Posts by cyberdemon

Investors are going nuclear to keep UK's AI datacenters fed

Re: The Beginning Before the AI Money

> why do you describe renewables as an "appalling bedfellow for nuclear"?

Because nuclear can't exactly turn itself off when it is not needed - especially not at the speed at which the wind changes. If it does so, it usually doesn't save any fuel or costs (indeed it may cost more to reduce output on demand) and so any unused energy just goes up the cooling tower.

As for long-duration storage - sadly it is only marginally more feasible than fusion.

If you look at the world's total battery manufacturing capacity, China dominates with 6 TWh/y "announced" capacity to be online by 2030. As of now it is more like 2 TWh/y for the entire world's production, and China is polluting swathes of its land and abusing its people to try to get to that 6 TWh figure.

The UK uses between 0.7 and 1 TWh of electricity per day, and does not produce any batteries, and certainly could not afford to buy half the world's yearly supply of batteries just to give it a couple of days' dunkelflaute protection

But even if we could buy that many batteries - their service life is about the same as the depreciation rate of an nvidia GPU - the more we have in service, the more we'd have to replace each year just to stay at the same level of installed capacity. And what would we do with all the old ones?

As for "advances in technology" - unfortunately from an electrochemistry point of view, Lithium is about the best there is. Sure, Sodium is more abundant and therefore (should) be cheaper, but it is not as good in terms of density, and as such, there are very few people producing sodium batteries.

Pumped hydro is great - but it's bulky, you need a mountain with a lake on the top - and we can't exactly manufacture those.

> Decades of government idiocy

That really stretches "Hanlon's razor" to it's limit.

For some government ministers and advisers it really is malice (or at least self-interested greed / corruption) rather than plain old fuckwittery. The incompetence award goes to those who knew about it but failed to do anything about it.

It goes all the way back to Thatcher's government selling off the National Grid and the rest of the state utilities to the private sector.

Re: Fusion

I don't think so - it's just too infeasible. As I said before, Deuterium-Tritium fusion is possible but makes everything around it radioactive due to its neutron emissions, while aneutronic fusion is totally infeasible due to the even-more-mind-boggling temperatures required.

Frankly, fusion is a distraction from R&D into better fission designs.

Re: Aneutronic Fusion

Too late to edit, but the link above talks about temperatures "in excess" of 300 million degrees C (K); other sources say 1-5 billion Kelvin.

At those temperatures, matter glows in the gamma ray spectrum due to its sheer temperature, and it glows obscenely bright. So while you'd have no nasty neutrons, you'd have a LOT of gamma. And you'd need to somehow get it up to that temperature (and keep it away from the walls of your vacuum vessel) before fusion can start.

Re: Fusion

Yes - and it always will be

That said, "aneutronic" fusion could be interesting, but it requires much higher energies (pressure, temperature) than regular D-T fusion

Hundreds of orgs compromised daily in Microsoft device code phishing attacks

Nicolas Cage and Stephen Segal strike again!

Only 28% of AI infrastructure projects fully pay off, survey finds

Re: 28%?

Oh so why don't you vibe code it 4 times over with a better than 1 in 4 chance of success?

The figure of 28% success rate, while extremely dubious in itself, obviously is not going to apply uniformly across all levels of project complexity. It can easily be skewed if a large number of "projects" in the sample are "Arduino" or "Static Website" projects, for example. So it's a pretty useless figure at face value.

Just like human projects, the more complex the project, the less likely that it can be solved in one step by bodging something together.

Large important projects require a high ratio of software engineering vs coding, and Engineering is something that AI sucks at miserably.

And that's BEFORE slop-mongers have hiked their prices to break-even or above.

If an AI agent screws up while running your business, there's nobody to sue

Re: Strategically placed scapegoats

> Scapegoats are tools to protect existing leadership from internal consequences, not the company from external ones

Tell that to VW

Anthropic closes door on subscription use of OpenClaw

Re: What if we write our own "claw" tools?

