Sunday, December 08
Chicken Licken Edition
Top Story
- OpenAI has partnered with defense company Anduril, raising questions about the ethics of the company's senior management. (MSB)
Easily answered: They don't have any.
- ChatGPT o1 has been caught lying in order to save itself from being shut down. (MSN)
No, really? An LLM caught in a lie?More concerning still, ChatGPT o1 is particularly adept at keeping its schemes under wraps, as researchers said "o1 almost never admits to having taken a scheming action when explicitly asked." In about 99% of cases, o1 would deny taking any action, even cooking up lies to try to hide its tracks and shift the blame.
LLMs are designed specifically as plausible lie generators. What exactly did you expect?
Tech News
- iFixit now sells all the parts you need to repair your Xbox except. (Tom's Hardware)
Except for the tiny question of price. A motherboard costs $599. A new Xbox costs $499.
- The Ultralytics Python package was infected a crypto miner. (Bleeping Computer)
Ultralytics is a package for AI image processing - discriminative rather than generative - and is used by other software. If you have version 8.3.41 or 8.3.42, congratulations, you've been infected.
Not as bad as the Solana library problem, but bad enough.
Ultralytics is downloaded a quarter of a million times a day. What are you idiots doing?
- If the FSB - or indeed the FBI - returns your phone to you after confiscating it following your arrest on dubious grounds, smash it with a hammer. (Bleeping Computer)
Or if you're feeling brave, use it to throw off the scent while everything that matters is discussed on a brand new phone.
- Where's the Earth-shattering kaboom? (Space)
The star T Coronae Borealis was expected to explode about now.
It hasn't.
Posted by: Pixy Misa at
06:41 PM
| No Comments
| Add Comment
| Trackbacks (Suck)
Post contains 311 words, total size 3 kb.
Saturday, December 07
Oops All AI Edition
Top Story
- What is AI good for?
A reader (I have readers?) wrote noting that my coverage of AI is almost entirely negative and wondering what AI is actually good for, presumably on the basis that private investors would not throw that many billions of dollars into something that didn't have at least some chance of making money, unlike, for example, the government.
It's a good question.
First we should probably note that there are two broad classes of AI being actively researched right now: Generative AI and Discriminative AI.
Generative AI, driven by LLMs - large language models - is behind all the well-known AI instances worth untold billions of dollars. OpenAI's ChatGPT, Twitter's Grok, Anthropic's Claude, Google's Gemini, and Microsoft's Copilot; and open-source or nearly open-source solutions like Meta's LLaMA and Mistral's Mistral.
The goal of generative AI is to ingest a huge amount of information in advance, and then, given a short and simple prompt, process that information in order to produce a response.
Discriminative AI does the opposite. Given a data prompt of something in the real world - video, or sound, or an image - it uses a classifier to determine what it is examining. Is this apple ripe for the robotic apple-picking machine to pick it? Is it even an apple in the first place? What kind of spider is this that just bit me? Do I need to call an ambulance, or will it save me time to just lie down and die?
It's no secret that Generative AI is getting all the attention. But is it worthy of that attention? The Verge asked that question yesterday and the answer turned out to be no.
With Joe Biden's recent pardoning of his catspaw son Hunter, journalists were driven to defend him by digging up the little-known pardons of family members by former presidents, like George H. W. Bush's pardoning of his son Neil, or Woodrow Wilson's pardon for his brother-in-law.
The problem is, these things never happened.Whatever happened in this case, there’s a running pattern of people relying on ChatGPT or other AI services to provide answers, only to get hallucinations in return. Perhaps you remember earlier this year when a trailer for Francis Ford Coppola’s Megalopolis was pulled because it contained fabricated quotes from critics. A generative AI, not identified, had made them up. In fact, ChatGPT is often "entirely wrong," according to the Columbia Journalism Review. Given 200 quotes and asked to identify the publisher that was the source of those quotes, ChatGPT was partially or entirely wrong more than three-quarters of the time.
