricky transportation problem of turning the flatbed around. They can then use that money for something sensible, like giving the Shuttle a good polish for the benefit of the visitors (and sneak the remaining $84 milion back to NASA).
> The only valid use of AI is to separate the gullible from their money
There are loads of useful techniques that have come out of the AI labs over the decades[1] - and some have been vastly overhyped in the past, then "died" - but they didn't die, they are still around, being useful. Except that it is not the done thing to refer to them as "AI" any more, they are just "obvious everyday algorithms".
Or, as I've said before, "if it works, it ain't AI" (any longer).
[1] some of which are already mentioned in other comments here, but consider Expert Systems, Planning Systems, anything to do with Natural Language, Machine Learning (in properly defined and controlled domains), a lot of "normal everyday" image processing techniques like edge detection and everything leading up to Vision Based Systems like Face Detection and the Face Recognition (and not just by dumping everything into an LLM and hoping, which is being tried nowadays).
> companies should be forced, by law, to...
In too many places, the companies own the law.
So what we'll get instead is:
>> companies will force, by law, the OSS authors to provide the fixes for free; if they don't, they'll be imprisoned which, in certain jurisdictions, means they can now be made to work for free. And it will serve them right, they are just common criminals after all.
The "worrying the original might be deleted" is prevalent, making sure that you and your user's will always be able to fulfill the dependencies. That is quite sensible, really.
As for just forking to get it under your name to pad the CV: if someone does that and the employer doesn't even care that none of the clones have any real commits in them? They deserve each other!
Those same people ignoring the honesty box are also leaving rude notes, telling everyone your free cakes aren't any good because they don't come in a range of pretty boxes and aren't as tasty as the one you get in Harrods - and why didn't you put out enough for all 170 guests at their daughter's wedding like they demanded in yesterday's rude note?
Then one goes over and reads that linked to Reddit thread and sees the number of upvotes the OP got at the start, even though he is excoriated later on: want to bet the "profitable businesses and wealthy corporations" are the ones who upvoted (and didn't bother to read any more of the page!)?
I do not in any way disagree with you, but the word "ought" is feeling very strained at the moment.
> what have you to lose by putting up the paywall for support and fixes?
Your time, your energy and, given the jurisdiction this is likely to be in, your house and livelihood.
The guys and gals who are doing the maintenance are programmers who are already overstretched (that is the basic problem). Just to set up the tech mechanisms for a paywall may be something they've no experience in, so have to put time & effort into learning that (without any guarantee that the effort will have any payback; at least you can see your code working as soon as it is done). Then they are suddenly faced with all the brand new legal questions: are they now in any form of contract, having accepted the money? To whom and for what? What is their new tax situation? Have they missed something in setting this up? Do they need an accountant for this? Insurance? If some dev is done behind this paywall, does it stay forever behind it? Does the software licence even allow that to be feasible?
Are their new legal responsibilities going to clash with their current employment?
"But all those things are the natural habitat of a foundation and there are lots of software foundations around, it can't be that hard".
Great. But if one of those foundations wanted to take on this project they've had the chance to do so. And the "community" of freeloaders^^^^^^^^^^users who are complaining now have had all the time to help with that as well as all the time they've devoted to helping develop the codebase.
Unless... Are you going to leap into the breach and provide the devs with all the necessary to manage this simple paywall?
> .So much of the rot on Github exists purely for the aggrandisement of the ego of the person who decided to fork. Hey, I'm not going to add to an existing repo, I'm going to invent my own version.
Um, you do know how Github works, don't you? Or any project in any other (distributed) version-controlled system?
Hmm, looks like you probably don't, so there goes: if you have an idea for an interesting/useful change to an existing repo, you fork it, use that fork to make your changes (which involves lots of commits - all the careful stages in your dev, your new test cases, fixes to your fixes to your new functionality). Then you sync up with whatever changed in the original in the meantime and finally send the existing repo a pull request. They spot a couple more things to tweak. You go around the cycle in your fork. Repeat until the original guys are happy and pull your changes in.
Ta-da, you have just "added to an existing repo".
And have, necessarily, got a.n.other forked copy of that project sitting on Github. Rinse and repeat for all the people who have made contributions.
Sure, there are people who never get around to the final pull request stage. Of course there are. But at least they tried.
And there are plenty of people who fork with the belief they'll create something so different it warrants a new name. Many of which just peter out. But at least they tried.
Plus all the students who are, you know, learning by seeing if they able to make functional changes to a working codebase.
> We don't need a million versions of video encoders, we just need one, so don't invent a new one for your onanism.
Leaving aside all if the above, and more, if you *seriously* believe things like "we can get by with just ONE video encoder that will handle *all* use cases, past, present and future", then you know a lot more about onanism than you do tech and probably ought to stick to what you are good at.
> a Creationist would posit that God hates pickups, but that seems a tad heretical.
Well now, don't y'all know that Jesus drives a pickup truck?
This is ripe for all sorts of fun:
An app that use the tablet-side camera to see when the user has pretending to read the news whilst peering over the top to gawp at the girl in the seat opposite: fill the outer screen with the camera feed and an enhanced leer. Or just an "AI" video of the user quietly picking their nose hidden behind the tablet.
Puts up a bumper sticker "My other phone is a Huawei".
You lot can come up with "better" ideas, I'm sure.
Once you use an iPhone, even a second-hand SE, *everything* is about Apple.
Just be glad the patent on the Electric Fanboi Field hasn't expired yet, 'cos once it does and *every* brand starts to use it we'll all be hoarse from shouting "shut about your EFFing...".
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
