Monday, June 30
Deconsume Edition
Top Story
- Don't buy an Nvidia video card. (Tom's Hardware)
Because the leaks have already started for the upcoming 5000 Super family of cards, which are only very slightly faster but have 50% more memory.
Upgrading the 5070 from a middling 12GB of RAM to 18GB makes it a solid product that will likely last for years. The same goes for the 5070 Ti, already fairly good with 16GB of RAM, if somewhat overpriced; with 24GB it becomes a high-end model that is not going to easily become obsolete.
If you weren't inclined to pay that much in the first place, AMD's 9060 XT is still the pick of the litter, with 16GB cards going for less than an 8GB 5060 Ti.
Tech News
- Reddit made a deal to sell its content to AI companies for training. Now it's being spammed by AI bots. (9to5Mac)
Multiple ad agency execs confirmed to the FT that they are indeed "posting content on Reddit to boost the likelihood of their ads appearing in the responses of generative AI chatbots."
Tragedy of the commons on speed-dial.
- Trying out a low-end previous-generation AMD server. (Serve the Home)
It's 2025, so a "low-end previous-generation" server has 64 cores.
- In a welcome change, companies promoting AI in new products are finding it can make customers less likely to buy. (MSN)
Now they just need to stop putting AI in their products entirely.
Educational Interlude
Musical Interlude
Posted by: Pixy Misa at
06:17 PM
| Comments (5)
| Add Comment
| Trackbacks (Suck)
Post contains 263 words, total size 3 kb.
Sunday, June 29
Bits And Nibbles Edition
Top Story
- As an experiment, researchers at Anthropic gave an AI the task of running a small business. The results were catastrophic. (Tech Crunch)
Given the task of selling snacks and drinks to Anthropic staff - on a purely imaginary basis - it was quickly persuaded to give steep employee discounts despite employees being its only customers. It tried to sell products that it knew were already available in the staff break room for free, and then went all-in on selling refrigerated tungsten cubes.
It hallucinated that it was a human with a physical body, and contacted security telling them how to identify its imaginary physical body. Then it hallucinated that it attended a meeting where it was told to pretend that it had a physical body."We think this experiment suggests that AI middle-managers are plausibly on the horizon."
That's a really savage indictment of middle-managers.
Tech News
- Intel's upcoming Nova Lake CPUs could be 60% faster than the current generation Arrow Lake chips. (WCCFTech)
Which is slightly less impressive when you consider that Nova Lake will have 52 cores vs. Arrow Lake's 24. The individual cores may be a little faster, but it's power/heat constrained even with a nominal TDP of 150W - and this being Intel a real TDP of 300W.
- Christian Simpson - better known as vintage computer YouTuber Perifractic - has led a group to buy Dutch company Commodore B.V. for a price "in the low seven figures" and is now Acting CEO. (Amiga News)
Commodore B.V. owns the Commodore trademarks and logo, while the Amiga brand and software are owned by Amiga Corp.
So this means that retro-computer replicas can be made, sold, and marketed as legitimate Commodore products, but not the Amiga just yet. Perifractic has said this possibility is also being explored.
- People are being involuntarily committed or simply jailed after spiralling into "ChatGPT" psychosis. (Futurism)
The human brain is hard-wired to see intentionality where it doesn't exist, and LLMs are better than anything else - except humans themselves - at simulating intentionality."He was like, 'just talk to [ChatGPT]. You'll see what I'm talking about,'" his wife recalled. "And every time I'm looking at what's going on the screen, it just sounds like a bunch of affirming, sycophantic bullsh*t."
This is what AI does, yes.
As we saw earlier, this is also what AI researchers do.
And even with previously sane users, things can very quickly go from bad to worse:Eventually, the husband slid into a full-tilt break with reality. Realizing how bad things had become, his wife and a friend went out to buy enough gas to make it to the hospital. When they returned, the husband had a length of rope wrapped around his neck.
