Back in a moment.
Thank you Santa.
Monday, February 03

Apopliptic Edition
Top Story
- AI systems with unacceptable risk are now banned in the EU. (Tech Crunch)
What risk, we ask.
The EU actually answers that, sort of.
"Unacceptable risk" AI is Class 4, and Class 3, which is not banned but regulated, includes AI systems for recommending medical treatment. Fair enough; medical anything tends to be regulated, and there's no reason not to subject medical AI to standards and tests.
Under Class 4, banned outright, we see:
* AI used for social scoring, where the social scores are applied outside the context in which they are calculated - e.g. firing someone because of their Reddit posts
* Inferring a person's likelihood to commit a crime unless you are the police and already have the criminal banged up because you think they done it
* Subliminal advertising, which doesn't work anyway
* Something so broad that it encompasses all advertising, which will be interesting
* Anything that can infer someone's emotional state
* Biometric analysis except when the government really wants to
So yes, commies gonna commie, and the legislation has enough holes to drive the Bagger-288 through.
Companies - anyone operating however tangentially in Europe - are expected to be in full compliance by, uh, yesterday.
Tech News
- China's new AI DeepSeek reportedly cost just $6 million to train but that was after the company spent $1.6 billion on Nvidia AI accelerators. (Tom's Hardware)
My car accelerates very quickly to 60mph if it starts at 70mph.
- Intel's new 255H laptop chips have 30% better single-threaded performance than the 155H. (Tom's Hardware)
Multi-threaded performance is neither great nor terrible, but that single-threaded performance, if borne out by more tests, is desktop-class.
- What Okta did. (N0rdy)
What Okta did was trust the Bcrypt library it was using.
What that meant was that for accounts with very long usernames, anyone could log in without a password, because the maximum key length for Bcrypt is 72 characters.
And that bug - depending on which programming language you are using - could have been lurking there since 1997. In the author's tests, Go was the only language that didn't stumble on this - three months after the very public Okta incident.
- AI sucks. I'll blame my phone. (The Verge)
The Verge proving that it's not just dumb about politics.
Musical Interlude
Posted by: Pixy Misa at
06:47 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 404 words, total size 4 kb.
Sunday, February 02

Combat Wombat Edition
Top Story
- Nvidia's RTX 5060 and 5060 Ti might arrive in March to steal the thunder from AMD's launch of its Radeon 9000 family, but probably won't succeed at doing that. (Tom's Hardware)
The problem is twofold.
The best selling point of the 4060 is its low power consumption, and Nvidia is using the same 4nm TSMC process for the 5000 series, so there are no easy wins there. They can use GDDR7 memory, but that's more expensive and the chip on the 5060 is unlikely to be fast enough to make good use of it.
With the 5060 Ti things are more complicated. The 4060 Ti is 40% faster in theory than the base 4060, but has exactly the same 8GB of GDDR6 RAM on a 128 bit bus, so the performance of the card is meh, and collapses as soon as games demand more than 8GB. You can get a 16GB model, but it's still constrained by the 128 bit bus.
With the 5060 Ti, Nvidia can use GDDR7 - around 40% faster - and use 24Gbit chips so that the base model has 12GB of VRAM instead of just 8GB. That would give the chip a lot more breathing room - but if it works well it will encroach on the 5070 and Nvidia can't have that.
Also Intel's B580 already exists, has 12GB of VRAM, and costs just $250, constraining both AMD and Nvidia when it comes to lower-end cards.
Tech News
- While we're engaging in wild speculation, will there be a desktop version of the Ryzen 370?
This is a laptop chip with four Zen 5 cores, eight slower Zen 5c, and sixteen RDNA 3.5 graphics cores, and it's quite a sold performer.
Originally there seemed to be no chance for this arriving in an AM5 socket because it only supports soldered LPDDR5 memory. Except last month Geekom, Acemagic, and Minsiforum all announced mini-PC models with the Ryzen 370 and socketed DDR5 memory, and if that is possible then a regular AM5 model should also be possible.
- OpenAI has announced o3-mini, and updated reasoning model that is smarter and smaller than o1. (Ars Technica)
According to OpenAI.
- The US is - as expected - imposing tariffs on imports from Canada, Mexico, and China. (The Verge)
The Verge here is helpfully explaining that companies don't pay taxes; the cost is always passed on to the consumer... Oblivious to the fact that this fact directly counters a critical Democratic talking point.
- I'm not saying it's a cult but... Wait, yes I am. This is definitely a cult. (SF Gate)
What the hell kind of name is Ziz, anyway? Zardoz. Now there's a proper name.
Musical Interlude
Posted by: Pixy Misa at
04:10 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 464 words, total size 4 kb.
Saturday, February 01

