Re: "he had therefore paid tax on it"
Because he had just been taxed on a "bonus" that was of zero value to him (a gift card that won't buy anything vaguely desirable).
5493 publicly visible posts • joined 9 Nov 2021
From a very stunted list of providers, which has strangely neglected to mention Zen, the provider I currently use (about whom I have, to date, no complaints about at all). Not does it mention A&A, or...
Maybe this is a list of providers who go over a particular percentage of complaints? Putting Virgin as "the least bad of the absolute worst"?
Not strictly true.
A crash is a landing that didn't go to plan.
If you plan to drop from Lunar orbit in only a spacesuit, after giving yourself a hearty retrograde shove out the airlock/docking port, screaming "For The Glory of Trump" all the way down*, then you'll have made a landing without a lander.
It's all a matter of perspective.
* It could take a while, bring water.
> No idea why anyone would've downvoted on yours.
Because he reminded us of the cypherpunks* list in the 1990s.Which the downvoters just found too embarrassing to recall.
Can you blame them? There they were, all dressing like Zero Cool at the end of Hackers (they'd just over their Case phase and were hiding the "I heart Wintermute" tats under the long sleeves) when suddenly cyberpunks* were out and the cypherpunks became The Thing. Now they had to ditch the glitz and neon, the roller skates and arcade games, to dress only in black hoodies and those damned dark glasses (no matter how many times they watched the Blues Brothers it just didn't work for them). Even worse, all too soon it went mainstream and they had a confusing "Gordon Gecko in striped shirts but still somehow in his mother's basement" moment.
Now they've all grown up and those are just the shameful memories of youth, all behind them now. Damn it, the arcade games were fun!
* if you were going to, the best group to ally yourself with were spending time in secretive groups that sounded like pirates and talked in a jargon inpenetrable to outsiders, all about "varietals" and "acidity regulators": the ciderpunks.
> They tout that the thing may be ready in 2027, but nuke approval wouldn't be until 2030
TFA> "We expect the power generation portion of the system to be available as early as 2027, with the nuclear module being available to customers about 2030 based on regulatory approval."
(Electric) power generation hardware available by 2027; tested and working, ready for more to be manufactured as required.
> What do you do with it for 3 years while you wait and hope for a yes?
Put the generators already made into a warehouse; if someone wants to buy one to test or use with their own heat source, quote them post & packing.
> when a search user's query gets answered on the search results page, so there's no need to click through to a source website.
If you are lucky.
OTOH a zero-click search is what happens when it is obvious from the search results that you're not going to find *any* website worth clicking on, so either try to change the search phrase (again!) or just mutter rude words and give up.
The latter zero-click situation is, to my eye, occuring with increasing regularity, especially as the search engines are trying to be "clever" and assuming that you want to hear about whatever is the "happening thing" today if it can possibly be confused with your search phrase. Plus, of course, all the sites that are copying each other (or using the same LLM to generate the text), so you see the same couple of lines of snippet time after time.
So are those two different reasons for zero-clicks ever differentiated in these sorts of studies - or am I simply the most unlucky web search user, an outlier to be excised from the reported data set?
> Each manufacturer has wiggle room on who they implement things.
The tying of BT and WiFi information to the Location services has been consistent across all the Android devices I've encountered so far.
See also support pages such as:
Why does Android require Location Permissions for Bluetooth?
Why location services must be enabled for Android users
and so on and so forth.
These are all reporting on behaviour baked into the OS, *not* something that is varying from manufacturer to manufacturer.
And on our phones, including the Motorola purchased this month, to allow webdav syncing to be enabled only when connected to our home WiFi (e.g. allowing davxv to see the SSID) requires location services be enabled.
> I keep Wifi, BT and data off when I'm not using them. I also don't load apps willy nilly
Bully for you - in other words, you just do what (probably) the rest of us here do! Fairly basic behaviour for any sane user. Doesn't alter what the OS is doing and when a User *does* need the functionality (in so far as *anyone* "needs" any of the things the 'phone does).
> . I have other SIMless phones for different things rather than compromising my daily driver.
Well, that is going above and beyond (and is not an extravagance that Joe Random User can afford) - although, don't get too smug: SIMless keeps you off the phone network's locators but the devilish device is still talking to its true masters over the WiFi...
