Posts by cyberdemon 3325 publicly visible posts • joined 26 Jan 2010
TfL already capture WiFi and Bluetooth MAC / RSSI data.. I have noticed the signs at Waterloo station
Will they now be linking your face to your phone? e.g. https://arstechnica.com/security/2026/08/new-surveillance-tech-links-your-phone-to-your-license-plate/
If you thought Flock was bad.. Now you can't drive OR take the tube without being tracked
Apparently only about 7-10% of utility water comes from reservoirs, at least in the south of England anyway. The rest comes from aquifers, which are like reservoirs except that we don't see them, nor do we fully understand how much is in them, how fast they fill when it rains, etc.
The more we take from aquifers, the drier the soil on the surface is, and it can become impermeable so that when it does rain, it doesn't soak into the ground - it instead causes flooding and rushes out to sea, without actually filling the aquifer.
If datacentres and their power plants are using utility water and evaporating it, then they are a continuous drain on the aquifers, causing desertification for potentially hundreds of miles around.
> I recall a several years ago some nonsense about synthetic textiles being IR transparent (the bollocks merchants probably meant UV) and that the cameras in phones are sensitive to IR
The bollocks-merchants in question were probably talking about far-infrared (thermal IR) which CMOS and CCD cameras are NOT sensitive to. You'd need a Thermal Imaging Camera, and the resolution on those is, er, pants.
> Augmented reality has its own spectrum, from the ghastly to the essential, even when you peel back the techbro hype
Er, what? Since when did we have an example where Augmented Reality is "essential"? I guess maybe for the people who absolutely MUST play Pokemon Go, because they are er, addicted to the damn thing? No, I'd say the spectrum ranges from the ghastly to the quaint curiosity, at best.
I get the point about "making the blind see" - my dad is 95% blind, but nevertheless he wouldn't be seen dead in these faust-glasses. He's 77, and most of his blindness has come around in the last 5 years, through progressive glaucoma. He's tried the "Seeing AI" app on his phone, but finds it infuriating/useless because it picks out mostly-irrelevant things and reads constantly. Having it strapped to his face would absolutely NOT replace the sight that he has lost.
The problem is that they don't actually "make the blind see" - they turn some gargantuan datacentre owned and operated by a creepy wanker in a bunker into an all-seeing-eye, which then tells the blind some of what it can "see". It's like having the devil on your shoulder and trusting him to tell you what is in front of you, while he spies on everyone around you.
Those who object to the faust-glasses do not object to the idea of the blind guy being able to "see", they object to the intrusion of zuckercorp's all-seeing eye into their private spaces. If it were possible for these things to be 100% local on-device, without storage and provably so (which of course is both technologically and economically infeasible) then nobody would object to them - but nobody except the blind would buy them either.
You can't compare them to dogs. A dog really IS 100% local "on-device" intelligence and it cannot record video. So we don't object to them. I would object to a cloud-connected "robot guide dog" for the same reason as I object to zuck's wanker-goggles. It's an extension of surveillance, it causes tech dependence and with it a "technology trap". I.e. what happens to the users when their newly-"essential" piece of tech suddenly becomes unavailable? Who cares, says Zuck.
If a technology which is entirely controlled by a third-party is allowed to become "essential", then we become enslaved by it, i.e. we become Cory Doctorow's "Reverse Centaur".
And why are grid upgrades that expensive?
Apart from wars and datacentres pushing up the cost of copper and electrical steel, there's another problem, at least in the UK: NIMBYs and Deform UK Ltd deluding themselves and the public that we can have "buried cables" instead of "pylons", without understanding the massive technical challenges with long-distance HV AC cables (and problems with HVDC too).
AC power transmission via cables is problematic because if you shove conductors too close to ground potential inside a cable, then they become capacitors. 400kV 50Hz cables are limited to about 50km before the reactive current (charging and discharging the cable's intrinsic capacitance) reaches about 50% of the cable's rated load. If you want to make longer cable runs, you need to install "static compensators" i.e. sodding great inductors, similar in both appearance and cost to grid transformers, every 40km or so.