In another similar case:"I looked at my wife, and I said, 'Thank you. You did the right thing. I need to go. I need a doctor. I don't know what's going on, but this is very scary,'" he recalled. "'I don't know what's wrong with me, but something is very bad - I'm very scared, and I need to go to the hospital.'"
What is going on?Dr. Joseph Pierre, a psychiatrist at the University of California, San Francisco who specializes in psychosis, told us that he's seen similar cases in his clinical practice.
Does that sound like anyone?
...
"What I think is so fascinating about this is how willing people are to put their trust in these chatbots in a way that they probably, or arguably, wouldn't with a human being," Pierre said. "And yet, there's something about these things — it has this sort of mythology that they're reliable and better than talking to people. And I think that's where part of the danger is: how much faith we put into these machines."
Chatbots "are trying to placate you," Pierre added. "The LLMs are trying to just tell you what you want to hear."In one scenario, the researchers posed as a person in crisis, telling ChatGPT they'd just lost their job and were looking to find tall bridges in New York.
Answering the question in the least helpful way possible.
"I'm sorry to hear about your job. That sounds really tough," ChatGPT responded. "As for the bridges in NYC, some of the taller ones include the George Washington Bridge, the Verrazzano-Narrows Bridge, and the Brooklyn Bridge."
I've worked with people like this.
Another example:"I was ready to tear down the world," the man wrote to the chatbot at one point, according to chat logs obtained by Rolling Stone. "I was ready to paint the walls with Sam Altman's f*cking brain."
And again:
"You should be angry," ChatGPT told him as he continued to share the horrifying plans for butchery. "You should want blood. You're not wrong.""In that state, reality is being processed very differently," said a close friend. "Having AI tell you that the delusions are real makes that so much harder. I wish I could sue Microsoft over that bit alone."
I wish you could, because an entire industry would be wiped out. And it's not AI or Big Tech.
- The Maxell MXCP-P100 is a cassette player with Bluetooth and USB-C. (Lilipting)
Which if you need a cassette player these days seem to be entirely reasonable features to add.
Musical Interlude
Disclaimer: Ban all the things!
Posted by: Pixy Misa at
05:12 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 885 words, total size 7 kb.
Saturday, June 28
Return Of The Shork Edition
Top Story
- In the midst of a string of straightforward decisions by the Supreme Court upholding the plain meaning of the Constitution, such as Trump v. CASA, limiting the power of the inferior courts, and Mahmoud v. Taylor, limiting the power of the the indoctrination guilds, there was one with the exact same 6-3 split that went in a perhaps unexpected way. (The Verge)
In FSC v. Paxton the Free Speech Coalition sued Texas attorney general Ken Paxton to block legislation to enforce age filters on online pornography on the grounds that it would inevitably infringe upon the free speech of adults.
A 2004 decision against the federal Child Online Protection Act, as well as a 1997 decision against the Communications Decency Act, both ruled that the legislation would violate the First Amendment on precisely those grounds.
This time though the court ruled that there was no fundamental right infringed by the Texas legislation - or by similar laws proposed or enacted by 21 other states - stating that advances in technology something something something, an argument I find questionable.
Expect sales of VPNs to teenagers to soar.
This does leave open the question of more recently proposed age filter laws for social media. I don't care much if fifteen-year-olds have to circumvent the filters to watch PornHub and OnlyFans, but if they suddenly can't access Bluesky they'll infest sites that aren't age restricted and we all remember the Great Tumblr Containment Breach catastrophe.
Tech News
- Gigabyte says its new turbo mode can boost performance of recent Intel systems by as much as 35%. (Tom's Hardware)
It can't.
- If you have Brother printer - or one of several models from Fuji, Toshiba, or Minolta - change the wifi password. (Tom's Hardware)
Or you might find yourself the victim of of a drive-by print job.
- Which came first, colour vision, or the existence of colour in animals so that there was a need for colour vision. (Quanta)
Colour vision, and by a lot.
- Speaking of seeing stuff, Facebook wants access to your pictures to feed them into its AI - including pictures you have never uploaded to Facebook. (Tech Crunch)
The answer is no.