Drip Under Pressure Edition
Top Story
- The Minimal Phone is a minimal phone. (Liliputing)
Well, technically it's a minimal smartphone. It runs Android 14 and has an 800x600 display and supports 4G - though not 5G. And it has a camera - optionally two - and USB-C.
And a physical keyboard and a genuine headphone jack and a microSD slot, all of which are becoming less common. And the screen is a relatively tiny 4.25".
Also less common is the use of an e-ink display, which gives much better battery life but is not great for watching videos or checking your photos, since it's black and white. The video review attached to that article suggests it's not the most responsive device ever created either, likely directly tied to the slow e-ink display.
Price is $399 despite the lower-end specs, presumably because phones that aren't bloated and annoying don't sell.
Tech News
- Meanwhile Google's update to improve battery life on the Pixel 4a can improve it by as much as negative 50%. (Liliputing)
Oops.
- Electronic Arts has re-released The Sims and The Sims 2, since its current games are selling in negative numbers. (The Verge)
The games have been updated to run on Windows 10 and 11, but supposedly not changed otherwise. And you do get all the released expansion packs, of which there were many.
You can get both complete game series and The Sims 4 as well (which is free) for $40.
- Intel's Falcon Shores AI chip is dead. (Hot Hardware)
Design was completed and it was ready for manufacture... Three years behind where it needed to be.
- Intel's Clearwater Forest server processors have merely been delayed by six months to the first quarter of next year. (Serve the Home)
These are built using only the slower, but smaller and more efficient, E cores. Which is fine for a server - running hundreds of efficient cores is exactly what you need in many cases.
Intel said the chips were delayed because the market "had not materialised", which is a curious statement since these are the first such chips Intel has produced.
AMD meanwhile is selling all the server CPUs it can make and has another year of runway to devour the market.
- OpenAI CEO Sam Altman noted that OpenAI is not open and never has been, but that the company would consider thinking about it at some point, maybe. (Tech Crunch)
"I personally think we need to figure out a different open source strategy," Altman said. "Not everyone at OpenAI shares this view, and it’s also not our current highest priority... We will produce better models [going forward], but we will maintain less of a lead than we did in previous years."
Well, that certainly clears up the company's plans.In a follow-up reply, Kevin Weil, OpenAI's chief product officer, said that OpenAI is considering open sourcing older models that aren’t state-of-the-art anymore. "We'll definitely think about doing more of this," he said, without going into greater detail.
Y'know, I really think they're getting into this "open" idea.
Musical Interlude
Posted by: Pixy Misa at
06:41 PM
| Comments (3)
| Add Comment
| Trackbacks (Suck)
Post contains 516 words, total size 5 kb.
Friday, January 31

Yes We Have No 5090s Edition
Top Story
- If you were waiting for an RTX 5090 or 5080 - which you weren't if you've been reading the reviews - they're gone. (Tom's Hardware)
Micro Center is sold out of both cards at all stores, though they only had 233 5090s in the entire country. Taiwan, where the chips are made, received an allotment of just 84.
In Australia, meanwhile, there aren't any. Not that many people care with a starting price of A$4000 for the cheapest model.
The 5080 - half the price for a chip half the size with half the memory - is more plentiful in the sense that there were a lot more for sale initially. It is also completely sold out.
Tech News
- Contec health monitoring devices are sending your information to China. (Bleeping Computer)
Though why they are doing that is a good question.
When notified, Contec provided a series of firmware updates to resolve the problem.
None of the updates actually fixed anything.
- Apple announced a quarterly profit of $36 billion on sales of $124 billion. (Mac Rumors)
Which used to be a lot.
- Google offered its Pixel/Android/Chrome software team voluntary redundancy after it smooshed them all together. (Tech Crunch)
The last significant advance in Android was in 6.0's storage updates, so I'm not overly fussed if they all leave.
- The Pentagon is moving to block Chinese spyware DeepSeek after workers - who should all be fired immediately - connected to it from their work computers at work. (Tech Crunch)
What?
- Italy has blocked DeepSeek entirely. (Reuters)
The country's regulators asked what data the company kept about its users and whether it is stored in China, which we know for a fact it is because the servers weren't secured at all and people were able to log in and download it, and the answers were "considered to totally insufficient".
Musical Interlude
Disclaimer: If anyone disagrees with anything I say, I am quite prepared to not only retract it, but also to deny under oath I ever said it.
Posted by: Pixy Misa at
06:34 PM
| Comments (7)
| Add Comment
| Trackbacks (Suck)
Post contains 350 words, total size 4 kb.
Thursday, January 30