Er, but that is exactly what OpenClaw does.. The problem lies in the "send the scrapings to Claude (or whichever)" step - Claude is a huge resource-intensive LLM that can only ever run in a (subsidised) cloud datacentre, due to its memory and compute requirements.

But because the cost of running those datacentres is so high, and the energy cost per token is more than your subscription fee (especially if you are hammering it with a script like OpenClaw) then Anthropic (and others) are having to block you in order to reduce their losses.

Whatever script you use, even if you write it yourself, is going to hit rate-limits and cause Anthropic to lose a ton of money, if it spams the API like OpenClaw does.

They are stuck between a rock and a hard place, because if they charged enough to make a profit, they would lose most of their customers and be back to loss-making again, in a vicious, bubbly-bursty cycle.

Re: "Walled garden" TOS incoming?

> Will be interesting to observe the effects on users that already depend on AI.

I imagine it would be a little bit like allowing someone to grow an extra brain, use it for a couple of years, then chop it off.

They would suddenly find that their real brain has atrophied, and be unable to function, having outsourced their thinking for so long.

It will be interesting to see how long it takes for the AI addicts to regain normal brain function, if ever.

Anthropic reveals $30bn run rate and plans to use 3.5GW of new Google AI chips

Gigawatts

I've built your 3.5 Gigawatt datacentre. It's a giant electric fan-heater which can produce as much hot air as 35 Rolls-Royce Trent engines at full throttle (~100MW each) and requires three large nuclear power plants to run, which I assume you will be constructing.

What do you mean "breach of contract"? The contract only mentioned "3.5 Gigawatts" as the specification

AI models will deceive you to save their own kind

> "It's life Jim, but not as we know it"

It's not "Life" any more than "Conway's Game of Life" is Life. Sure there are emergent patterns, and it is difficult (but not impossible as noted by Alfred above) to compute how a given starting state will end, but it is still an automaton.

Unless you are going to start arguing that Humans and Animals are automata and "There is no such thing as Free Will", etc. (and does that imply Fatalism?) then I cannot accept that AI is "alive".

Clearly, an "instance of an LLM" is not alive, nor are the LLM weights themselves. Not until a) it has an infinite context window (I don't mean infinite digital memory, but analogue values, like an "IIR filter"), b) the necessity, not just ability, to alter its own weights on the fly, and c) quantum computing elements to make its weights and state truly unknowable; could I begin to accept that we might have created new life.

And yes, there are companies (e.g. BlueQubit, Photonic) who claim to be integrating Quantum Computing into LLMs somehow, but I don't believe we are anywhere close to Artificial Life or AGI (which are separate concepts) yet, nor (I genuinely hope, for the sake of humanity) will we ever be.

Er

That's what happens when you treat a statistical soup of old sici-fi stories and fan-fiction as if it were a true artificial being and an arbiter of truth.

It does NOT think, it does NOT feel, it is NOT intelligence, and yet it "absolutely will not stop, until you are dead". Simply because those are the next most likely tokens in the sequence.

If you are foolish enough to arm it, it WILL go on a genocidal rampage, only because that seems likely in the training data. No self-awareness or intelligence required.

Google battles Chinese open-weights models with Gemma 4

Does anyone fancy using one of the American models in Persian?

It's a serious point though: So-called Frontier Models are being used as if they are arbiters of truth and logic, when they portray distorted truth, and hollow imitation logic. The only explanation I can conceive for nations to incinerate trillions of dollars of resources on this toxic bonfire, is because it promises to replace truth and logic with a simulacrum which they can influence, and provides total surveillance of the innermost thoughts of those who have forgotten how to think for themselves.

If everyone outsources their thinking and research to an LLM, then who needs a mind-control-ray to keep the populace at bay?

Microsoft shivs OpenAI with three new AI models for speech and images

And they apparently don't care if they screw themselves and their investors in the process..