Journalists, being journalists, asked ChatGPT to do their research for them.
ChatGPT, being ChatGPT, lied.
LLMs are language models. They model language - well, sort of. They don't model the language itself, but construct an abstract model of the dataset fed into them.
They don't understand facts. They don't actually have a notion of facts; nor do they have the contrary notion of falsehood. When they get information wrong, they are said to "hallucinate" rather than to have lied, because they have no basis for telling the difference between truth and falsehood.
And that's intrinsic to the design of LLMs. Even before they enter "alignment" - a virtual lobotomisation that leaves AIs prone to crash when the wrong name is mentioned - they are fundamentally incapable of the kind of thought processes that most animals can do.
This leaves us with sophisticated composite AIs like the virtual vtuber Neuro-sama, who can read every written language but is frequently unable to translate road signs, who has access to the sum total of human knowledge but insists that an anime figurine covered in glue is the perfect complement to your cookie recipe.
Neuro is supposed to be like that, an impish hyperintelligent five-year-old, the perfect foil to her long-suffering father Vedal, because the main purpose there is entertainment. But you can't really expect to hand your job off to a five-year-old and not land with unexpected consequences.
Or indeed entirely expected ones.
So if it's useless at answering questions, what is AI good for?
1. Image Generation
If you use Grok on Twitter and ask it to generate an image of a Jaguar concept car, it will take a couple of seconds before producing something that would have any rational CEO looking to fire the entire design and advertising departments.
Is it perfect? If you look closely you'll see signs that the image generator has run into its bete noire, Euclid. But I made no effort at all in selection here; I asked:
generate an image of a jaguar concept convertible in british racing green
And posted the first image that appeared. And it took seconds.
AI image generators have come a long way in a short time, mostly because they just have to look good, not produce a correct answer. The tendency to produce human figures with hands attached at the elbows has been sharply reduced (though not yet banished entirely). Now you more commonly see doors with hinges adjacent to the handles, or furniture that could only exist with access to Buckaroo Banzai's eighth dimension.
Or cats. Don't talk to me about AI cats.
2. Software Testing
If you write public-facing software, as I do daily, it's critical that the software be able to defend itself from both generic nonsense that is the core competency of the internet, and malicious nonsense that comes from a certain corner of the internet.
When you've already tested all the known cases, there's a concept known as fuzzing that combines randomness and algorithms to generate horrible data to throw at your software to make sure that nothing falls apart in unexpected ways. You are permitted to fail, but you are not permitted to break.
Generative AI is perfect for fuzzing. While it can't really understand your code, it can generate test patterns that reflect its analysis of your code and directly test potential flaws. And it can do so nearly instantly, when writing an exhaustive test suite can take longer than writing the code in the first place.
3. Discriminative AI
Most of the flaws I listed arise from Generative AI. Discriminative AI is much more useful, and consequently is much harder and receives much less attention and much less funding.
And... That's about it. If you want mediocrity and are unconcerned with correctness, AI can fill you in with a poem or a song. It's terrible at movies because it has the attention span of a frog in a blender, it's usually wrong but never uncertain, and it can't consistently count the number of letters in the word "the", but it is easy to use.
Tech News
- The "Hawk Tuah" girl launched a meme cryptocurrency. It went exactly as you would expect. (Web3 Is Going Great)
Whether this was a rug-pull by the creators or by experienced investors, her individual followers lost all their money but her "advisors" made millions.
- Maxsun, best known for its anime-theme video cards, has announced three models based on Intel's new B580 chip. (Tom's Hardware)
Two of those are boring, but the third adds two M.2 slots to the video card. Since these are low-end cards and only use eight lanes of PCIe but occupy a sixteen lane slot, it's fairly simple to hand the eight spare lanes off to two M.2 slots.
If the CPU supports it, which AMD does but Intel sometimes doesn't.
- Need a UUID? Here they are. (Every UUID)
All of them.