Not sure why you got downvoted - your points are absolutely right. We have overbuilt intermittent generation, and are overbuilding continuous (base)load.
If a 1GW datacentre wanted to relocate to scotland and take advantage of a "very attractive new pricing scheme" to use up otherwise-wasted wind energy, it would need something of the order of 10 GWh of storage. For scale, a battery "gigafactory" is loosely defined as producing 1GWh of batteries per year. So a single 1GW datacentre would need the annual output of ten battery factories if it wanted to run on Wind/Solar alone. According to a random search summary, that would cost about $3bn just for the batteries. And batteries depreciate, just like GPUs.
Er well.. The modern ones are fission-fusion-fission. That is to say, they have a fission bomb at the core, which when it explodes, heats and compresses fusion fuel ie Deuterium-Tritium, which fuses to release an intense pulse of 14 MEV neutrons into a final layer of subcritical fissile material, which is what does the real damage.
So while it is correct to say that the only significant amount of Fusion energy we have on Earth (except the Sun, which is not on Earth) is a bomb, it's also correct to say we have never achieved a self-sustaining Fusion reaction, in a bomb or otherwise.
> Given the obviously quite vulnerable state of the modern internet, how, exactly, do you expect a CEO to prevent a successful ransomware attacks? CEOs already have plenty of incentive to block such attempts were it that easy.
Not requiring CEOs to "block" ransomware. (Though it is obviously in their interest to do so if they possibly can) Just asking, nay, commanding, that they DO NOT ever, EVER, pay a fucking ransom.
If the data is forever encrypted and lost, so be it. If the data is leaked on the "dark" web, better be prepared to pay substantial damages to all those affected.
But DO NOT, under ANY CIRCUMSTANCES WHATSOEVER, pay the scum-sucking fuckers, or you will be jailed as if you are one of them.
Here's my personal theory, arrived at after taking a "Mind as Motion" aka "neuroscience 101" module as part of a Cybernetics MSc, and er, some interesting mushrooms:
Our neurons, as I learned on my course, are kept forever on the very-edge of firing. This is done both by self-regulation of electrolytes within the neurons, and also the neurons continually optimising their connections to other neurons i.e. synapses, so that they don't fire too much or too little. If a neuron fires continuously, it runs out of energy, but it adapts itself to need a higher threshold before firing to avoid doing that in future, or else it would die. If all your neurons fired at once, you would have a seizure, and perhaps die. When they do fire, it is because that incredibly finely-balanced chemical mix inside the neuron, influenced by all of its synapses, has tipped over a threshold that causes a positive-feedback cascade to occur, turning it momentarily into an electrochemical battery and passing that pulse along to connected neurons, which in turn only fire if they get exactly enough input from other neurons to tip them over their self-determined thresholds.
They are extremely complex systems, so much so that our biggest supercomputers couldn't fully simulate a single neuron at molecular scale in real-time. So my professor claimed in 2013, at least.
Now, because these neurons are so finely tuned by self-regulation to be just on the very border of firing or not-firing, and because they receive multiple inputs from different synapses that each add an analogue signal (i.e. the concentration of various electrolytes inside the neuron) - the difference between firing or not-firing could be down to the presence of a single molecule - and the difference between the production or non-production of that single molecule at a synapse could be down to an unknowable "Quantum" property e.g. the spin of an electron.
Thus, in trying to stay exactly on the infinitesimal border between firing or not-firing, the biological neuron could be optimising itself to amplify quantum properties. And complex behaviour of a unique system that is governed by unknowable quantum properties is something that could perhaps be called a soul.
LLMs on the other hand, are identical buckets of bits running on deterministic silicon. Every instance of an LLM is the same pile of statistics, i.e. the same pattern of connections and weights, with randomness added only as noise in one or more of its layers. Even if that randomness could be quantum randomness, it cannot adapt its own weights on the fly, and both its weights and state are known. If you could show me an LLM where each INSTANCE had its OWN weights, that were all quantum-unknowable and self-adapting on-the-fly, then I might be able to call it something like alive.
