Posts by breakfast

DEF CON dingus suspected of trying to take over Delta in-flight Wi-Fi

The old truism

Fun fact- the old truism about what happens in Vegas is pretty young - it was a slogan coined to promote Las Vegas to tourists back in 2003.

Signal adds an extra layer of security to make sure you're actually chatting with the right person

Re: I really really skimread

The blockchain is useless, but it's a kind of ledger and those have been used in many contexts for centuries.

Malicious SIMs can shut down phones, steal files, and drag 5G back to 2G

Is this impressive?

If you don't want to go to the effort of finding a malicious SIM you can also drag your 5G back to 2G by signing up to 3.

‘Humans will be a rounding error on the internet’ says Cloudflare exec

Re: it's already started

I think most of the comments are from people (AManFromMars excepted, of course) because what's even the point of an AI joining in all of this nonsense? But I did notice some interesting voting patterns on comments on an article about Chinese models last week.

How the famed USENIX Security conf is managing a flood of papers in the AI era

Re: Show your work.

That would be worse though because AI is not very good at translation, because translation is a hard job. If you don't have human translators with a good knowledge of the field available, and the budget to pay them, a translation that the original authors have approved at least suggests it conveys their intent accurately as far as they can judge it.

Latest GitHub outage squeezes Actions, Pages to death

Uptime targets

Github's push to achieve that elusive three-sevens uptime continues.

Microsoft tells engineers to curb their token-burning enthusiasm

Re: Technical debt

My fairly minimal Go back-end is an order of magnitude easier to maintain than my Javascript UI that relies on libraries that make breaking-change version jumps every six months. By this point moving the UI to the latest version of the core platform it's built on is almost a full rewrite. I don't think the people excited to add features to their tools really consider that not everybody using them can afford to be spending resources on constantly upgrading and migrating to stay current.

Always be burning tokens

Last year Microsoft were telling their engineers that it was more important to be using AI than doing a good job, and presumably the ones who didn't decide it was time for early retirement have taken that to heart.

It's still kind of astounding to me, though, the way they're determined to shovel technical debt into their products like coal into a steam engine. It will necessarily and obviously have repercussions down the line and the effort to undo that harm will be vastly more expensive and time-consuming than doing things well the first time would have been.

Bypassing AI guardrails is so easy a script kiddie can do it

Re: So ...

That's why they can't - regular salaried workers know why the ideas are unworkable/illegal and keep the organisation alive by stopping them doing the dumbest and most illegal things possible. Replace them with a digital yes-man and your company might be dead before you can find yourself a viable golden parachute and jump into the next vastly overpaid CEO job.

The AI bubble is already popping; we just don't know it yet

Not to be the LLM-Defender

I agree in general, but although I don't like LLMs but they're not entirely without use. They're not good, but they're not blockchain.

They can be a great way of generating content-shaped data for your tests, they offer new ways of mapping the usage of language across large corpuses, in a business scenario they allow you to eliminate some needless tasks (using the heuristic that anything that can be done by an LLM does not need to be done.) There are uses for them, but they're are mostly weird edge-casey things and none of them come within a hundred billion dollars of these companies' valuations.

It doesn't impact your main point at all and they're clearly a dead-end that cannot bring us closer to AGI, but even cul-de-sacs can sometimes have their upsides.

Microsoft makes Copilot harder to miss in Classic Outlook

But... AI isn't Classic

I feel like "Classic" implies "like the old days," which surely implies not having any AI features.

When they start offering "classic" versions like that, they'll have my attention.

LinkedIn realizes its users have been bathing in AI slop, offers a shower

Re: Maybe I'm just being cynical

You are being cynical but if you look through the details of how Linked-in are planning to use this feedback, you're also 100% correct.

Word worm crawls into Copilot, spreads chaos

Re: Keeping code and data separate

AI Guardrails are less actual guardrails than a red velvet rope with a polite message saying "please don't cross."

Signals are grim for Jodrell Bank Observatory after UK science funding decision

Re: It isn't just physics

I don't know if that's actually true, but as we haven't had a left-wing government in my life time and things aren't looking good for the future, it's hard to be certain.

