Posts by find users who cut cat tail

Wine 11 runs Windows apps in Linux and macOS better than ever

Re: Gaming is the Key

Never had a game that did not work in Linux with Steam. Granted, I do not play the type of games that want kernel-level anti-cheat protection. So YMMV.

Hong Kong’s newest anti-scam technology is over-the-counter banking

Re: Yep, the right step.

> I like the idea of face-to-face banking.

Especially since nowadays the alternative is a shouting match with a chatbot.

Apple blocks dev from all accounts after he tries to redeem bad gift card

Re: Local copies of, almost, everything here

> you might find your "average" user doesn't have a lot of paid-for account-locked content

Even if that were true, the Big Tech is working hard to change it and make virtually everything account-locked (if you do not know what you are doing).

British Airways fears a future where AI agents pick flights and brands get ghosted

Re: Worst case scenario

If this is your worst case, wait until LLMs start designing planes (from what I've heard in other industries, it is coming soon).

User insisted their screen was blank, until admitting it wasn't

Re: screen messages

The first character? That would be progress for my bank.

If you select a field in their form, but then decide to fill some other bit first and switch there, it starts complaining about missing values.

Welcome to America - now show us your last five years of social media posts

Re: Pulling out the old foot-gun?

If you do not have social media (like many of us here), you must surely be a TERRURRIST or something. If you had nothing to hide, you would spend your days blathering about yourself like any upright citizen, comrade. So best not even try…

UK finally vows to look at 35-year-old Computer Misuse Act

Re: Until....

Pretty much all ads are malicious code run by a third party (i.e. not even the website you are browsing) on your computer against your will. No prosecution there.

India demands smartphone makers install a government app on every handset

Re: And

People in several of those countries still think a plain piece of paper or plastic (i.e. an id) somewhere at the bottom of their backpack is ‘government spying on them’. It will be a rough awakening.

Stealthy browser extensions waited years before infecting 4.3M Chrome, Edge users with backdoors and spyware

Re: Drink me

So it should be un-derated.

Gorge on Microsoft Store apps with 16-at-once installer

Re: Sorry, this is news ?

A mass uninstaller would be more useful. But well…

Sony rolls out a standard way to measure bias in how AI describes what it 'sees'

Two words: Goodhart's law.

Microsoft warns of 'payroll pirate' crew looting US university salaries

While in Soviet Russia…

At a university on this side of the pond, you could phish as much as you like. Employees need to fill and sign a form and bring it to personal dept. to change anything related to the salary. In person. Antiquated? Maybe. But how often do you really need such changes? You probably change employers more frequently than that.

AI gets more 'meh' as you get to know it better, researchers discover

Re: Chatgpt was useful to me

Basically, if you use LLM ‘advice’ to fix surface-level problems, you set up yourself for being absolutely lost later – when tasked with solving the subtle and non-local problems the previous fix created. Or you may be lucky and doing something simple for which the LLM-regurgitated stuff is sufficient. It can happen. Usually, could just do a web search and copy'n'paste from some guide/tutorial in such cases, but whatever.

Re: PROPOSAL: Let's stop calling it “Artificial Intelligence”…

They already lost me at ExplanAItions…

AI chatbots that butter you up make you worse at conflict, study finds

We are doomed.

At the same time this is happening, the society tries pretty hard to shield idiots from the consequences of their idiotic acts. Sure, there will be reckoning, eventually. But it will be too late.

India's tech talent pipeline is sputtering

Re: Optimism until it pops

At least some impact on cognitive faculties seems to be long-term. They may be able to recover. We do not know since good long-term data need long-term studies and the entire LLM thing is still pretty new. But it is easier to acquire new skills at a younger age.

AI devs close to scraping bottom of data barrel

Re: Latest 'AI' excuse ... or Why 'AI' does not work !!!

Thank you for my weekly dose of punctuation.

AI: The ultimate slacker's dream come true

Re: Empowerment

Sure, if you want one of those degrees that were already pointless, go ahead.

Here you need to spend a lot of time in the lab to get the degree (and some useful work done).

Google's dev registration plan 'will end the F-Droid project'

Re: Trapped

Same here, not longer supported, blah, blah. Both lynx and links.

Text browsers worked for a while, demonstrating Google's hypocrisy – if Search worked in a text browser they could obviously send the same plain HTML to a GUI browser. That's probably why they were killed, not a technical reason.

Google-sponsored DORA report reframes AI as central to software development

Please stop this ‘rapid advance’ nonsense, in other words breaking everything constantly and completely unnecessarily (unless you consider being bad at your job a valid reason for breaking things).

Give me my FFTWs, SQLites and plainTeXs and FORTRAN77 code doing some useful maths and leave me be.