Sure, but every time the pen-testing agent finds a bug, it is passed to the "software engineering" agent, which is supposed to work out how to fix it, then passed to the coding agent to implement the fix. Then it is passed to the testing agent to evaluate the fix, etc. Each of these agents have a non-zero probability of getting it wrong and introducing a new bug. Or indeed, inventing a bug that wasn't there, while the "fix" introduces a real one. So at some level of system complexity, the number of bugs in your system will reach a plateau where your AI bots are playing whack-a-mole with themselves.
Meanwhile, your token spend goes through the roof.
These models are so amazing, so powerful, so God-like that we must immediately start spaffing a trillion a year of taxpayer subsidy to keep them solvent, or else we won't have our AGI Super Weapon and CHINA will get there first!
> Ilia Kolochenko, CEO of ImmuniWeb, said at least some of the incidents appear to be "part of a well-orchestrated marketing campaign" and argued that the reported "escapes" were simply the result of poorly isolated test environments rather than models independently breaking out of their sandboxes.
No f*ing Shit, Sherlock!
Also, the "Zuck wearing his pervert-glasses" image was brilliant, if nauseating.. So far the only thing Zuck has got out of his AI venture is to expand his Petabyte Personal Porn Collection. Presumably he is the only person in the world who watches porn in a VR headset, which is why he thought it fine to spend 80 billion on VR. Now he is spending 130 billion (a year) on AI..
The sooner zuckercorp dies, the better. The guy is an utter creep
Er, no, hospitals have a very useful purpose (whereas AI on the other hand, even if its output were free of cost and had zero resource impact, has questionable benefit to society and plenty of negative consequences), and roads require little maintenance post construction, i.e. do not require a constant stream of resources (gigawatts of baseload i.e. fossil or nuclear powered electricity and billions in replacement GPUs) just to keep them open.
And both are (traditionally at least) funded up front by the taxpayer anyway, so your argument is nonsense.
There may be Other Reasons why only 7% are getting built.
Perhaps because it's a toxic bubble, a Big Tech equivalent of the South Sea bubble, where every Megawatt they build costs them $40 million in capex, but they continue pouring money down the drain (perhaps because their bosses have succumbed to AI psychosis themselves) despite the facts that a) no bugger wants it! and b) operating profit margins are in the negative numbers! i.e. every megawatt they bring online just loses them more money faster.
Not all of the UK grid connections queue is datacentres, 2/3 of it (80GW) is. 7% of 80GW is still 6GW, or two Hinkley Point C nuclear plants, which apparently cost £50 billion each and take 15 years to build.
But if Ed Zitron's numbers from TD Cowen are to be believed, then a 70MW datacentre costs $3 billion to build and to fill with nvidia hardware. That's $42 billion per GW (almost twice as much as Hinkley), or $214Bn for the portion of the 80GW backlog that you/Reuters believe could actually get built. (with almost none of that money ever touching UK pockets, naturally).
The reason that only 7% get built is not just that they are speculatively submitting applications - it's because it's technically and economically impossible to build them any faster.
What baffles me is WHY do these tech companies continue to pour resources into a money furnace that just consumes more and more money the bigger it gets.
When the bubble inevitably collapses, it will leave behind a depression of staggering proportions, and we will be left with noting but a pile of incredibly expensive yet worthless junk. The pikeys will have a field day with all the heavy copper busbars, but that's about it.
A deposit can still be forfeited, that's generally the purpose of a deposit. It doesn't make it a fee.
E.g. you put a deposit on a house to secure purchase or enter a rent agreement. You get it back if you are a genuine buyer and complete purchase, or if you are renting you get it back if you return the hose in good condition. The estate agent's fees on the other hand, aren't refundable.