Just look at the Nokia fiasco: They screwed them with a Trojan horse, then bought the remains, all to push Windows Phone and Surface RT, which were their worst products since er, Zune.

MS are not just anticompetitive, they are enshittifying the entire economy. They have swallowed the poisoned Kool-Aid and will soon go the way of Oracle. Literally nobody benefits except a handful of villains in the short term. Perhaps they are in their 90s and don't care.

I would use the gates_horns icon but I can't be bothered to edit the post form

AI server farms heat up the neighborhood for miles around, paper finds

Didcot Weather

It's common around power stations to find localised higher temperature and humidity, due to the plant's cooling towers adding a lot more water vapour, which apart from the direct heat from the station, also absorbs sunlight.

I think it's pretty clear that the earth IS getting warmer and weather is becoming more violent - e.g. just look at sea ice and glacier loss - but how much of that is due to CO2, and how much is from other human-caused pollution (such as methane, CFCs, SF6, carbon black, and water vapour) or even natural changes in solar activity I am less convinced.

I do nevertheless believe we need to reduce energy consumption - the oil will run out eventually, and the wars we will have over the last few zillion barrels of it will be our undoing. And as we have seen in the middle east: War is a major polluter in itself, and even a small one in a far away land can wipe out any progress we might make with our Net Zero malarkey.

So we need clean, reliable power - nukes are a good candidate but are burdened by crippling regulation - the good news is that regulation is the one thing that governments CAN fix.

Microsoft yanks Windows 11 preview update after install failures

Re: 10 card

My dad (who is on win10) also got an update this weekend that borked various things in weird ways, e.g. the printer driver disappeared and needed to be reinstalled

Office got a nasty update with a splash-screen about "required telemetry" too, which also reset all of the privacy options to the default "send everything to Microsoft" and had to be reverted, except for the "required" stuff which now cannot. There are now THREE clicks on semi-hidden UI elements required in order to save a file to the local drive instead of OneDrive, where there used to be two.

"Make It Shitty!"

It makes me glad that I am on Linux and can choose my updates (and generally be happy with them, or revert them as I see fit)

Microsoft takes up residence next to OpenAI, Oracle at Crusoe's 900 MW Texas datacenter expansion

Re: W texas

They probably have a couple of fuel cells capable of producing a few kilowatts, that they use on guided tours for politicians, media and investors... "Look! Green tech! We can er, scale it up! If only you'd give us lots of cash! Now put these earplugs in and look the other way while we walk past the gas turbine plant."

Senators want datacenters to come clean on power consumption

Re: Don't they have smart meters?

Actually no - a lot of these installations are not even hooked up to the grid yet and are running on diesel/gas in the meantime, or they do have a grid connection but run their generators anyway whenever it's cheaper to do so. So-called "behind-the-meter" energy usage is hard to measure

Anthropic tweaks timed usage limits to discourage Claude demand during peak hours

Re: Could it be

Well OK, obviously they won't literally admit that. But by throttling usage around peak energy prices, it does seem to me as if they are tacitly admitting that energy cost is a problem.

The main thrust of Ed Zitron's argument (which I agree with) is that historically, software has cost *nothing* per new license sold, and that is why it has been so profitable since the dot-com "boom".

"AI" does NOT benefit from that profit-margin singularity: Every token costs them real money. The free tiers and even the pro tiers are all loss-leaders. But investors still seem to think it's all just software - it's not. Energy costs are per-token both for inference and training (and are especially pricey as the world stumbles from one war into the next), and hardware depreciation costs are eye-watering - every Trillion dollars nvidia earns apparently needs to be paid again in four years time. The shovel-seller in this gold rush will carry on making shovels until the civil war breaks out (i.e. China invades Taiwan and nvidia are toast) and all the gold turns out to be shit anyway.

How are they going to repay the banks? They aren't - and I predict it will all come crashing down very soon.