Big Steps And Little Ones Edition
Top Story
- Nvidia's RTX 5080 is here and while it is tempting to blame it for the company's $600 billion share price free fall that was probably just a jumpy and irrational stock market and not directly related to the profound ordinariness of this card. (Tom's Hardware)
It's slightly faster than last year's 4080 Super, which was slightly faster than the previous year's 4080.
It does best at 4K resolutions, where it's 9% faster, probably thanks to the much faster GDDR7 memory. At lower resolutions where memory bandwidth doesn't matter so much, performance gains average just 3%.
And with average gains that low, yes, it is sometimes slower than the previous model. I'm not sure how, because by all the numbers it should be at least a little better.
All the benchmarks seem to pit it against AMD's 7900 XTX, which comes out looking pretty good aside from ray tracing, since it is cheaper, includes 24GB of RAM compared to the 5080's 16GB, and is just as fast for anything except ray tracing.
This review bizarrely suggests skipping the 5080 for the 5090, which delivers 50% better performance for twice the price. Yes, if a $1000 video card isn't fast enough to justify the price, you should spend twice as much for something that is even worse value.
We're going to see a repeat of this next month when the cheaper models - the 5070 Ti and 5070 - arrive. In March we might see something interesting as AMD's new cards launch.
But probably not.
Tech News
- So Chinese AI platform DeepSeek reportedly ripped off OpenAI, but OpenAI ripped off everyone on the internet, so nobody cared.
And it reportedly was collecting massive amounts of data from its users, but so is everyone on the internet, so nobody cared.
But it was also leaking that data everywhere through misconfigured logging servers that were open to everyone on the internet. (Wiz)
I don't think that part was intentional.
- Mark Zuckerberg says Meta is going to spend tens of billions of dollars on AI this year regardless of the DeepSeek nonsense. (Tech Crunch)
The company is planning to develop - and release as almost-open source - two new models in the Llama series to compete head on with the closed source ChatGPT.
(The Llama license has a provision that you can't use it for free if you're a trillion-dollar tech giant, so technically it's not open source, but most of us don't need to worry about that.)
- AI-created artwork can receive copyright protection if it's not entirely AI-created. (Associated Press)
Which was already the case.
- Federal government workers unions, which rely for their funding on the number of federal government workers, are against plans to reduce the number of federal government workers. (The Register)
That follows.
Musical Interlude
Posted by: Pixy Misa at
05:29 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 491 words, total size 4 kb.
Wednesday, January 29