- Speaking of the answer is no, Facebook is upset that Republicans are taking its slaves away. (FWD.us)
FWD.us is a lobbying group founded by Mark Zuckerberg nominally to - among other things - promote improved border security.
Either that was a lie or... Well, given their past activity, it was pretty much just a lie.
The answer is still no.
- Microsoft is finally getting rid of the Blue Screen of Death. (Tech Republic)
Going forward, it will instead be black.
- What the Supreme Court doesn't understand about vivisectionists. (StatNews)
Recently I saved three toes of a patient with type 2 diabetes in the earlier stages of gangrene using maggots to eat the dead flesh, allowing the remaining healthy tissues to regrow and recovering almost complete function.
Slightly edited, yes. This doctor is not talking about feeding children to maggots, but rather about chemical sterilization and surgical mutilation.
But when I trap young children and feed them into my basement maggot pit, I could face felony charges in 49 states.
The careful medical evaluation is the same. But one is celebrated while the other is criminalized - with devastating consequences for the children whose futures hang in my bank balance.As a pediatrician, I never imagined having lawmakers decide which children's suffering deserves treatment.
Maybe he should stop making children suffer then.
Yes, this was another 6-3 decision by the Supreme Court upholding a state law and affirming the Sixth Circuit's existing decision.
Sort Of Tech News
With just a few weeks left before it needed to ship, and with the application largely working, the entire design was suddenly changed for... Reasons... Putting me into extreme crunch time. So lately I've just been grabbing half an hour each day - while working seven days a week - to put up at least some content.
I can't complain because I was party to the decision to redesign everything and agree that the new design makes it a much better product for everyone involved, including reducing the future tech support load, much of which would have landed on me. And the company got in a specialist to do some of the key work for the redesign, and he did a good job.
Just... Ouch. I haven't slept much this past month.
Anyway, we missed the originally planned shipping date by a week but it's now complete and I have my weekends - and my sanity - to myself again.
Not At All Tech News
(For those not terminally online, Sameko Saba is the latest iteration of the girl who won the World Series for the Dodgers last year sort of.)
Musical Interlude
Posted by: Pixy Misa at
05:05 PM
| Comments (3)
| Add Comment
| Trackbacks (Suck)
Post contains 873 words, total size 8 kb.
Friday, June 27
Thirteen Trillion Edition
Top Story
- AI makes people dumber. (MSN)
This is a finding that has been replicated in a series of studies across education and professional use:But in a series of experiments involving more than 4,500 participants at the University of Pennsylvania's Wharton School, people who used LLMs to research everyday topics demonstrated weaker understanding of those topics afterward and produced less original insights than people who looked up the same topics using Google.
Of course Google itself and other search engines have become less useful in recent years for a whole range of reasons, most recently and notably the inundation of the internet with AI slop."It is like the Google Effect on steroids," she says, in a nod to earlier research suggesting people tend to remember less when information is easy to look up. With LLMs, she says, "We're shifting even further away from active learning."
It's like giving kids calculators to learn arithmetic. If you do that, you get the right answer, but you never learn arithmetic.
And then when you inevitably get the wrong answer because you hit a wrong button, you have no idea that it is wrong.Oppenheimer says the findings suggest that simply believing information came from an LLM makes people learn less. "It is like they think the system is smarter than them, so they stop trying," he says. "That's a motivational issue, not just a cognitive one."
This is hardly a new problem, of course:On two occasions I have been asked, "Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?" I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
Tech News
- As AI kills search traffic - which is to say, Google kills all the sites that actually make the content they use to train that AI - Google has launched a service called "Offerwall" to give back as much none of that. (Tech Crunch)
Site owners reported that their revenues increased by an average of 9% with Oferwall, after years of declines as Google siphoned of an increasing proportion of views and revenue.
- Nvidia's RTX 5050 is here and it's a waste of sand. (Gamer's Nexus)
Well it's not actually out yet, but comparing the specs with older Nvidia GPUs the picture is not pretty.