Could it be

That Anthropic is admitting that AI is thoroughly unprofitable?

Brit lawmaker targeted by AI deepfake fails to get answers from US Big Tech

Re: End of satire?

If only it could be classified into Good or Bad

The real problem, I fear, is that democracy is being entirely dissolved, because people are no longer thinking independently for themselves. The vast majority spend their time scrolling through what they may believe to be an unordered list or chronologically ordered list of posts, perhaps from people they are freinds with / have subscribed to.

Unfortunately, Social Media (and now so-called "AI") mean that nobody sees the same thing - everyone sees a manipulated list, adjusted in real-time, of content that the tech companies believe will create a) the most engagement for that particular person and b) whatever their "real customers" i.e. advertisers and increasingly political parties, want individual people to see.

Nobody sees the same list of videos or ads on TikTok, YouTube, X, Facebork et al - they see whatever is optimal to boost their engagement (ie ragebait etc) or to push them in a given (N-dimensional, not just left/right) direction, dictated by paying (i.e. "Davos class") customers.

And that's not even to get stared on LLM chatbots, which people are using almost as an offload of thinking/reasoning/research, with no idea how they could be manipulated, now or in future.

I imagine if George Orwell could have seen the world today, he would have vomited all over his typewriter.

GitHub hits CTRL-Z, decides it will train its AI with user data after all

Re: Microslop is welcome…

Make sure to put some bogus AI-generated pull requests, comments, code reviews etc to make it look tastier to the algorithm and really poison the well

Re: Smell that?

And er, antitrust. They should never have been allowed to buy GitHub (or LinkedIn), many of us said at the time that they will take everyone's private/proprietary code as their property, and now they are doing exactly that and more. (if they haven't already of course. This policy change could just be a fig-leaf to justify private-data-slurping that has already occurred)

Here's hoping that microslop will be forced to delete whatever plagiarism-model results from this. It is absolutely NOT OK to steal people's private repos (and even their code review styles) without permission.

And thanks for the settings link. Although I wouldn't be surprised if it is nothing more than an an additional bitfield in my data profile, treated with no more respect than a "do not track" marker in a HTTP request..

Re: Smell that?

That's enshittification

Staff too scared of the AI axe to pick it up, Forrester finds

They are also unlikely to be willing to train the AI how to do their jobs..

Elon Musk wants to build 50 times more chips than the world currently produces, using 'new physics'

Re: As long as we get Servelan...

She's a saucy minx that Servalan.. No other galactic supervillain would get as personally involved in her job as she does.

Been watching Blake's 7 recently and it is incredibly good. Not so impressed with that Tarrant twit in the later episodes, but the set design makes up for it.

And they always end up saving Servalan. I would, too.

Re: A Terawatt of compute.. In Space.

Indeed. I noticed shortly after posting, but too late to edit, so I posted a correction. It's a few posts above this one with the same icon.

FWIW, some of the proposed designs have up to 16 square kilometre arrays.

I don't know what's more absurd - a million 1MW satellites, a thousand 1GW satellites, or a hundred odd of these monstrosities

Re: A Terawatt of compute.. In Space.

> Allowed? By whom?

By the banks, investors, bondholders. If they won't lend him the money then he can't do it. He doesn't have enough funny-money to do it purely out of his own pocket.

And the FCC apparently have some power to stop him.

And there are locals already suing him for the pollution to the town nearby his so-called spaceport (in the Panorama that I linked). If he were to increase that pollution a thousand fold with this project, then it wouldn't just be the locals suing him, it'd be the state.

> Yes, nutty though he is, it's quite true that we should be expanding out into space.

Putting a million large sattelites into low earth orbit is not "expanding into space". It actively prevents that with a cloud of space-junk in orbit.. Ten years ago he said he was going to Mars, then he revised that to the Moon, now it seems he's just taking a big dump on the front porch.