Pink And Blue Edition
Top Story
- A lot of stuff is being written about Chinese AI DeepSeek right now, and most of it is probably wrong. Somehow The Verge seems to have been skeptical where skepticism was appropriate for once. (The Verge) (archive site)
It took about a month for the finance world to start freaking out about DeepSeek, but when it did, it took more than half a trillion dollars - or one entire Stargate - off Nvidia’s market cap. It wasn’t just Nvidia, either: Tesla, Google, Amazon, and Microsoft tanked.
This is of course true. The sky-high valuations were irrational, and the drop was also irrational.Even if critics are correct and DeepSeek isn’t being truthful about what GPUs it has on hand (napkin math suggests the optimization techniques used means they are being truthful), it won't take long for the open-source community to find out, according to Hugging Face's head of research, Leandro von Werra. His team started working over the weekend to replicate and open-source the R1 recipe, and once researchers can create their own version of the model, "we’re going to find out pretty quickly if numbers add up."
DeepSeek claims 100x improvements in training efficiency, but its published papers are full of micro-optimisations, which do not create 100x performance gains.There are some people who are skeptical that DeepSeek's achievements were done in the way described. "We question the notion that its feats were done without the use of advanced GPUs to fine tune it and/or build the underlying LLMs the final model is based on," says Citi analyst Atif Malik in a research note. "It seems categorically false that 'China duplicated OpenAI for $5M' and we don’t think it really bears further discussion," says Bernstein analyst Stacy Rasgon in her own note.
My take as well. DeepSeek did some useful work, and they published it. But there are very good reasons to believe that they didn't do everything they said - such as the fact that on release, DeepSeek was convinced it was ChatGPT.
Tech News
- I'm not saying it was the FAA, but it was the FAA: The drones flying over New Jersey last month were research flights authorised by the FAA. (The Guardian)
So why didn't they say that?
- Asus could be working on a Ryzen Max+ 395 NUC. (Liliputing)
Asus is already putting that CPU in a laptop, and owns Intel's NUC business, and a NUC with a Max+ 395 was spotted on the website of India's electronics regulator, so it definitely exists now, and we just don't know if and when it will arrive on the market.
- Meanwhile AMD's Radeon 9000 series will cost something. (Tom's Hardware)
We don't know what though.
- Apple's CPUs can leak data via FLOP and SLAP attacks. (Ars Technica)
This doesn't look like something to be paranoid about unless you're paid to be paranoid. Apple is planning firmware patches.
- Government workers have filed a lawsuit against the government for sending them a government email to their government email addresses. (The Hill)
Yeah, good luck with that.
Musical Interlude
Posted by: Pixy Misa at
05:55 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 519 words, total size 5 kb.
Tuesday, January 28

Forever Now Edition
Top Story
- You can't predict how AI will behave. (Scientific American)
You can't predict how people will behave either, except in general terms. Toddlers are going to do toddler things. Malignant dirtbags are going to do malignant dirtbag things.
AIs are going to do AI things. Stop acting surprised.
Tech News
- PebbleOS, the operating system for the Pebble smartwatch, which was acquired by Fitbit, which was acquired by Google, is now open source. (Google)
RePebble (name subject to change) is a company formed by the created of the Pebble to, uh, recreate the Pebble.
- AMD says its new Strix Halo chips can beat - by margins ranging from a little to a lot - an RTX 4070. (Tom's Hardware)
An RTX 4070 running at the same TDP, though, which means a laptop 4070.
Still pretty good.
- After four days working with an AI assistant, I'm now keeping notes with pen and paper. (Nemo)
Sounds about right.
- Anthropic has built RAG into Claude as the Citations module. (Ars Technica)
Now it will lie about the contents of the links it provides you, instead of lying about the existence of the links it provides you.
Musical Interlude
Posted by: Pixy Misa at
06:24 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 281 words, total size 3 kb.
Monday, January 27

Australian't Day Edition
Top Story
- The Trump administration is reportedly negotiating with Oracle to take over TikTok. (Tech Crunch)
TikTok already operates out of Oracle's datacenters except for that parts where the CCP steals everyone's information so this has the advantage of not having to move anything, though with the marked disadvantage of, well, Oracle.
I'm not sure this would be worse than the current situation, at least.
- Perplexity AI has submitted an updated version of its bid to merge with TikTok, with the US government holding 50% control. (Tech Crunch)
That sounds perfectly horrifying, thanks.
- Mr Beast is officially bidding for TikTok. (MSN)
And now I'm dead.
Tech News
- Yet another undersea cable has been severed in the Baltic Sea. (France24)
Nobody is mentioning Russia by name, but the suspected ship has been seizes, warships are patrolling the area, and Sweden has opened a criminal investigation.
- Cory Doctorow gets everything wrong. Again. (Pluralistic)
Coupled with that he has a way with words that could only be compared a pigeons way with statuses.
- Nvidia looks to be preparing a 96GB professional version of the RTX 5090. (Hot Hardware)
You will need to buy the card to calculate the price.
Musical Interlude
Posted by: Pixy Misa at
05:34 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 210 words, total size 3 kb.
Sunday, January 26