Even if you have allowed the storage of your location data by Google, you should be able to switch off location tracking by the phone itself. Say by tapping on the icon in the settings.[1]
Except that tapping that icon then prevents access to full Bluetooth and WiFi functionality. So you either switch location back on, including the GPS[1] or you don't get to use some pretty basic stuff[2].
We[3] all know WHY this happens[4] and it is insidious. And, of course, it gets worse over time. To the unbounded joy of all the wrong people.
[1] and drain the battery even more, yay
[2] you want a program to know the SSID of the AP you are already connected to? Turn on location! Of course it is location data, APs are all glued to the ground! What, you have a pocket NAS? Deviant! The cue the arguments that SSIDs are "trivial to change, you can't trust them, of course you validate by using GPS" yada yada; all very collaborationist.
[3] except we El Reg readers are a privileged lot; read Amazon etc reviews for gadgets and toys that present WiFi or BT control and you'll find confusion about "why are you snooping on my location - my child's location - when we are using this LEGO RC car?"
[4] just in case you don't, Google - and the other lot - are using you to wardrive and map the locations of every WiFi Access Point and BT beacon "for the convenience of their Users". Ho ho.
Do you manage to get the fork into them?
I can never manage when they have non-integer Spin: I have to wait until the Gluons have cooled down and the meat is firmly stuck to the plate (after which, as Feynman showed, the plate only wobbles half as fast as it all spins and is therefore easier to hold down on the table).
"WILLIAMANMARY for some reason was known as The Orange in their own country of Holland, and were popular as King of England because the people naturally believed it was descended from Nell Glyn. It was on the whole a good King and one of their first Acts was the Toleration Act, which said they would tolerate anything, though afterwards it went back on this and decided that they could not tolerate the Scots."
I hold out hope that the engineers* will quietly attach a smoke stack, ball regulator and a steam release valve onto ATLAS.
If only to see the look on Brian Cox's face when he fronts the live re-opening on TV, the curtains are drawn back and he gets to see all the latest additions for the first time.
* aka postgrads, fresh postdocs and other navvies
I'd like to see the Standard Model fitted with the Fairchild Truss Flange: that was able to prevent Gravity's attempts to topple the factory chimney, Big Tom, so should be sufficient to tame the G-beast and wrestle it into the fold, alongside the quantum ducks*
* You know, the ones that go "quark quark"
> USB cable only plugged in when there is a conference
Or use a USB hub that has an individual power switch on every socket (one example); I can enable/disable the webcam, mic, BT[1], WiFi[1], random devboard etc etc at will, without having to plug/unplug all the time. Even reboot my ancient and sometime recalcitrant keyboard (*still* on the todo list to rebuild that, or at least check all the solder joints...)
[1] each on separate dongles - this is for a desktop, not laptop
> You need a lens to focus the incoming light...
From a (very shallow) reading of the Nature article that TFA links to, this device appears to act more like a recorder of holographic images than focussed images - a hologram doesn't need[1] a lens for its creation nor for viewing.
> let alone be able to take quality selfies.
The light-emitting mechanism appears to be the inverse of the light-detecting, so to make your selfie viewable you'd send the data from sensor to emitter and ta-da. But sending from your flashy new Fourier Pixel camera & display over tour friend's tedious old Android OLED is more of a pain.
This does bring up interesting questions about how much processing would be required if you did try to use it to create a JPEG! We can calculate the hologram for generating simple images (it has been possible for a long time to calculate those, draw them using a pen-plotter and then photoreduce them to the necessary scale allow a light-based holographic image to be viewable) and complex ones "only" requires banging away at the same calculations for longer. Inverting that, calculating what the source was that created these patterns, is - harder[2]. And then you are left with a model of that source from which you have to infer the result of "normal" 2D focussed light process (raytracing!) and then you can get your JPEG to feed into your mass-surveillance Big Brother system.
Of course, you don't have to directly analyse/interpret a recording for it to be useful if you can just play it back via an inverse physical system (e.g. we didn't - generally - try to pull data directly from old wax cylinders, wire recordings or VHS tapes: instead, we just played them, recreating the output sufficiently that humans observed & understood. So the lack of anything as "useful" as a bitmap coming (almost directly) out of a sensor does not make it any less of a camera. And no doubt the same ML techniques will work just as well (!) for facial recognition direct from this form of data as it does from JPEGs, once it has sufficient data to train upon.