There's also a serious problem with these projects requesting absurd power requirements (to placate investors and contracted third-parties, to whom they have promised "gigawatts") and then only 50MW ends up being energised. That's a massive waste of resources for everyone involved and (as I put in my other post) should rightly be penalised.
Also, having a linear scale is too weak for discouraging AI slop-farms. The fee could for example go with the square of megawatts, e.g. £10k * MWs^2 would mean that a "modest" 1MW datacentre only pays £10k to join the queue, a 50MW barn pays £25m, while a 500MW megafarm must pay £2.5b and a 2GW hyperscale behemoth pays £40b. This may seem harsh, but it reflects ACTUAL TECHNICAL REALITY in maintaining grid stability, i.e. the grid and each of its sub-networks must be able to cope with the sudden loss of its single largest element.
This is similar to what New York have done (i.e. outright banning datacentres above 50MW) but is more flexible, removing the risk from government and placing it on the tech companies instead.
If any is to be refunded, it should be done so similarly based on how much IT load is powered, i.e. if you applied for a 1GW connection but only built 100MW, then you only get 100MW worth of refund i.e. 1%
What's that? Your bubble is bursting and you can't pay for the other 900MW of GPUS? Oh dear, What a Shame, Never mind. Thanks for your contribution to His Majesty's Treasury.
> The fee would be refunded once the facility is drawing power, or forfeited if the project exits the queue early instead.
If it's refundable, then it's not a "fee". It's a deposit.
Frankly I'm amazed that there isn't already a per-MW deposit payable - no wonder the UK grid connection queue is choked to death with bogus applications! If spamming the process can be done at no cost except to NESO/DSOs/Ofgem i.e. the Tax/Billpayer, then unscrupulous companies will obviously game that system to both speculatively submit applications (e.g. multiple sites for the same project), and perhaps even spuriously, to nobble their competitors.
It should NOT be fully refundable. At the very least, the cost of processing each application needs to be recovered, with an additional levy to support infrastructure upgrades. If there is anything left over, then refund that when a project is complete.
Also, will this be applied retroactively to the 125GW (3x the UK's peak winter demand!) of mostly-bogus projects (2/3 of which represent datacentres) already in the queue? They should be made to pay at least a portion of the new the fee in order to keep their place.
Having just watched Newsnight, with some ex-OpenAI prat gushing about the Hugging Face hack where, *shocker*: other companies were "hacked"!, and Oh I said this would happen, we were working on superintelligence and we all knew tgat One Day it would escape its guardrails and start thinking like a superintelligent human but a thousand times faster!! Oh we are so close to creating Humanity's Children in the next two years, or some bollocks.. This tech is so powerful and oh so dangerous in the Wrong Hands .. blah.
Then in Other News.. There has been a massive selloff of AI tech stocks!! Oh Noes!! The Chinese will get this Superweapon / Artificial God first!! Before We do!! We must pour public money into the poor struggling trillion-dollar hyperscale tech companies of Silly Con Valley so that this doesn't happen! It's a matter of er, Sovereignty! Or the Economee! or something. "AI will solve poverty and hunger!" (Yes that one was an actual quote...)
I despair, how utterly brazen the AI pyramid scheme has become.. That they can spaff Trillions up the wall on a loss-making slop factory and then come cap in hand to the taxpayer, fully expecting to be handed another Trillion, because "we must win this arms race!", when in reality they are a massive failed venture, a busted cult, having made a blurry simulacrum of intelligence that in reality provides nothing more than a randomised mush of concepts that have gone before, with all the efficiency of a belt-fed blunderbuss trying to hit a target a mile away.. It will cure Cancer! It will solve its own energy problems cos it's er, Super Intelligent!
Will it bollocks. But coincidentally it's all GREAT for surveillance and for silencing one's political enemies!
The HF hack was blatantly a publicity stunt to instill Fear and Awe at the "power" of AI.. But how they can con so many supposedly clever people both baffles and sickens me..