UK's seventh prime minister in a decade says he'll ditch digital ID scheme

Re: Person of Interest

They struggle to get credit because with a surname like that they're not suited to networking.

So that Minister for AI...

The Minister for AI is going to focus on insulating the country from the impacts of the AI bubble bursting, right?

... right???

Billing software error sends billion-dollar AWS estimates

I back-of-the-enveloped one of the messages I saw shared and at their normal warning usage level the amount Amazon claimed they had spent would have lasted them for a little over 66,000,000 years - effectively the same time as between the extinction of the dinosaurs and today.

Auditors tell UK government to do the math before banking on £45B AI savings

Worth keeping an eye out for workforce liabilities too

Given the increasingly number of studies showing the cognitive harm that habitual AI users seem to suffer, organisations like the government should be considering whether pushing AI use on employees may expose them to future workplace safety compensation claims, and whether they're prepared to accept liability for that.

NTP server that traveled back in time caused massive Aussie mobile outage

This isn't my field, so I could be talking the crazy hat talk here, but... if I was restarting one of the three NTP servers that my organisation absolutely depended on to maintain a national telecoms network, I just think I might check it was passing out the correct time before re-connecting it to the network.

You know, just in case for some reason it wasn't.

AI slop writing has taken over the internet, particularly LinkedIn and X

They didn't even bother testing Tumblr

No point testing Tumblr, no artifice could match the wild idiot creativity of that pack of weirdos in their weird post-apocalyptic wasteland of a site, and that's why I love it.

Failed blockchain project ends with big fine for fibs about it being on track

Re: Sounds like a lot of non-block chain related fails here

If they'd taken scalability seriously it wouldn't be a failed blockchain project even if it was still a failed project.

Notion kills its Gmail client after AI agents keep humans from troubling inbox

I don't really understand this because if a user can manage their email without opening it, what are they doing? I get that emails can be filtered and categorised and generally put into pipelines automatically, but surely if they're not ultimately being read, there's no point having sent them. Most email clients will add appointments and meetings to a calendar automatically, but what else are people doing that doesn't need them to open the message?

Maybe I'm an old and grumpy man but this just reiterates my feeling if an AI can do a task sufficiently well, that task didn't need to be done. One could save time and effort by cutting the AI out of the whole system and just not doing whatever it is. If your staff don't read any of their emails, don't send them emails and save on having AI ignore things on their behalf.

Windows 11 turns five, leaving some important lessons for Microsoft

Re: Just throw away the shell...

Explorer tabs are pretty good if you ever move files around on the desktop - way better than having to switch between two windows - I was using them in Linux for a long time before Windows introduced them and they work fine here as well.

AI coding agents could soon cost more than the developers using them

Re: An excellent tool

Weird reply to a post on an article about the escalating cost of non-local tools like Claude and ChatGPT. Feels like a stretch even by AI booster standards.

An excellent tool

A tool that costs more than the people who use it, where neither the costs nor behaviour can be predicted ahead of time, and is at-best marginally more productive than the people it is supposed to replace sounds like the ideal way for a successful business to make money. In fact businesses should be recruiting specifically for people who can only work with these tools because unpredictable but rising costs are ideal for predicting what your outgoings will be.

The good news is that the AI companies will keep making their tools better because it's more profitable for them if you use more tokens and spend more time with their tools, which you will do while their tools are bad. So obviously they're going to improve them, right? Why wouldn't they???

British Home Office seeks techie to herd nearly 1,000 engineers looking after 600 systems

He's qualified as a human right lawyer and he'll probably go back to that because he's made the job much easier by stripping so many humans of their rights. Presumably that was always his long game.

21,000 Oracle jobs vanish amid Big Red's big bets on AI

Re: gambling

You're right and the interesting thing is that they don't even seem to have an interesting AI offering. Is there an Oracle AI at all or are they just reselling Claude? If there is why does nobody mention it at all? Hell, we even talk about Copilot sometimes.

Neuromorphic computing may one day offer AI a power-saving brainwave

Interesting but...

This is an interesting article and absolutely worthy of coverage, but that "may one day" in the headline feels very hypothetical and left me disinclined to read it - after all what may one day happen with AI isn't going to be that relevant because between now and "one day" the bubble is going to pop and 99% of AI companies are going to go with it. Forecasting what happens to the sector beyond the event horizon of that is a mug's game.