PS
To anyone else who understood that paper *way* better than I did (not a high bar): please feel free to correct my total misapprehension, but - be gentle!
[1] some setups do use optical elements, but not for focussing light from the subject onto a focal plane, and not all setups require them.
[2] however, I am woefully out of date - been ages since I talked to anyone about this and I freely admit the maths in the Nature article flew so far over my head I could hear it scaring the pigeons on the roof - so if anyone here happens to know that such calculations are mere pocket-calculator stuff these days it would be great to know about.
True - neither LLMs nor any other form of ML is "intelligent" in any way that we like to ascribe to ourselves.
BUT the large part of that line of argument comes from the usual misrepresentation of a field of investigation and the name it chose; misrepresentation that is inevitable given the behaviour of humans in certain fields[1].
So we have people who misrepresent "the Big Bang"[2], "Evolution" - and the modern favourite, "AI". People who work on the Big Bang are not looking at/for any kind of detonation[3], evolution does not come in flavours that you can pick and choose[4].
The field of AI (as it was described[5] when I started taking the classes) is about looking for ways to create useful, or at least interesting, processes that can be applied to perform, or aid in performance of, behaviours that appear to require intelligence. That is, to find any ways we can MIMIC those behaviours; there is good reason that the Turing Test is also referred to as the IMITATION Game. When we have some machinery that does a decent job at something it is made glaringly obvious that it is ARTIFICIAL.
Ok, there are confounding factors, such as "if it works, it's not AI (any longer)"; i.e. the tendency for AI Lab creations to be applied out in the Real World and become mundane and (shrug) "well, yes, so what?". Remember, Back In The Day, playing chess was something only humans did (it was one of the "open challenges" in my book of "100 programming exercises", way after "print the Fibonacci sequence, with and without recursion"[7]) and now we have Toledo microchess born from an IOCC entry, a far different goal. It would be nice if people gave the attribution ("I'd like to thank the Academy for this award, and the tireless AI boffins, without whose heuristic search my runtimes would have been unfeasibly long"), hey ho.
Other confusions: people just throwing around "smart" and "intelligent" around willy-nilly (home automation, 'phones). And a scarily large number of people who seem to be really, really angry at finding out that the human system is "imperfect" (i.e. has limitations) and get upset at finding out there are mechanical methods of doing things they find hard or impossible (thankfully, most now seem ok with the pocket calculator) and/or that they can get so easily tangled up by interactions with simple mechanisms (oh, ELIZA, why, why?). Or, so much worse, "simple": "I am better than 200 lines of code!". Um, have you considered all the human intelligence applied to allow those 200 lines to exist and execute?
My conclusion:
Trying to split up or do away with the phrase "artificial intelligence" is as sensible as trying to be rid of the phrase "artificial hip": no surgeon or biomechanic is trying to claim that that lump of titanium is a "real hip", any more than any (non-misrepresentationalist) AI bod is trying to claim[8] that software is "real intelligence". But both AH and AI are doing a sterling job helping someone/something to run.
[1] basically, anyone who believes they can make money by misrepresentation: salesmen, hucksters, fraudsters. Oh, and philosophers - and even more "philosophers" - who keep trying to muddle things up by trying to imply they have anything useful to say about either physical reality or mathematics; for money.
[2] all in quotes to indicate their misrepresentation; without quotes, these are all perfectly respectable fields.
[3] there are a *lot* of those to study, but they all came a lot later! And we should all be very glad that they did.
[4] it doesn't cover abiogenesis at all and the attempts to create a split between "micro-evolution" and "macro-evolution" are complete twaddle.
[5] probably in a lot better fashion than I'm going to manage here, but also over a longer period of time: like any introductory class, we started with flashy[6] demonstrations before getting on with it
[6] sadly not as flashy as the rocket scientists get; but then they do have to take their maths more seriously, as getting it wrong tends to do more than waste a chunk of runtime, so let them have their fun.
[7] so, did he buy that as a first edition the day it came out or pick it up years later? Ah, the mystery deepens.
[8] yes, yes, you get them all "tired and emotional" and they'll probably wax philosophical; which is as useful to reference as the fact that you're my best mate, you know that, I love you, I really do (clunk).
"The right tool for the task" was the heavy-duty geometry modelling, the continual incremental improvements in x-ray imaging to detect the ink placement - and the substrate placement to feed the geometry modelling...