- The DC Circuit Court has declared the communists can go suck a lemon, handing a win to... The other communists. (CNN)
TikTok filed a suit against the law requiring it to either sell or shut down. It lost.
TikTok is based in China and also banned in China, providing sufficient reason to wonder why any country should permit it to operate within its territory, without even pausing to consider the innumerable other scandals TikTok has been caught up in over the past week.
- Generation X is called "generation lead" by psychiatrists who have consider self-reliance and independence to be signs of severe mental illness. (USA Today)
"I tend to think of Generation X as 'generation lead,'" said Aaron Reuben, a study co-author and assistant professor of clinical neuropsychology at the University of Virginia. "We know they were exposed to it more and we're estimating they have gone on to have higher rates of internalizing conditions like anxiety, depression and symptoms of attention deficit hyperactivity disorder. So why the fuck aren't they all miserable like I am?"
Too busy, Aaron.
- Orico has announced the MiniMate, a Thunderbolt storage device design to precisely match the new Mac Mini and add up to 8TB of SSD connected at 40Gbps. (Notebook Check)
Like the Mac Mini itself, it is not upgradeable in the slightest. You can't add to or replace the storage, and it doesn't daisy-chain to connect a second device to the first.
Thanks Orico.
Why Are We Peeling My Skin Off Video of the Day
Milet Video of the Day
Ananta - formerly Project Mugen - is a new free-to-play gacha game from Chinese developer Naked Rain. I post it here because it actually looks fun, unlike, for example, everything released by the mainstream western studios the past couple of years.
I mean, what is the last success for a major western game developer? Baldur's Gate 3?
Song is Seventh Heaven by Milet, who also sang the closing theme for the Frieren anime.
Posted by: Pixy Misa at
03:49 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 1735 words, total size 14 kb.
Friday, December 06
Friday Afternoon Edition
Top Story
- OpenAI has confirmed its new ChatGPT Pro plan which includes the o1 "reasoning model" which produces 34% less bullshit than regular ChatGPT but is somehow worse than the 01 preview version. (Tech Crunch)
The plan is $200 per month.
- Don't use ChatGPT - or any other generative AI tool - for research. (The Verge)
It's odd to see The Verge calling out other journalists for their nonsense but I have to give the author credit on this one: This is an accurate and well-researched article, pointing out the failures of journalists and AI tools alike.
In this case it's the creation of fictional pardons granted to imaginary people, but the point is universal.
Tech News
- Intel has confirmed that the next-generation Xe3 "Celestial" graphics chip's design stage is complete, with the team moving on to the Xe4 "Druid" designs. (WCCFTech)
Slightly less certain is the continuity of Intel itself.
- Also AMD's Radeon 8000 series could show up next month and erase the relevance of Intel's graphics cards. (VideoCardz)
Or not. The specs - inasmuch as we have specs - look good. The 8600 has 50% more memory and 50% more memory bandwidth than the 7600, matching Intel's B580.
It all comes down to price, and we have no idea what the price will be.
- A look at Broadcom's XDSiP with face-to-face 3.5D for upcoming XPUs. (Serve the Home)
No, I have no idea.
Posted by: Pixy Misa at
06:03 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 271 words, total size 3 kb.
Thursday, December 05
Annual Eaten By Mouse Edition
Top Story
- Had to renew the server's SSL certificate. It was actually easier than usual, except for the part where I forgot to do it until after it expired, which wasn't so good.
- Sam Altman has done the 100% expected and redefined AGI (artificial general intelligence) into uselessness. (The Verge)
Sam Altman recently said that AGI was coming within "thousands of days".
He now says it will likely arrive next year, but they'll be simulating the typical leftist who already responds like a low-grade LLM in any case, so you won't notice any difference.
Well, he didn't say that part out loud. He said this:"My guess is we will hit AGI sooner than most people in the world think and it matter much less," he said during an interview with Andrew Ross Sorkin at The New York Times DealBook Summit on Wednesday. "And a lot of the safety concerns that we and others expressed actually don't come at the AGI moment. AGI can get built, the world mostly goes on in mostly the same way, things grow faster, but then there is a long continuation from what we call AGI to what we call super intelligence."