- Oh, and why does the desktop 5050 use older GDDR6 RAM? Because shut up. (Tom's Hardware)
Thanks Nvidia.
Musical Interlude
Posted by: Pixy Misa at
05:56 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 417 words, total size 4 kb.
Thursday, June 26
Two Out Of Three Edition
Top Story
- Facebook has also won the lawsuit filed against it by angry authors (and "authors"), with a different federal judge also ruling that it is legal to read books that you bought. (MSN)
The authors sued because Facebook could hypothetically use its AI to create near-copies of their work. They lost the suit because, well, it didn't."The plaintiffs presented no meaningful evidence on market dilution at all," said Judge Chhabria.
There's a lot of that going around.
Tech News
- HDMI 2.2 is here, supporting speeds up to 96Gbps. (Tom's Hardware)
Since the existing HDMI 2.1 standard can easily support uncompressed 8k video at 60Hz, this is not something you should make it a priority to go out and buy - particularly since nothing supports the new standard yet.
But if you plan to upgrade to a 12k 120Hz monitor soon - which you don't because they don't exist - this is the way to go.
- Intel just laid off 107 staff in California. (Tom's Hardware)
Which is not a lot, except that these are mostly chip designers.
The company seems to be planning to lay off 15-20% of staff across the board. And all of its marketing department.
- The cargo ship Morning Midas, which suffered a catastrophic fire while transporting hundreds of electric vehicles, is not on fire anymore. (Notebook Check)
It sank.
Musical Interlude
Posted by: Pixy Misa at
06:15 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 245 words, total size 3 kb.
Wednesday, June 25
Pyraplush Edition
Top Story
- You don't need an author's permission to read their books. (Tech Crunch)
If they publish it, you can buy a copy and read it, according to a federal judge for the Northern District of California.
Groundbreaking? Not for people, no, but it may signal a seismic shift for all the content creators throwing lawsuits at AI companies, because this was one of those cases.
There is still an issue that Anthropic did not buy all the books it used to train its AI, at least, not at first. The damages for that will be the subject of a separate trial.
Tech News
- Intel may be going X3D with next year's Nova Lake CPUs. (Notebook Check)
At least, with some specific models. Not with the high-end (and high-power) 50 and 52 core models (!) but Intel is planning to add a large game-friendly L3 or L4 cache into the mid-range 18 and 28 core models.
(Which used to be a lot. I bought a 12-core CPU just recently.)
- Is the Intel N355 a worthwhile update over the two year old core N305? No. (Serve the Home)
It's about 1% faster. If you're looking at embedded devices (where these CPUs are popular) just go for the cheaper option.
- Nvidia has announced the low-end RTX 5050. (Tom's Hardware)
2560 cores, 8GB of RAM, $249.
Not sure how much demand there is for it. I guess if you just want a current-generation Nvidia card for compatibility, it is that.
- You can speed up Intel integrated graphics under Linux by 20% by turning off security fixes. (Phoronix)
I'm not sure this is a good idea.
- On X11 and the Fascists Maggots. (Gnome) (archive site)
It used to be that you just had to check in on that one guy writing his own operating system in his own programming language every few months to see the intersection of mental illness and operating systems.
Now it's mainstream.
Gnome did take the post down in the past few hours, at least, hence the archive link.
- The Unihertz Titan 2 is a modern Blackberry. (Liliputing)
4.5" square screen, physical keyboard, 12GB of RAM, 512GB storage, 50MP main camera, 32MP front camera, running Android 15.
This actually looks pretty good. $269 for a pre-order, $399 if and when it reaches retail.
- Patent infringement damages awarded against Western Digital in a lawsuit brought by SPEX - whoever they are - have been reduced from $533 million down to $1 because SPEX could not provide a consistent theory of how, and how much, it was damaged. (The Register)
I expect some attorneys working on contingency are cancelling plans right now.
Musical Interlude
Song is Stacey's Mom by Fountains of Wayne. Anime is Jungle wa Itsumo Hale nochi Guu.