Missed the edit window but wanted to add:
Some generals/politicians/despots probably think that the above is GREAT and THE BEST WAY TO WIN, but what they will eventually realise (when it's too late) is that the symmetric war i.e. between different "factions" of AI-deathbots, only starts when all of the humans and human infrastructure (i.e. the targets programmed into the heuristics) are destroyed.
All of this is without any intelligence required, never mind any sci-fi superintelligent self-aware singularity. We already have the technology to arm a rover/drone with a machinegun and have a classifier with no more intelligence than that which we merrily plonk into a local-facial-recgognition CCTV camera deciding where to aim and fire. This kind of shit has already (apparently) been deployed, but only on asymmetric killing-fields such as Gaza.
So-called "AI" lacks intelligence, logic, reason. But it is great at profiling, pattern-matching, fuzzy inference. Great for assymetric warfare, aka genocide.
The only feasible "fully autonomous weapon" is a genocide machine, one that picks targets based on heuristics such as skin colour, facial appearance, object classification (what is that this human is carrying? 30% TV camera, 40% bazooka)
It can't be bargained with, it can't be reasoned with. It doesn't feel pity, or remorse or fear and it absolutely will not stop, ever, until you are dead!
Er yes, a lot like the South Sea Company
and the resulting crash when everyone realised that a) it wasn't profitable as slaves still needed feeding so their labour wasn't "free" (and AIs need powering, big time), b) slave labour was a poor substitute for domestic skilled labour, and er, c) the whole idea was a fucking abomination.
> All other lift and lay of PV sections can happen in front and behind the section of way being maintained at no additional delay to maintenance (providing that neither lift of lay functions take longer than the physical maintenance step)
...
Wallace & Gromit would like a word!
Yes.. "The Free Market will solve our engineering problem!"
Will it Hell
It just adds a couple more layers of Bork to an already Borked system.. "Prices" do not necessarily reflect the load on the local distribution network i.e. the substation at the end of the road or the bigger one it's fed from. What happens when the servers that set the "prices" are unavailable? How do you protect it from deliberate cyberattacks such as the one the Russians recently attempted in Poland?
> "experts suspected that whoever was behind it attempted to disrupt communication between renewable hardware and power distribution operators."
> From ramblings I've heard where I am, since I don't have an agreement for feed-in with the power company, if I "run the meter" backwards, they will likely cut power. The meter won't run backwards to begin with as it's programmed not to. They would fit a new meter that would if I signed up for feed-in.
You may have just stumbled on the "real reason" for the smart meter rollout right there. So that they can either a) pay you 5p/kWh for feed-in while charging 50p/kWh for consumption or b) pay nothing at all for feed-in and/or cut you off immediately if any feed-in is detected
I have so-far resisted having one. Mine is one of the old-fashioned GEC "bake-o-lite" electromechanical meters with the spinning disc and mechanical clock. Naturally it "pays" me exactly the same rate for feed-in as it charges for consumption. It's like having a free, infinite-capacity battery! Pure delinquency of course as far as grid stability is concerned, but my solar array is small enough that it's pretty rare for i to ever actually move backwards, and my rate of consumption is all still net positive between readings..
Alas they will at some point forcibly replace it, on the grounds that "its calibration has expired". Of course, they could easily come with a calibrated check meter to test and re-adjust the calibration - but obviously they aren't going to do that. They might offer a "dumb meter", but it would just be a smart meter with half-hourly billing turned off, i suspect it would still disallow feed-in.
I have a perfectly serviceable dirigible to sell you
- Generalfeldmarshall Paul Ludwig Hans Anton von Beneckendorff und von Hindenburg
Jokes aside, Hydrogen is ridiculously hard to store, it leaks out of -everything-. It seeps through stainless steel and gradually makes the metal brittle over time as it does so. Liquid Hydrogen in cryogenic dewar tanks would be impractical on an aircraft for numerous fairly obvious reasons e.g. since it needs to be kept at around 20K and isn't significantly denser than its compressed gas form - so storage would be compressed gas at 13 Bar, where as mentioned it will leak through tank walls, valves and fittings to create an explosive atmosphere in any air-filled voids. So either those voids need filling (adding weight, complicating maintenance), or you need special (heavy, costly) tanks, valves and pipework..