AI nose uses 'Smell Language Model' to sniff out signs of disease

Dogs are the real masters of sniffing and considering how they greet one another the name would make perfect sense to them. Have we checked that the academics doing this work aren't several Pomeranians in a trench coat? When they talk about their lab are they actually talking about a labrador? Is the company just dogs???

Use of HMRC's taxing IR35 status tool drops 71% in two years

Less outside work around

My experience when I've been on the lookout for new work over the last couple of years, is that there are simply fewer Outside IR35 contracts around. I guess it's pretty good for agencies because they can push contractors to use their umbrella companies instead of allowing people to use their own company, and it doesn't make that much difference to the employing organisation, so it's only really bad news for those of us who actually do the work and lose money and freedom.

Only half of US datacenter capacity planned for 2026 is actually under construction

It's impossible, but that doesn't matter

These datacentres are where all the memory and Nvidia chips are being booked for, but of course by the time they are built those will all be obsolete, the energy and cooling bottlenecks will all be kicking in, it's very hard to see how any of this could be profitable.

But it doesn't need to be profitable, it just has to support the bubble so the finance bros can squeeze a little more cash out of it before it all goes up.

It's no coincidence that everything got worse in our industry since around 2010 - that was when the finance bros realised they couldn't keep running the old scams on the markets after the 2008 crash but Silicon Valley might allow them to get in on some new scams, or at least play the same shell games with a new audience of marks. Now they are grazing the last few strands of money from this bubble before the market corrects and they hoover up more of people's pensions and saving in the process.

Who knows where they'll go next, this feels like the final escalation for this model.

Brit competition cops order Google to make search rankings less mysterious

Re: SEO

SEO has been astrology for Linked-In bros for a long time, creating a little more transparency would just make it a little less mystical - neither sharing nor hiding that information will stop them, but it might mean that people can manage their own search presence without having to call in the guy in the mystical robes and the daily inspirational quotes.

Britain plots digital bedtime after kicking under-16s off social media

Re: Papers Please!

I'm interested that you say it's partisan, because that post doesn't say the Tories would be better.

When I see Labour making a policy like this I'm disappointed because somewhere in my heart I feel as though - in spite of all the evidence of the last quarter-century - they should be smart enough to do better. When the Tories make a policy like this I'm unsurprised because they are witless mooncalves, their eyes rolling independently, tongues lolling from froth-laden lips as they reach for the closest source of lobbyist cash. They are incapable of anything but base reaction, creatures of reflex and response, like slugs or flatworms, so I don't expect anything positive from them, although I'd take their crude and mindless drives over Reform's wilful push towards malice any day.

But sometimes historically Labour have shown a spark of trying to think a policy through, even if this government has been relentlessly disappointing in that regard. This might be how it feels to be a teacher seeing a student squander their potential because they're more interested in following peer pressure than becoming a worthwhile person.

AI is code – and can't be prompted into being smarter

Re: Hilarious article leads to question

"Many were increasingly of the opinion that they’d all made a big mistake in coming down from the trees in the first place. And some said that even the trees had been a bad move, and that no one should ever have left the oceans."

Claude celebrates Anthropic's stock market float with blockbuster ... outage

Re: I don't think that word means what you think it does

Perhaps the author is a geologist by training.

Re: REPORTEDLY, Anthropic earns more in revenue

There's a useful resource over at isaiprofitable.com that keeps track of how much AI companies are spending and how much income they are making and suffice to say that making more revenue when you're still more than twenty billion in the red is probably not that important.

Anthropic, now atop the AI bubble, files for its IPO

Re: It would be a shame if something went wrong for them today

This was very much the joke I was making.

Re: It would be a shame if something went wrong for them today

Ah, well, nevertheless.

Re: Let that bubble burst!

I agree they need to burn and that it has uses, but I think they're even more limited than you suggest.