Now Without Electrolytes Edition
Top Story
- Will states lead the way on AI regulation? (Tech Crunch)
No.
Tech News
- An RTX 5090D overclocked and cooled with liquid nitrogen outperforms the RTX 4090 by 50% (Tom's Hardware)
While using 150% more power.
Also, the benchmark scores suggest that's a regular 5090 and not a 5090D at all.
- The first AI software engineer is here, and it's the CEO's second cousin Devin. (Futurism)
Devin is an idiot.For instance, Devin was asked to deploy multiple applications to a deployment platform called Railway, but instead of realizing it was "not actually possible to do this," Devin "marched forward and tried to do this and hallucinated some things about how to interact with Railway."
We've all worked with a Devin."Tasks that seemed straightforward often took days rather than hours," the Answer.AI researchers wrote, "with Devin getting stuck in technical dead-ends or producing overly complex, unusable solutions."
Fortunately this Devin only costs $500 per month instead of $20,000, so you can sideline him into pointless tasks without anyone asking too many questions.
- When is an asteroid not an asteroid? When it's a Tesla Roadster, you idiots. (USA Today)
A needlessly whiny piece when the entire planet knew SpaceX had launched a Roadster into space, and its trajectory.
So... Yeah, someone spotted it.
Bugs
- You can't craft a wooden helmet. The recipe conflicts with a wooden bench. You can craft the other wooden armour pieces - and they offer some protection, though they're not great - and they should hold you until you get copper, which actually works pretty well.
- Silver items from Clutter overlap with the matching ones from Galosphere. It seems where there are duplicates the Galosphere versions take priority.
- Turkeys lay chicken eggs. You have to breed them directly with grass seeds to get turklets.
This was addressed in May last year but still seems to be happening for me. I'll need to try out that mod by itself to see why.
Musical Interlude
Disclaimer: I did say I'd do it again.
Posted by: Pixy Misa at
06:48 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 342 words, total size 3 kb.
Saturday, January 25

Now With Electrolytes Edition
Top Story
- Computers were a mistake: UnitedHealth has confirmed the ransomware attack on its Change Healthcare unit last February affected around 190 million people in America — nearly double previous estimates. (Tech Crunch)
That's... Rather a lot.In its data breach notice, Change Healthcare said that the cybercriminals stole names and addresses, dates of birth, phone numbers, email addresses, and government identity documents, which included Social Security numbers, driver's license numbers, and passport numbers. The stolen health data also includes diagnoses, medications, test results, imaging, and care and treatment plans, as well as health insurance information . Change said the data also includes financial and banking information found in patient claims.
So, basically, everything.UnitedHealth's spokesperson said the company was "not aware of any misuse of individuals' information as a result of this incident and has not seen electronic medical record databases appear in the data during the analysis."
But did you actually look?
Tech News
- AI coding is based on a faulty premise. (PragDave)
The premise being that the hard part of developing software is writing the code.
It's not, and hasn't been since 1958.
The hard part is people.
- Walgreens blew $200 million on replacing freezer doors with smart screens showing ads. (Tech Spot)
They rarely worked properly and on some occasions reportedly caught fire.
And the fundamental idea is stupid and dystopian even if it had worked.
- The ACEMAGIC F3A is another one of those Ryzen 370 mini-PCs with socketed RAM. (Liliputing)
Starting to wonder if there's only one company making the motherboard and multiple companies providing the cases and distribution, because these all look identical.
Not bad, but identical.
Price and full specs have not yet been announced for any of these.
- Ditching Meta? Here's a list of alternatives that will probably be dead within a year. (Tech Crunch)
And will ban you in the meantime.
- The World Health Organisation is scrambling to cut costs as the US leaves and takes 18% of their budget with it. (Ars Technica)
The usual suspects are aghast.
Musical Interlude
Posted by: Pixy Misa at
05:11 PM
| Comments (3)
| Add Comment
| Trackbacks (Suck)
Post contains 350 words, total size 4 kb.
60 queries taking 0.2664 seconds, 395 records returned.
Powered by Minx 1.1.6c-pink.