[Icon of old man yelling at cloud]

India’s IT minister moves to Zoho’s spreadsheet and word processor, urges 1.4 billion people to do likewise

Re: If it hurts Microsoft, it must be a good thing

You cannot escape AIfication so easily. Quoting Zoho website:

> Now with Zia across the board

> Zia is Zoho's AI-powered intelligent assistant trained to help you work better.

> In Writer, she's a writing assistant who helps you find grammatical errors, style inconsistencies and more, while in Sheet she's a data-analyst who helps you make data-driven decisions better. She's also your note-taking assistant who can search the web for you, draw references and add them to a note card on your command.

So, I am keeping my LibreOffice.

Word to the wise: Don't tell your IT manager they're not in Excel

Re: Excel

Even though everyone being like this would be probably a disaster, we NEED such people.

Because of course the spreadsheet formula does not include the last two cells, one reference is off by a column, it says *2 instead of +2 and no one has noticed MS Excel silently converted rows seven and eight to dates…

OpenAI says models are programmed to make stuff up instead of admitting ignorance

Re: However

Bad doctors? Hard to tell. My doctors certainly do say it occasionally – usually in the form ‘we would need these additional tests to know’.

AI in your toaster: Analyst predicts $1.5T global spend in 2025

Re: Just a thought

Forget the cloud. My question is:

Why?

What value are we getting?

Rust-style safety model for C++ 'rejected' as profiles take priority

Re: How do profiles enforce their rules

Ehm, mrum, ghrm, beer time now?

Boffins detail new method to make neural nets forget private and copyrighted info

Re: Ahhh. The need to publish is still as strong as ever. Add in some keywords (AI, LLM, copyright)

Admittedly, I have not read the technical details. But you do not need to decrease entropy to forget some data – you can do the exact opposite. It will make the LLM a tiny bit worse overall, but… who cares.

Larry Ellison bankrolling £118M AI vaccine research at Oxford University

There is a first for everything

I did not expect to ever agree, even remotely, with anything Larry Ellison is doing. But here we are.

Programmers: you have to watch your weight, too

Re: Binary advice ... part 2

I am sorry to inform you that Ø is slashed uppercase-Oh (used in Nordic languages). It has nothing to do with the digit zero, which has quite a different shape. Also, consequently, I now hate you. But you are in a good company as most people can't tell the differences between dash and hyphen…

AI spies questionable science journals, with some human help

Re: Sadly

Yes, my reaction to ‘identifying 1000 journals’ was (unfortunately): That's pretty easy, just make a list of those that e-mail me weekly.

I do not understand who publishes there. Even if you are desperate to publish something, papers in such ‘journals’ will not help you. It will make you an academic pariah as soon as people find out.

Vision AI models see optical illusions when none exist

> The three leading commercial models tested … all recognize actual visual illusions…

Do they?

Common optical illusion images are ‘standardised’. People had tried different things but then finally arrived at a few ‘canonical’ versions which work well to demonstrate the particular illusion. These are widespread with only minor variations – and plenty of labelled copies in the training data. So the classifier recognises these standard images (and similar).

In order to say the classifiers recognise optical illusions, they would have to test *novel* (or very obscure/atypical) optical illusions. Then maybe we could say they learned some fundamental properties of images which make them visual illusions. But the testing images look like canonical examples.

So, nothing to see here.

GitHub CEO: Future devs will not code, they will manage AI

Re: Can't wait to be a former developer

Passing unit tests is easy. The AI knows, approximately, what the report from a successful unit test should look like, and so it will generate one (it is how it does all things). Problem solved.

Meta training AI on social media posts? Only 7% in Europe think it's OK

Exactly. LLMs should be trained exclusively on social media posts.

Politically hot parts of US Constitution briefly deleted thanks to 'coding error'

Random chance

> all three missing bits related to one Trump controversy or another

What is the chance that if some code randomly deletes a few bits of the US constitution, they will *not* be related to Trump controversies? I would say pretty slim at this point…

Rampant emoji use suggests crypto-stealing NPM package was written by AI

But why?

Presumably the LLM was trained on lots of emoji-free code. So where or how did it pick up this silly habit?

Trump AI plan rips the brakes out of the car and gives Big Tech exactly what it wanted

> From now on, the US government will deal only with AI that pursues truth, fairness and strict impartiality.

These do not exist. At the minimum, it would require the training data and process to have such attributes.

So a favourable reading of such statement would be that they will not purchase any AI snakeoil. Of course, in reality it means they will only purchase from their cronies. For these idiots, truth = the content of Trump's last tweet and impartiality = spewing out our side's propaganda.

Copilot Vision on Windows 11 sends data to Microsoft servers

Let me get this straight…

They cannot make a usable OS. Like one where you can just find how to change screen resolution because it is in a logical place.

So they instead add an unstructured natural-language command line thingy (let's call it what it is). Which does not even behave deterministically. But on the plus side, it spies on you!

And people still complain Linux is complicated…

Please, FOSS world, we need something like ChromeOS

Re: If only...