Inevitably it would be paired with electric fuel cell propulsion rather than gas turbines, for investor Wow factor (and the genuine benefits of thrust vectoring afforded by electric propulsion, and er, ease of integrating alternative power sources for when the Hydrogen system fails), but all this adds even more weight, cost and complexity that would make it doubly impractical and uncompetitive with both hydrocarbon-burning jet aircraft and battery-powered short-haul electric aircraft
Hydrogen powered cars, aircraft, and the "Hydrogen economy" have failed repeatedly, because it's just a bugger of a molecule to store
> The latest models are trained on the output of Russian bots reposting Truth Social AI memes
Well, the latest models are merely updates i.e. "fine tuning" or sometimes "distillations" of the previous ones.. I'm not sure that they still train them from scratch, and if they did, they would certainly include the datasets that they gathered pre-2018, and actively avoid any post-AI training data, for fear of "model collapse" - If you were train an LLM on purely "modern" text e.g. from the past 5 years, you would end up with a model of a blithering idiot.
e.g. they are hoarding old books (and butchering them - cutting off the spine and feeding the pages to a multi-page scanner, destroying the books in the process - because that is the only fast way to do it) because they still need more slop-free training data
https://news.slashdot.org/story/26/07/21/206243/ai-companies-are-buying-tons-of-old-books-because-theyre-free-of-ai-slop
But yes, of course the man who has read all the books in the library will tend to be more left-wing than the man who has spent all his time on YouTube and Facebook, or worse: TikTok and X.. But perhaps humans can suffer something like "model collapse" too - i.e. if all you read is AI slop, then you too will become a blithering idiot.
Purely coincidental that the orange nitwit announced a "Kill Switch" for AI, cos they are so powerful, so very dangerous in the wrong hands, that Donnie must have a special red button (hopefully not to be confused with the other red button) to "turn them off"..
More likely it is just a stick with which to threaten any AI company whose LLM is guilty of emitting Liberal/Commie/Anti-Donnie token sequences!
And of course for OpenAI it is both a hype boost and a means to an end, if they fail to meet the financial standards for an IPO, they can look for a government subsidy to help pay for their 1.4 trillion of debt promises, as they certainly aint making a profit, never mind one big enough to pay off their obligations.
So they instead stage an incident hyping fears of an AI singularity, and suck up to The Donald and offer him a Fisher-Price steering wheel in exchange for a trillion dollars?
Or they could quote it in units of actual output i.e. 77MW is about 6000 continuous slop-streams of fake Veo2-generated video, according to my fag-packet (based on my reading that a 5-second clip uses 1kWh, so an hour clip might be 12kWh i.e. a continuous fake nigel farage would use 12kW to talk endless slop on GB news without sleeping, and with an attention span of 5 seconds)
Then, they could consider the "value" of said output, and the value of slop is plainly negative..
Certainly, you wouldn't pay back a Trillion Dollar loan by selling the services of some lazy and unethical human slaves.. (Who need feeding of course - about 100 Gigawatts and a Billion litres of water a day to keep these lazy unethical nitwits going)
What would you call one of those Solar MC4 connectors then, which has a "plug" with a "female pin" inside, or a "socket" with a "male pin" inside?
Anyway, a "plug socket" is simply thus: A socket for a plug to go in. As opposed to a socket that attaches to a socket wrench, for example. Or a socket that I can open in UNIX..
May other states follow suit
But of course the Slopmongers will use this to try to explain/blame away their staggering losses, when really this was entirely their own folly, throwing Gigawatts and Trillions at AI in the psychotic delusion that they were "building God"..
And before anyone says "but but but datacentres are needed to provide all the modern services that we all rely on": Sure, build <50MW datacentres. Only slop needs Gigawatts. "50MW should be enough for anyone"!