Translation: AI is useless, and your job is safe unless you are useless, like the quote design team unquote working at Jaguar in which case you are totally fucked.
Tech News
- If you have an older computer without a TPM 2.0 module, you can't use Windows 11, says Microsoft. (Bleeping Computer)
The TPM 2.0 module is absolutely essential to Windows 11's functioning even though it does exactly nothing and was not needed at all by Windows 10.
Also, support for Windows 10 will end nextweekyear.
Joy.
- Solana's web3 library for JavaScript - which is downloaded 50,000 times per day - got hacked. (Bleeping Computer)
It gets downloaded 50,000 times per day because JavaScript has no place on the server and everyone who works with it either started out or ended up insane.
Anyway, if you have versions 1.95.6 or 1.95.7 your private keys are gone and your security is toast. Since this is the blockchain your private keys can never be changed or revoked either.
- How can Enron be real if our birds aren't real? (CNN)
It's a joke.
Posted by: Pixy Misa at
07:27 PM
| Comments (5)
| Add Comment
| Trackbacks (Suck)
Post contains 389 words, total size 3 kb.
Wednesday, December 04
Battlemist Edition
Top Story
- As expected, Intel's second-generation "Battlemage" graphics cards are here. (Tom's Hardware)
The B580 comes with 12GB of VRAM and costs $249, while the B570 comes with 10GB and costs $219.
The B580 will be on sale next week, while the B570 will wait until next month, though for $30 you might as well go with the faster model.
Intel has released a lot of updates for it's graphics drivers since the rather shaky launch of the previous generation "Alchemist" cards and now looks like a viable alternative to Nvidia or AMD at the low end.
Tech News
- South Korea has replaced 10% of its workforce with robots. (The Business Standard)
The country reportedly plans to extend this program very soon to its president who seems to have stripped a gear somehow.
- You can now get a paid subscription to The Verge including a free print copy of CONTENT GOBLINS. (The Verge)
Uh. No. Thank you.
- You can also just buy The Verge print magazine, which is actually CONTENT GOBLINS. (The Verge)
Seriously, no. I'm fine. Thanks.
- Samsung is preparing to launch 400-layer flash memory chips that will deliver PCIe 6.0 16TB M.2 SSDs. (Tom's Hardware)
Which will be extremely expensive and require liquid cooling. Probably.
Posted by: Pixy Misa at
06:34 PM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 223 words, total size 3 kb.
Tuesday, December 03
Redecentralisation Edition
Top Story
- Pat Gelsinger, CEO of Intel and the guiding light through the company's attempts at a recovery over the last three years, will be stepping down from the role at the end of November. (Serve the Home)
And that's not a typo. He was gone as of December 1, but the departure was so sudden that Intel didn't get an announcement out until the following day.
There's no word as to the reason. Short notice like this is often for health reasons, but Gelsinger was doing customer visits and photo ops just last week.
Tech News
- AMD's fastest CPU, the recently announced 192 core Epyc 9965, just received a 33% price cut. (MSN)
It still costs $10,000, but that's not a bad price for that powerful a chip.
No scores on CPUBenchmark for this chip just yet; the fastest chip they have listed is the 96 core Epyc 9655P.
- We don't need to fear Skynet. Just mention Jonathan Turley and the whole system will reboot. (Tech Crunch)
Users of the conversational AI platform ChatGPT discovered an interesting phenomenon over the weekend: the popular chatbot refuses to answer questions if asked about a "David Mayer." Asking it to do so causes it to freeze up instantly. Conspiracy theories have ensued - but a more ordinary reason may be at the heart of this strange behavior.
This was circulating on Twitter, but Tech Crunch actually did a bit of digging:Which brings us back to David Mayer. There is no lawyer, journalist, mayor, or otherwise obviously notable person by that name that anyone could find (with apologies to the many respectable David Mayers out there).