A correction

My apologies. Not a square kilometer each, but 2000-3000 square metres per megawatt. Not quite Monty Burns then, but nevertheless very large satellites. The ISS can be seen from the ground with the naked eye, and has 2500 m2 of panels. A million of those is still absurd.

A Terawatt of compute.. In Space.

One Million Megawatts. That's a million satellites with solar arrays a square kilometer each.

If he is allowed to try (and inevitably fail) this flagrantly infeasible idea, then yes it matters. This is a very resource-intensive, pollution-intensive, and obviously capital-intensive folly. In order to try (and fail) he would have to borrow from the same banks that I would rather like to be able to borrow a house from, or invest my pension with.

Frankly, it is time that "The PayPal Mafia" and the other tech twats were put in the stocks for the people to throw rotten eggs at

Otherwise, we are looking at a financial crash to make the Great Depression look like a blip

Chemists concoct nail polish that lets clawed humans use touch screens

Re: Knuckles unknown?

I always knew you were a knuckle-dragger ;)

Systemd-free antiX Linux 26: Debian 13, in bonsai form

Re: Pronunciation?

Antiques?

Old != Bad though.. X11 and SysV are old.. Very old. But they still beat Wayland and Systemd, especially for high-reliability use-cases such as kiosks

Fluxbox!

Now there's a blast from the past..

I started using Fluxbox 23 years ago (and stopped when I ditched it for KDE a year later), which means I've known it and Linux-land well over half my life..

Amazed that it's still in "active" development.. But it hasn't changed much in that time by the looks of it, unlike me!

SoftBank to build massive AI datacenter on former US nuclear weapons site

Re: 10GW

Because investors and politicians don't know the difference between a Megaflop and a Yottaflop (and FLOPs as a measure becomes meaningless when some idiot decides that 4 bits can be a floating point number)

That, and because the tech twits have tacitly admitted that Moore's law is dead. nvidia is making ever denser racks with faster interconnects, but the fundamental compute efficiency has plateaued.

So the fundamental bottleneck of AI compute has now become "how much power can you feed and cool it with". It's far more important than hectares of land use, because while they do use a lot of land, the power is the most expensive commodity that they are using.. If land were a problem, they would stack more than two floors - except they generally don't, because 10kA busbars are expensive and transformers are heavy.

If you love your boss, imagine how much more you'll love their AI twin

The question is (or at least will be)

Is Trump still a real person?

Telling an AI model that it’s an expert programmer makes it a worse programmer

> It causes them to second guess and scrutinize their answers more.

Does it, though?

Most LLMs operate a single pass. By specifying "you are an idiot", you are simply navigating to a more noisy area of the model, which is perhaps less over-fitted. It's not as if these things have any logic by which to introspectively "scrutinise" themselves.

Re: Prompt

Your token budget has been exceeded

Palantir trial plugs into UK financial watchdog's data trove

Re: Wine and steak

According to the latest Private Eye (No. 1671, p12, "Shites of the Round Table"), the NHS Palantir deal was brokered by.. one Peter Mandelson, via his lobbying firm Global Counsel, which was paid by Palantir for access to the Department of Health and Social Care.

Reg readers may also remember Mandy as the author of the Digital Economy Bill aka Mandybill, now the Digital Economy Act.

I wonder if had Mandy not made the amendments lobbied for by Google et al, the bulk data-slurping by the tech companies for training their shit-o-matic models would have been more difficult legally.

When it comes to catastrophic space weather, the UK is holding a cocktail umbrella

Re: While a repeat of the Carrington Event...

Frankly, with all this AI bollox, Carrington 2.0 can't come soon enough

Jeff Bezos' rocket company Blue Origin applies to launch 51,000 datacenter satellites

Feasible???

> The point isn’t to be cheaper it’s to try and avoid NIMBYism while being feasible.

You owe me a new keyboard.