As TFA says, a full workflow, pulling together best techniques from physics, biochemistry (materials) and maths.
Which created a flattened image of the marks upon the unrolled scroll.
Finally, the well-understood use of Machine Learning was applied to aid character recognition - generating a result that can be verified by humans looking at the highlit portion of the image.
No doubt ML was used to look for good discriminators for signal analysis, which discriminators - the output of those MLs - are running within the x-ray machinery.
All of which is good - ONLY you MUST be careful to differentiate between continued use of (by now) "old-fashioned" and efficient/verifiable ML versus the current hype of equating "AI" with "LLM".
Hang on, that's not right.
John and Virgil (plus optional passenger) were delivered by tube to TB1 and TB2, which (aside from some comics adventures, that required weeks of significant upgrades by Brains) were Earthbound.
Any space-going crew were delivered to TB3 by a very comfortable sofa.
> never been a huge issue
Well, not since the desperadoes stopped riding in the next car from the Western Union bullion van, watching for the flash up in hills as the train passed through empty scrublands on its way to deliver payroll for the mining town.
Thankfully, there was less of that in the UK, although there have been reports of piracy on the mainline.
> In order to operate the brake you'd have to be sitting in the one seat where that is.
Nah.
Give each passenger access to a brake pedal. Only when *all* the passengers feel safe about being driven away and all release their brake will the vehicle be allowed to move off.
During travel the amount of "reluctance to kep going on" is the sum of all the passengers' pedal pushing; if this reaches a quorum then it becomes brake pressure applied: if you all scream and stamp down together the car stops!
You may scoff, but how many times have you, as a passenger, found yourself stamping on a brake pedal that isn't there? And that is with a human driver you (hopefully) can reason with* about their driving.
* "reason" can be at varying volume; quality of language may go down. You marriage may be at risk if you don't bite your tongue.
> Stripping down the ... driver repository
Needlessly preventing us from running peripherals we own, creating yet more e-waste.
> modern instruction sets ... without hardware emulation bottlenecks
If *you* happen to have that hardware, it will be used without "bottlenecks"; emulation (more likely, simply a second copy of the library with different build options) are (should be) already only installed on kit that requires them.
> Memory (RAM)1 GB
Then you try opening Edge (guessing this OS is stripped of the ability to bloat itself installing a.n.other browser) and open up Microsoft's latest SaaS: better be ready with those RAM sticks!
> (Idles at 400MB)
Ya know, if we take off the wheels from your car, you can still start it up and happily tell us how efficient it is at idle; good luck getting in the weekly shop.
> We can dream ?
Of an OS that no longer runs the useful software that makes owning a computer actually useful?
> So, I asked Ai the question
Ah, silly me, this whole thing was just a joke in the first place, wasn't it!
We still need the spies, on both sides, to infiltrate the running dogs'/cunning devils' secretive hunters and bring home the latest details on security holes that have been found in our systems.
Only this time, instead of using the information to arrange little accidents for the moles, a vast widespread response is required, one that reaches across the globe and touches so much more than the assorted security services' personnel. Now, the superpowers are forced to patch all our desktop software to stop the Other Side's encroachments.
Although this does raise the SPECTRE of increased Corporate Sponsorship of espionage, the proper stuff, not that penny-ante "industrial espionage": is the most cost-effective route for Microsoft to locate bugs in Excel to train their own Matt Helm* or to hire Blofeld? Well, we all know management would rather spend money on contractors than risk increasing their own company's warm body count...
* singing optional
> Many paid models can be run on local hardware and often for free.
For certain specific values of "free". I.e. those that discount capital cost to get the kit capable of running a non-trivial model, the cost of the energy that kit requires... Oh, and ignores all risk that the license might change on a whim do that the paid model now has to be paid for even when run locally.
At least when discussing the average piece of "free to own a copy of" software it'll run on kit that you almost certainly already have to hand (and which will still be worth maintaining after you've finished running that software as it is suitably sized for many other common tasks).
Annoyingly, I don't have the URL, but there was a fairly recent comment - I fear from an El Reg commentard and not one of those ghastly other places where one sometimes slums, such as Reddit - when the subject arose that developers were suddenly finding that they were running out tokens faster than they expected and this was causing them to spend more.
This person had described a typical tale of woe, where, in one sweaty working day, making a few bug-fix edits had barely touched his token budget. But, to his shock and horror, the next step, getting the same LLM to generate documentation for his finished project, emptied the coffers.