This and thousand times this!

MS Windows is incomprehensible. No regular user actually understands how to do anything in MS Windows (and it applies equally to Apple users, if not more). At best they fumble about, achieve something randomly by googling and Brownian motion and then memorise the resulting completely stupid way of doing something as the only way of doing that thing…

The root cause is not that Linux (or any other OS) is complicated. The solution is not to make it simpler.

Exif marks the spot as fresh version of PNG image standard arrives

Glad to learn my native language is Old Tongue (from the Age of Legends?).

PNG is pronounced something like p(ə)n(ə)g(ə) around here. Difficult to describe if you are not used to words formed only by consonants.

More trouble for authors as Meta wins Llama drama AI scraping case

If tech bros slurp my GNU GPL licensed code for machine learning, I am absolutely for – provided that all LLM output then becomes automatically GNU GPL licensed.

LLMs can hoover up data from books, judge rules

Re: Farenheit 451

Someone should train an LLM on the US Declaration of Independence. You need the original, of course, to train it properly. No cheap substitutes.

Xlibre forks to the rescue – but Kubuntu gives X11 the boot

Re: Eddies around a sinking ship?

Wayland is certainly not delivering what I desire. So what are you talking about again?

Techie went home rather than fix mistake that caused a massive meltdown

Re: Honestly I'm bilingual

The feet for height and altitude I always find weird. You use imperial units because you want nice human-sized quantities, right? You have a plethora of units to choose from. So why has every mountain height and plane altitude be some ridiculously large number? And you can't even say 40 kft because that would be too SI…

Deal to 'save' UK colleges £45M in Oracle Java licensing fees followed audit requests

You need a separate PhD to understand Oracle licensing… And, as other pointed out, there are lots of DIY and BYOD going on, often simply because everything would fall apart if people were not doing it. In other words, a large grey area Oracle would love to audit and cause mayhem (whether they have any actual right to do so or not).

Microsoft broke DHCP for Windows Server last Patch Tuesday

Unfortunately, our advanced AI has not come up with any suggestions how to do that so far. But we keep asking…

KDE Plasma 6.4 ships with major usability and Wayland improvements

Re: Remind me

KDE has actually one of the better Wayland implementations (with a bunch of additions to make it reasonable).

But, yeah. The entire debate has always been between ‘we have this software and users love how it works across all platforms – except it breaks in Wayland because [SOMETHING] is impossible by design there’ and ‘we have this grand philosophy how GUIs should work in abstract and if you do not write your synchrotron control system to work with windows wrapped around a 3D bunny rabbit[*] you are a bad person’.

[*] See https://lists.freedesktop.org/archives/wayland-devel/2021-May/041844.html if you do not believe…

European consumers are mostly saying 'non' to trading in their old phones

What?

Why is ‘limiting supply of second hand phones’ fundamentally bad? When everyone uses their device until it is dead, completely dead, beyond any hope of upgrade or repair – that is the most eco-friendly thing to do:

(a) It means every device is utilised to its maximum possible lifetime, and

(b) the device is not shipped around the world needlessly to switch users.

The only missing piece is that you should not throw it to the nearest river but dispose of it responsibly. And Europeans are relatively good at this part too.

MiniMax M1 model claims Chinese LLM crown from DeepSeek – plus it's true open source

Re: You do know where Shanghai is, right?

It seems, unlike me, you did not have mandatory Marxism–Leninism classes at all levels of school. So… Not impressed. Next.

Firefox is dead to me – and I'm not the only one who is fed up

Re: OK, but what now?

I use Palemoon, and user agent string are not the main compatibility problem. Occasionally something does not work right or is too slow (including some of our silly internal systems – solved by using FF). Occasionally something complains but works acceptably nonetheless.

The biggest problem seems ‘feature canaries’. Websites check for some random feature X, which they may not use at all, and when the browser does not support it they conclude it does not support Y and Z they actually need, throw a tantrum and refuse to load at all.

But then, most of modern popular web is utter crap and I only touch it when necessary and with a burner browser profile. So what do I know…

Researchers claim spoof-proof random number generator breakthrough

Even though things like Bayesian interpretation of quantum mechanics or superdeterminism exist, quantum mechanical randomness behaves as ‘truly random’ as far as we know. We have ruled out all the naïve ideas that/how it is not actually truly random.

And the other interpretations are generally of the type that results of QM measurements still appear truly random to us, just the universe is even crueller than you thought.

The trendline doesn’t look good for hard disk drives

Re: There is a lot of "non-active" data out there

Exactly. You have SSDs for everyday fast access. But your longer-term high-capacity storage is unlikely to be SSDs. It does not have the capacity/price ratio. And trust.

We all have HDDs from the era before SSDs became common still happily spinning and storing data (even though the capacity is usually laughable now). There are practically no really old SSDs still working. The longevity might be improving but gaining trust in storage takes decades.