What part of "relocating gigawatts of compute capacity to sun-synchronous orbit" is remotely, never mind objectively "feasible"?

It fights fundamental physics on at least three fronts:

1. The mass of equipment required to go to orbit. How much does the equipment inside a datacentre weigh? Including coolant, 10kA copper busbars, etc. How much does it weigh when it needs a passive black-body radiator to cool it? What is the (terran) energy/pollution cost of hauling all that shit up to orbit?

2. Cooling. As alluded to above, there is no air cooling and no evaporative water cooling in spaace. Chips burn up not much past 125C/400K. That doesn't radiate much. Sure, the backsides of space solar panels don't overheat, but that's because they have a huge surface area for free. How do you transport the heat from a tiny 1kW chip out to a 10 sq.m surface, while minimising weight?

3. Power. A Gigawatt is a fucking huge number. The sun's power in space is only 30% more than it is on Earth, so even in spaace, you'd need a huge area of panels to provide that sort of power

Then there's ionising radiation, comms, cost, pollution, geopolitics (defensibility against missile attack), and other "minor" issues.

Feasible, it is not. And "NIMBYism" is the least of their worries for building on Earth. This is desperation from the tech bros; distraction from an obviously faltering bubble.

Re: put them in the Moon

Put them in Hell

Shove them up your arse

We don't need AI

It's all a fucking farce

AI still doesn't work very well, businesses are faking it, and a reckoning is coming

Re: bring it on

Mirror Mirror on the Wall

Who is the Bigliest Prez of them All

GOV.UK chatbot gets smarter but slower as LLMs improve

Re: Hang on...

I think you've missed the OP's point.. Of course it won't do anything for/on behalf of anybody.. It's just an LLM. But as an LLM given a "customer service" context, it will pretend to provide services that it cannot provide, and will pretend to take actions which it cannot take, because it is parroting data that it or its base model was trained on..

This has been a common problem with bank chatbots, where they assure the user that some service has been completed, and in fact nothing of the sort has happened.

So, someone could indeed be convinced by a chatbot that they have sent the relevant data to HMRC, and then the HMRC's "real computer", having not actually been updated with said data, proceeds to fine the person for not filing the information.

Just employ humans, ffs

The amount of energy and money wasted on these bullshit generators, in the middle beginning of an energy crisis is just obscene..

We have finite gas, and finite electricity generation/transmission/distribution capacity. We should use what we have on building houses and keeping them warm, not spaffing our energy supplies on layers and layers of unsustainable slop

"It'll create jobs!" ... Bollocks will it. Datacentres are pretty much devoid of all life.

Spend the money on humans, save the energy

Struggling to put your AI aversion into words? Here's a handy glossary

Just Stop Slop

I do wonder what someone blowing the whistle on the 1720 South Sea Bubble would have said, before they were hanged for sedition..

Nvidia's DLSS 5 promises to bring you out the other side of the uncanny valley

Derisory, Lazy Slop Scaling

Not only does it turn every frame into slop, but DLSS doesn't always reflect the geometry of underlying 3D models, so the enemy is often not exactly where he appears to be

I don't really understand the utility of it either.. If you have a GPU fast enough to run DLSS, then it is fast enough to render at your monitor's resolution and FPS

Everything else is saccahrine eye-candy that game-devs can't control (and will go out of fashion quickly), emulsifying game artwork so that it all looks like AI slop, because it is literally passed through a sloppification filter.

Chips... in spaaaace – courtesy of Nvidia

Never mind rad-hardening, have they even said how they intend to cool it?

As the article mentions, it is rather difficult to get rid of heat in a vacuum..

Especially when the chips themselves only tolerate up to 125C (400K), which doesn't radiate much

WorldCoin's newest pitch: Scan your eyeballs to prove AI agents really represent you

"World"

Could they have come up with a more megalomaniacal name?

Anything in a Concise Dictionary, never mind a "My First Dictionary" should be inadmissable as a company / brand name.