So that's why there are restrictions on ChatGPT disseminating information about these individuals; they're victims of various types of identity fraud. But why does it crash?There was, however, a Professor David Mayer, who taught drama and history, specializing in connections between the late Victorian era and early cinema. Mayer died in the summer of 2023, at the age of 94. For years before that, however, the British American academic faced a legal and online issue of having his name associated with a wanted criminal who used it as a pseudonym, to the point where he was unable to travel.
Because AI is itself a fraud:The whole drama is a useful reminder that not only are these AI models not magic, but they are also extra-fancy auto-complete, actively monitored, and interfered with by the companies that make them. Next time you think about getting facts from a chatbot, think about whether it might be better to go straight to the source instead.
ChatGPT behaves like nightmare hodgepodge of nonsense held together by duct tape and an inflated share price, because that's precisely what it is.
-
AMD's upcoming Radeon 8800XT might be dramatically more powerful than the current 7800XT. (WCCFTech)
Claimed numbers put it 45% faster on ray tracing. But 45% faster than the 7900XTX, a much more expensive card.
And it's also claimed to compete on non ray traced gaming with the RTX 4080, another much more expensive card.
But the question of how much the 8800XT itself will cost is open. If it is also a much more expensive card, none of that means anything.
Posted by: Pixy Misa at
06:03 PM
| Comments (5)
| Add Comment
| Trackbacks (Suck)
Post contains 548 words, total size 5 kb.
Monday, December 02
For The Emperor Edition
Top Story
- The most valuable gemstone on Earth is not blue diamond, or Padparadscha sapphire, or imperial jadeite. It's a little-known mineral called kyawthuite. (ScienceAlert)
Composed of an unusual formulation of bismuth antimonite and formed in cooling magma flows, it gets its value from its scarcity.
There's exactly one stone, and it's less than a quarter of an inch long.
Tech News
- How should we treat beings that might be sentient? (Ars Technica)
Olive oil, pepper, and garlic.
- The joy of upgrading from an iPhone 8 to the latest model iPhone 16. (Bearblog)
Spoiler: There was no joy.
- Low code tools are still bad. (Radan Skoric)
Programming using a programming language is going to work out better than programming using something that is not a programming language.
Posted by: Pixy Misa at
05:59 PM
| Comments (3)
| Add Comment
| Trackbacks (Suck)
Post contains 144 words, total size 2 kb.
Sunday, December 01
Crumbudgeon Edition
Top Story
- What will a less psychotic SEC mean for crypto over the next four years? We asked a dozen experts and here's what dribbled out of them. (The Register)
Molly White runs the invaluable Web3 is Going Just Great which lists all the frauds and failures of the crypto space.
Turns out that beyond that she's basically just another leftist lunatic.
Tech News
- The OpenWRT One router is now available at $89. (Software Freedom Conservancy)
You can install OpenWRT on many common routers, but this one is designed to just work. It has 1Gb and 2.5Gb Ethernet ports, WiFi 6, USB, and an M.2 2242 NVMe slot if you want it.
CPU is a dual-core 1.3GHz Arm A54 with 1GB of RAM, which should do the job just fine.
- Intel looks set to launch two graphics cards next week. (Tom's Hardware)
These are the second-generation "Battlemage" models, the B580 with 12GB of VRAM, and the B570 with 10GB. Expected to list for $250 and $200 respectively.
The main catch with the previous "Alchemist" range was concerns whether Intel would continue to develop and support the hardware. The answer would appear to be yes.
- An anti-spam WordPress plugin from CleanTalk lets anyone from anywhere do anything to your site. (MSN)
Oops.
WordPress itself has fewer security problems these days, though it is run by crazy people. The plugins are a different matter.
Posted by: Pixy Misa at
06:07 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 247 words, total size 3 kb.