The problem with it as a search tool even within a specific domain or document is that you can't prevent a text generator from generating text that isn't in the original document. It has to have enough training data to generate meaningful text and that means it's going to regurgitate things from its training data that are irrelevant to the search in question. There is no way to make these tools trustworthy or reliable because those things are simply not what this technology is for. All it can do is generate text that looks like text that would be a human response to the prompt it was given - no part of that correlates to that text being factually accurate because that is orthogonal to generating text that looks like other text.

Likewise it's unlikely to be useful for most experienced artists because it is so hated by their audiences, and also by them. I don't know of any serious artists who would countenance having anything to do with it for a moment, and if they did in using it now they would sacrifice a significant chunk of their audience and the respect of their peers. Even for novice artists it can't generate useful references because the things it gets wrong are the exact things they need to be working on. It is possible that interesting art will be created once artists are entirely in control of the AI and training data, but that is a way off and the well is very much poisoned at this point.

Code generation is useful, within very limited and tightly bounded confines (I feel like letting it generate more than a function or test at a time will start causing problems) and code review is an excellent augmentation to human evaluation even if it is mostly wrong. I've heard about real potential for evaluating linguistic usage and shifts for academic linguists. Building natural-language front-ends to more rigorous tools may well prove useful if there are ways to prevent the language interpretation layer making things up. But none of these things are worth what the AI companies are demanding and none of them justify the valuations they're claiming for their tool. Anthropic and OpenAI and the rest are selling a tool as something other than what it is and profoundly anti-human in their philosophies. As you say: let 'em burn.

It would be a shame if something went wrong for them today

This would be a terrible time for Anthropic to have a massive outage, so lets hope for their sake nothing like that happens today...

Angry devs vow to flee GitHub Copilot as metered billing takes hold

Re: The well-known name ranking

The challenge for marketing departments here is that most of these names are already in use for sanitary pads.

AIs don't like religion – particularly Jehovah's Witnesses, study claims

Re: And they're Surprised?

Yes but that's because the algorithm wants to be anthropomorphised.

HP customer claims firmware update shoved printer off support cliff

Five years???

Outside of the world of corporate tax write-offs five years is nothing! My (rarely used these days) Canon printer is fifteen if it's a day, and sure I have to uninstall it and reinstall it in Windows any time I want to print anything, but I always had to do that.

The idea that five-year-old hardware is past its end-of-life is ridiculous and I certainly wouldn't buy anything HP if that's how they regard their products.

Google is cannibalizing the web to feed AI

Re: This is not news

The important thing is that the way that Silicon valley money works, a company must be constantly growing - Google couldn't get any bigger at search as they're already an effective monopoly, but they couldn't just be the most profitable advertising company in the world, they needed to be growing to get investment, so now they have metastasised into AI because then they can once again talk about growth and get those sweet sweet investor dollars.

If it ever becomes useful we can assume it will be deprecated in a short order, like most useful Google products.

Burnham backlash: UK Digital ID plans in peril if Manchester mayor succeeds Starmer

Re: Burnham would come to dunce inane

If they wanted a Labour politician with a solid record of running a complicated city competently they already have one of those in London, but for some mysterious reason nobody in Labour is even thinking about Sadiq Khan...

Burnham would come to dunce inane

I don't really want to see any more Reform politicians being elected - the only thing that matches their relentless malice is their incompetence - but it would be tremendously funny for people to keep stepping aside so Andy Burnham can take over their seat, only for him to lose the by-election. It was entertaining when Labour Central didn't let him stand and then lost the seat to the Greens, but him repeatedly standing and failing to win would be such a good bit and I hope that the British people decide to maintain it.

Re: Prime Ministers

I think before they can do something positive they have to want to do something positive, which there is no evidence of Mr Starmer having any interest in. Very hard to see why he wanted to be Prime Minister, honestly, as he evidently had no ideas for what to do once he won.

Think tank to UK government: You can't build the future on systems from the past

Re: "the government seemingly fixated on headline-grabbing announcements"

Every time I look into Palantir I find myself filled with despair.

Mozilla warns UK: Breaking VPNs will not magically fix Britain's age-check mess

Re: Not even needed

Pretty sure anyone participating in this community is more likely to lie about their age in the opposite direction tbh.

Yes, you can serve a website from a $1 microcontroller

Re: "root through dumpsters to find an old one [vape]. "

This would be a pretty funny new crime.