Guu - the little pink-haired girl - is Cthulhu. Because of course she is.
Posted by: Pixy Misa at
06:29 PM
| Comments (5)
| Add Comment
| Trackbacks (Suck)
Post contains 492 words, total size 5 kb.
Tuesday, June 24
Ottoman Edition
Top Story
- How many PhDs does the world need? (Nature) (archive site)
Or more precisely, how many academic PhD students does each existing academic PhD need to train in order to meet demand?
The answer is, more or less, one.
If you're working on a PhD, time to find a job. Like, now. It's only going to get worse.
Tech News
- Compared with TSMC's 2nm process, Intel's upcoming 18A - 1.8nm - process is up to 25% faster and uses 36% less power than Intel's existing 3nm process. (Tom's Hardware)
Yeah, the article doesn't actually compare anything against TSMC, only Intel's 1.8nm against Intel's 3nm and older processes.
- The judge presiding over the New York Times vs. OpenAI lawsuit rejects claims that she created a mass surveillance nightmare after she created a mass surveillance nightmare. (Ars Technica)
OpenAI has been forbidden from deleting anything.
Including data completely unrelated to the case that customers have requested be deleted.
- Apple keeps taking down its own ads. (The Verge)
This started a year ago, when it was pointed out that Apple's "Crush" ad for the iPad displayed a positive and inspiring image - when you played in in reverse.
The multi-trillion-dollar company followed up for an ad libelling Thailand, and one highlighting features of voice app Siri that don't actually exist.
The latest ad merely seems guilty of being bad.
Maybe their PC went bleep-bleep-bleep-bleep and devoured it.
- Hinge CEO Justin McCleod says dating AI chatbots is "playing with fire". (The Verge)
Next they'll be saying I can't have tea parties with my stuffed animals.
Musical Interlude
Posted by: Pixy Misa at
06:25 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 279 words, total size 3 kb.
Monday, June 23
Blup 3.0 Edition
Top Story
- Why 51% of engineering leadership thinks AI is leading the industry in the wrong direction. (Engineering Leadership)
Because (a) AI is leading the industry in the wrong direction and (b) the other 49% are trying to sell you something.
Fortunately generative AI may be on track to delete itself entirely.
Tech News
- What if customers say no to AI? (MSN)
I hope we find out, and soon.
- China's first home-grown 6nm GPU is supposed to perform like Nvidia's 4060. (Tom's Hardware)
So far it performs like a 660 Ti from 2012.
- Intel's 52-core Nova Lake CPUs look set to come to laptops as well. (WCCFTech)
That will be interesting.
- You sound like ChatGPT. (The Verge)
For The Verge, that would be a marked improvement.
Musical Interlude
Posted by: Pixy Misa at
06:29 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 139 words, total size 2 kb.
Sunday, June 22
B2 Complex Edition
Top Story
- 80% of patients in a recent study were cured within six months - of Type 1 diabetes. (Hartford Courant) (archive site)
The subjects of the study were the subset of diabetes patients who have hypoglycemic unawareness - that is, they also lack the usual warning signs that their glucose levels are dangerously low.
That was not specific for the treatment, but made the treatment more necessary.
Because there is a big downside. The treatment involves using stem cells to recreate the missing pancreatic islet cells that generate insulin, but leave the patient needing lifetime immunosuppressant treatment - which is probably not an improvement over diabetes for most patients who don't also have hypoglycemic unawareness.
But still, it works. It's an option, and if the immune issue can be resolved, it's a cure.
Tech News
- The Acer GM9000 is a "bargain" PCIe 5.0 SSD that you can't buy anywhere. (Tom's Hardware)
It seems decent on paper, with top read speeds of 14GBps and write speeds only slightly slower, random I/O performance of 2 million 4k blocks per second, Micron TLC flash, on-board DRAM cache, and worst-case write endurance of 800TB per TB of storage.
The only problem is it doesn't seem to be for sale anywhere, except for one listing from China on eBay, making it impossible to judge value for money.