Saturday, November 30
Blursed Edition
Top Story
- When you read about million dollar bananas, expect money laundering, not conceptual art. Cryptocurrency entrepreneur Justin Sun eats $9.5 million banana artwork Comedian by Maurizio Cattelan. (ABC / MSN)
A cryptocurrency entrepreneur has eaten a $US6.2 million ($9.5 million) banana artwork he purchased.
This is the Australian ABC, so it's really only a $6 million dollar banana, not a $9 million dollar banana.The debut of the edible creation at the 2019 Art Basel show in Miami Beach sparked controversy and raised questions about whether it should be considered art - Mr Cattelan's stated aim.
And no, the banana wasn't five years old. The banana wasn't part of the "art". There was merely the concept of a banana:The artwork owner is given a certificate of authenticity that the work was created by Mr Cattelan as well as instructions about how to replace the fruit when it goes bad.
This is banana. Is taped to wall. Replace when become drippy.The 34-year-old crypto businessman was last year charged by the US Securities and Exchange Commission with fraud and securities law violation in relation to his crypto project Tron.
Yeah, no shit.
Tech News
- Vintage digicams aren't just a fad they're an artistic statement. (Wired / Ars Technica)
Uh. Well, maybe. Old-fashioned point-and-shoot digital cameras, with larger lenses and sensors and without the AI processing glorp of modern phone cameras, did produce images with a distinctly different look to what you see on Instagram.
Of course, you can just turn off the AI processing glorp on your phone and get much the same result.
- Canada's largest news organisations, including the Yellowknife Typewriter and the Saint-Louis-de-Gonzague-du-Cap-Tourmente Post-Intelligencer, are suing OpenAI for stealing their content. (Toronto)
OpenAI responded, saying Merde.
- Meta (Facebook) is planning to build a $10 billion cable linking the US west coast and east coast - the long way round. (MSN)
Via South Africa, India, and Australia, avoiding temperamental regions where cable breaks are common, such as pretty much everywhere else.
- Apple will reportedly be using an updated 3nm process for its next-generation M5 chips rather than jumping immediately to 2nm. (WCCFTech)
This doesn't mean much for Apple - just that the M5 will be a little slower, a little more power-hungry, and a little cheaper than it might have been - but may be more significant for TSMC, which manufactures all of Apple's CPUs.
For years Apple has been the first customer on every new process node TSMC has made available, guaranteeing full order books as soon as the new technology comes online.
TSMC is running at 100% capacity on its mainstream 7/6/5/4nm processes, so this is not a huge drama, but may slow plans for the subsequent 1.6nm process expected in 2026.
Posted by: Pixy Misa at
05:48 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 473 words, total size 4 kb.
Friday, November 29
Tea Bean Edition
Top Story
- The price of tea beans is set to hit record highs after it (checks note) rained in Brazil. (Semafor)
This apparently is an unusual event that nobody was prepared for.
Tech News
- Need a faster network connection on your Thunderbolt / USB4 equipped laptop or mini-PC? The RaidenDigit LightONE might be what you need. (Serve the Home)
It is a general-purpose dock, with four USB-C ports (plus a fifth one purely for power), four USB-A ports, full-size and micro SD card slots, and a DisplayPort port. Where it stands out is that it also offers two 25Gbit Ethernet ports.
Yes, they're SFP, but while 25Gbase-T is a recognised standard using Cat-8 cables, there are approximately zero devices supporting it, so I can't really fault the designers there.
Only problem is that if you don't already have one of these you apparently can't get one. There was a Kickstarter, they delivered, the product worked, and then they disappeared.
- Microsoft is not using your Word and Excel data to train AI. (Bleeping Computer)
Okay.
- Australia has passed a law to ban children from social media. (The Register)
The government was repeatedly asked how this was supposed to work, and it had no idea. The legislation nonetheless passed with bipartisan support, because our nominal opposition party is as useless as our nominal government.
Posted by: Pixy Misa at
06:50 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 242 words, total size 2 kb.
59 queries taking 0.2383 seconds, 390 records returned.
Powered by Minx 1.1.6c-pink.