How could this possibly happen? The bug fixes were awfully clever and tricksy things to do, but the documentation was mostly just a load of boilerplate and dragging out factoids* from the comments, any fool could sit down and churn it out by the gallon.
But in amongst all the commiserations and similar tales (plus a bit of the good-natured joshing these forums are infamous for) the question never answered was: just how big was the text of the generated documentation compared to the size of the corrected lines of code? And documentation for the entire project, not just the portions containing that day's fixes? The only useful response** would have been "Just because YOU think that the k lines of bugfix represents a lot of clever thinking whilst the k * big_number lines of documentation can be cheaply spat out by any random useful idiot, that does NOT mean that the LLM agrees, that it sees one as cheaper than the other to generate; you are paying by the token which essentially means by the raw amount of text, irrespective of whether some of that text is more 'interesting' than the bulk".
* factoid == something that has the outward appearance of a fact, but isn't one (cf human/humanoid, cube/cuboid, Anne/Annedroid); an apropos word for so much concerning LLMs and their output.
** yes, I missed out on the opportunity, drat the need to ever take a break from constant monitoring of the Interwebs!
We had plenty of end-of-year parties, even - shock horror - ones with two or three different classes* invited at the same time!
Just no need to go overboard on the whole thing: it's was a normal shindig, albeit a bit bigger and way more inclusive than the average "hiring the room at the back of the community centre" birthday do. If you wanted to encourage dressing-up it'd be less "Under the Sea" and more "come in fancy dress".
* are we supposed to call them "cohorts" nowadays? That always sounds a bit - conspiratorial. The entire cohort was in cahoots.
What? No howling wind that lasts for minutes, threatening to pull our heroes into The Void? No gale that only increases in strength (the deafening noise drowning out the depressurisation klaxon) until finally the leaking barrels of radioactive materials are lifted from their pallet and sent hurtling away before exploding at a safe distance?
Have the historical documents been lying to us? Next you'll be telling us that Gilligan escaped the island (those poor people).
And no longer redirects straight to the history /page.
But today it reads like a bad Max Headroom imitation, repeating itself itself itself, some stories appearing up to three times for no obvious reason (no doubt, each one is a fresher take in the subject?).
Still, at least the /archive page has regained the links to the comments (and even the pictures), which had vanished recently, prompting the peek to see if /week was working any better at that point.
A Loop with more realistic scenarios and far fewer chances of anything unexpected ever happening or going dreadfully wrong.
"We found it and we'll let you have a look at one - if you can afford the licensing fee and sign the NDA*"
Get a couple of gullible governments into a bidding war over the patent and ka-ching!
Not our fault if the winner dropped the entire World's supply when they struggled to get the child-proof lid off; we have sworn testimony that "all of the Majorana Particles currently in Microsoft's possession were poured into the labelled container".
* including the "if I couldn't see it sitting there it is because I forgot to bring my reading glasses, mea culpa" clause.
Except for those that do; then again, the tyresome lot are just metro rail, not mainline services.
PS
before they spill their weak lemon drink in the rush to get to the keyboard, yes, "normal" train wheels do have tyres, just not *pneumatic* tyres: the outermost layer(s) in contact with the rails. And those can even involve rubber layers sandwiched between the metal (which have had their own problems...). And metro systems (can) do a sterling job in improving mass transit and making cities much nicer places to be, didn't mean to strike a nerve by suggesting they are "lesser". Look, I just wanted a chance to coin "tyresome" in response to the pavel's punning use of "tire", ok? Get off my back!
PPS
pavel.petrman - have an upvote, we don't see around here very often.
> We note that Codex, presumably running GPT-5.3, reviewed this particular series of commits. That makes it all the more surprising that the code was so ill-conceived.
Excellent.
Now, class, I hope you were all paying attention here: always put in a twist at the end. Stab, twist.
"US companies are the best and it is only fair that they dominate across the World, we will fight you if you say otherwise"
"US companies shouldn't be made to pay taxes commensurate with their earnings in each country, that isn't fair, we will fight you if you say otherwise"
Hope we are all keeping track of the US gunboats: now Trump has given up on peace-making he'll be trying to speed-run across the World the ideas from the Perry Expedition to the Enola Gay, in order to make an impression for the mid-terms.