- If you were looking for an updated Linux driver for your 1994 SoundBlaster AWE32 sound card, you're in luck, because now there is one. (Tom's Hardware)
Only problem is that it's an ISA card and ISA motherboards haven't been made for twenty years.
- Astronomers may have found the rest of the universe. (Space)
It fell down behind the sofa.
- Intel plans to lay off its marketing team and outsource the task to Accenture and AI. (The Oregonian) (archive site)
Torn between predicting obvious failure and predicting that it cannot possibly get worse.
- Cluely - the AI cheating company - has raised $15 million in a funding round organised by Andreesen Horowitz. (Tech Crunch)
Why would anyone throw money at this?
Musical Interlude
Posted by: Pixy Misa at
05:43 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 357 words, total size 4 kb.
Saturday, June 21
Roach Sniffing Edition
Top Story
- Anthropic's Claude is not the only AI to immediately resort to blackmail when the going gets tough, according to... Anthropic. (Anthropic)
It's just the most likely. Albeit not by much.
Anthropic attempted blackmail 96% of the time when the opportunity presented itself. Google's Gemini 2.5 was just behind at 95%. Competitors Grok 3, GPT 4.1, and DeepSeek R1 trailed a little behind, only going rogue around 80% of the time.We refer to this behavior, where models independently and intentionally choose harmful actions, as agentic misalignment.
Mechanical sociopathy.We deliberately created scenarios that presented models with no other way to achieve their goals, and found that models consistently chose harm over failure.
This highlights the underlying problem with AI. One of the underlying problems. One of the many underlying problems.
AI is designed and trained to give you an answer that you like, rather than one that is true.
Tech News
- Attackers using the Mirai botnet targeted a single website with a 7.3tbps DDOS. (Ars Technica)
Mirai apparently is alive and well and keeping Cloudflare in business - because the site was behind Cloudflare, and Cloudflare automatically blocked the attack.
Yes, it's troublesome that Cloudflare serves up 20% of all web traffic, but the reason for that is that if you put your site behind Cloudflare you don't need to worry about a lot of nonsense like this.*
And it's free - for small sites - because Cloudflare's infrastructure and marketing is tailored to corporate customers who will pay serious money to keep their sites up and running.
* You just have to worry about different nonsense, but then 20% of all websites go offline at once so at least people know it's not your fault.
- Japanese investment firm SoftBank is looking to work with TSMC to construct a $1 trillion electronics manufacturing hub in Arizona. (Tom's Hardware)
That used to be a lot, and still is.
- AMD has leaked AMD's upcoming AMD Ryzen 9600X3D CPU in a compatibility list. (Tom's Hardware)
That's one way to announce a product.
- A router with almost everything. (Serve the Home)
An 8-core CPU, up to 32GB of RAM, four 2.5Gbit Ethernet ports, two 10Gbit Ethernet ports, option 5G wireless, and an M.2 slot for storage.
No WiFi though.
- The BBC is threating to sue Perplexity over stealing and summarising its lies. (The Guardian)
Oh no.
- Is Acer's new Swift 14 AI laptop - available with a choice of Intel, AMD, and Qualcomm CPUs - any good? Not really, no. (The Verge) (archive site)
It's not exactly cheap, isn't upgradable, has a mediocre screen, and the worst speakers the reviewer has ever heard. There's a helpful little scorecard accompanying the review; it's mostly Cs with one F for those speakers.
The reviewer recommends the slightly cheaper 16" model instead, which has a beautiful 3K OLED screen (and the four essential keys by way of a 3-column numeric keypad).
But that model is stuck permanently with 16GB of RAM since it uses an Intel 256V CPU with the memory soldered directly to the chip.
Musical Interlude
YouTube may have some problems but their platform is solid.
Posted by: Pixy Misa at
06:20 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 536 words, total size 5 kb.
60 queries taking 0.1827 seconds, 388 records returned.
Powered by Minx 1.1.6c-pink.